the new england journal of medicine - march 24, 2011 - vol. 364 no. 12 364 12

Perspective

The NEW ENGLAND JOURNAL of MEDICINE

march 24, 2011

n engl j med 364;12 nejm.org march 24, 2011 1085

controlled Senate. Although Re-publicans’ best chance to overturn the law won’t come until 2013 — and then only if they win a major-ity in both Congressional houses and the presidency in 2012 — they may meanwhile pursue tar-geted repeal of controversial pro-visions.

No provision is currently more beleaguered than the individual mandate to obtain health insur-ance or pay a penalty. Many ana-lysts view this mandate as crucial to ensuring that healthier people join state-based insurance ex-changes: since the law prohibits insurers from charging higher premiums to or turning away people with preexisting condi-tions, exchanges would other-

wise attract disproportionately sicker, costlier enrollees. That adverse selection would drive up premium costs and threaten ex-changes’ stability.

Reformers had hoped the man-date would also confer political advantages. Some Republicans, including former Massachusetts governor Mitt Romney, previously supported the policy. In a 2006 opinion piece, Romney defended Massachusetts’ decision to im-pose penalties on people who didn’t purchase insurance as a “personal responsibility principle. Some of my libertarian friends balk at what looks like an indi-vidual mandate,” he wrote. “But remember, someone has to pay for the health care that must, by

law, be provided: either the in-dividual pays or the taxpayers pay. A free ride on government is not libertarian.”1 The Massachu-setts mandate has in fact encour-aged healthy people to obtain coverage.2

But Massachusetts proved to be only an oasis of bipartisanship. Among Republicans in Washing-ton, pro-mandate arguments about personal responsibility gave way to concerns over individual lib-erty and the political priority of handing the Obama administra-tion a defeat.

The mandate now confronts a legal and political backlash. Flori-da’s Roger Vinson recently became the second federal judge to deem it unconstitutional, and the issue appears headed for the Supreme Court. Some state legislatures are seeking to block the mandate’s implementation. A few Democrat-ic senators, including Claire Mc-Caskill (MO) and Ben Nelson (NE),

Under Siege — The Individual Mandate for Health Insurance and Its AlternativesJonathan Oberlander, Ph.D.

The battle over the Patient Protection and Af-fordable Care Act (ACA) rages on: in January,

House Republicans passed legislation repealing the ACA, but the measure failed to clear the Democratic-

The New England Journal of Medicine Downloaded from nejm.org at MEDIZINISCHE UNIVERSITAT WIEN on March 26, 2011. For personal use only. No other uses without permission.

Copyright © 2011 Massachusetts Medical Society. All rights reserved.

PERSPECTIVE

n engl j med 364;12 nejm.org march 24, 20111086

who are up for reelection in 2012, say they’d like to find alterna-tives to the mandate. And while conservatives rail against the mandate and politically vulnera-ble moderates run away from it, many liberals support it reluctant-ly because of concerns that insur-ance remains unaffordable and that the requirement is a gift to the insurance industry.

Increasingly, Democrats may wonder whether the provision is an albatross that should be jetti-soned to save reform. Inasmuch as the health care reform debate is defined by the mandate, Demo-crats have a problem. The policy is highly unpopular — 76% of Americans view it unfavorably3 — and makes reform seem puni-tive. Although Democrats can highlight other consumer-friendly provisions that target health in-surers, in the mandate fight they are allied with the insurance in-dustry. Furthermore, President Barack Obama could in 2012 be in the uncomfortable position of defending a mandate that he ar-gued against in the 2008 Demo-cratic primary. (The mandate’s prominence also complicates Romney’s potential bid for the Republican presidential nomina-tion.)

Still, it’s not clear how much Democrats would gain politically by dumping the mandate. Many Republicans also oppose the ACA’s requirements that employers offer coverage or pay a penalty, the Medicaid expansion, increased government regulatory authority over the insurance industry, re-ductions in projected Medicare savings, and the cost of the sub-sidies for the uninsured to pur-chase insurance. Opponents of health care reform are attacking the mandate because it’s the most politically vulnerable part of the

law. Were it to fall, they would not endorse the ACA, but move on to attack other controversial provisions.

If the mandate cannot be sus-tained, alternatives exist. The country could adopt single-payer, tax-financed national health in-surance as a universal entitle-ment. There would be no need for penalties and no worries about adverse selection, since the gov-ernment would operate one risk pool. Enrollment would be much easier to administer and universal coverage would be ensured. Yet single payer would require com-pulsory participation, new taxes, and the political transformation necessary to displace private in-surance. It remains infeasible.

Staying within the reform’s ex-isting boundaries, one alternative is to limit enrollment in the ex-changes to a fixed period each year and impose premium penal-ties for eligible people who choose to wait and buy coverage later — and to make the penalty apply not just the first time they pur-chase insurance, but across their lifetimes.4 This model is employed by the Medicare Part D program for prescription-drug coverage, which the Bush administration and many Congressional Repub-licans supported.

However, to induce healthy un-insured people to sign up, the late penalty might have to be substan-tial, in which case this arrange-ment would be operating similar-ly to the mandate. As the health care economist Len Nichols points out, if we don’t have the political will to impose a strong penalty in conjunction with an individual mandate, we probably wouldn’t have the will to impose one as part of a fixed enrollment system. Furthermore, it’s not clear how a lifetime late penalty would work

when people’s insurance coverage could shift over time among pri-vate plans in the exchange, pri-vate plans outside the exchange, Medicaid, employer-based cover-age, and eventually, Medicare.

Another alternative would be to automatically enroll people in health insurance plans, permit-ting them to opt out. Auto-enroll-ment would occur primarily at the workplace but could also happen at state offices such as the Divi-sion of Motor Vehicles. This mod-el could have bipartisan appeal: a 2009 health care reform bill co-sponsored by Republican Senators Richard Burr (NC) and Tom Co-burn (OK) and Republican Con-gressmen Paul Ryan (WI) and Devin Nunes (CA) relied on auto-enrollment.5 Auto-enrollment can be combined with a premium penalty for people who opt out but later decide to purchase cover-age. Medicare Part B, which pays for physicians’ services, works this way.

Once again, though, the late-enrollment penalty might have to be substantial for auto-enrollment to effectively induce healthier peo-ple to pay for insurance coverage. Inertia alone may not be a suffi-ciently strong force to get younger, healthier workers to stay insured, given high and rising insurance premiums. Auto-enrollment may also not work as well outside the workplace or for workers’ depen-dents.4

Substituting either of these al-ternatives for the mandate would, as economist Jonathan Gruber ar-gues, attenuate the ACA’s bene-fits, resulting in an increased number of uninsured Americans and higher premiums in the nongroup insurance market (as healthier people decline cover-age).4 Implementing the rest of the ACA without any substitute

The Individual Mandate and Its Alternatives




PERSPECTIVE

1087

policy would similarly reduce cov-erage gains and destabilize insur-ance pools. The substantive case for the mandate is still strong, even if its political and legal foundations are shaken.

It’s also unclear whether past Republican support for alternative policies would be sustained or, as with the individual mandate, evap-orate in the heat of the political spotlight. Insofar as Republicans continue to support alternatives, they will be interested in them as part of a broader conservative health care reform package, not a means of bolstering the ACA. Republicans’ good chances of winning control of the Senate in 2012 further reduce their incen-tives to cooperate now on an alter-native plan.

What reformers may need even more than a policy alterna-tive to the mandate is an alter-

native rationale, especially since any provision meant to ensure broad participation in insurance pools must include financial penalties. The mandate’s de-fenders could again invoke the rhetoric of personal responsibil-ity. Or they could emphasize that the mandate makes possi-ble the insurance reforms that guarantee the availability of cov-erage to sick people.3 Alterna-tively, reformers could appeal to the mandate’s communitarian foundations, arguing that there are some public programs — such as Social Security and Medicare — that produce in-valuable social benefits and that succeed because everyone par-ticipates in them. Ultimately, the furor over the mandate un-derscores the reality that soli-darity remains elusive in U.S. health policy.

Disclosure forms provided by the author are available with the full text of this arti-cle at NEJM.org.

From the University of North Carolina, Cha-pel Hill.

This article (10.1056/NEJMp1101240) was published on February 16, 2011, at NEJM.org.

1. Romney M. Health care for everyone? We’ve found a way. Wall Street Journal. April 11, 2006.2. Chandra A, Gruber J, McKnight R. The im-portance of the individual mandate — evi-dence from Massachusetts. N Engl J Med 2011;364:293-5.3. Kaiser Family Foundation/Harvard School of Public Health. The public’s health care agenda for the 112th Congress. January 2011. (http://www.kff.org/kaiserpolls/upload/ 8134-F.pdf.)4. Gruber J. Health care reform without the individual mandate. Washington, DC: Center for American Progress, February 9, 2011. (http://www.americanprogress.org/issues/ 2011/02/gruber_mandate.html.)5. Coburn T. Individual auto-enrollment: an alternative to an individual mandate. (http://coburn.senate.gov/public/index.cfm?a=Files .Serve&File_id=e87f06bf-d429-4eac-8e7e-ade046b8b882.)Copyright © 2011 Massachusetts Medical Society.

The Individual Mandate and Its Alternatives

Early Accelerated Approval for Highly Targeted Cancer DrugsBruce A. Chabner, M.D.

The striking results of recent phase 1 trials of targeted can-

cer drugs have provoked serious discussion about shortening the road to drug approval. A typical cancer drug takes 7 years from entry into human trials to approv-al by the Food and Drug Admin-istration (FDA), which requires proof of efficacy in “well-con-trolled clinical trials.” In these randomized phase 3 trials, a new treatment alone or added to a drug combination is compared with a “standard” drug or com-bination. The most convincing end point for such trials is im-proved survival, although the FDA has accepted surrogates such as tumor progression, response rate,

or rarely, symptomatic relief. These trials are invaluable for es-tablishing the benefits of new drugs in instances where reason-able alternative therapies exist.

But phase 3 trials are expen-sive and time-consuming, usual-ly taking at least 2 to 3 years to reach survival end points. The news of a highly successful new compound in phase 1 or 2 rap-idly reaches physicians and pa-tients, creating demand for early access. For drugs aimed at dis-eases with limited effective treat-ments, delaying access during phase 2 and 3 trials creates dif-ficult ethical issues for regula-tory agencies and pharmaceutical sponsors and agonizing decisions

for physicians and patients. As Miller and Joffe recently empha-sized, the concept of equipoise (uncertainty as to which treat-ment carries greater benefit), underlies the ethical basis for randomization.1 Strong early re-sults in phase 1 or 2 may tip the balance and argue against delay-ing access by performing phase 3 trials, but Miller and Joffe con-tend that such trials may none-theless be justified by the need to establish new drugs’ long-term efficacy.

New understanding of the molecular and genetic lesions that cause cancer has sharpened the discussion, however, by enabling rapid development of drugs that



PERSPECTIVE


specifically target tumors bear-ing aberrant signaling pathways.2 Inhibitors of the BCR-ABL kinase in chronic myelogenous leuke-mia (CML) and inhibitors of the epidermal growth factor recep-tor (EGFR) in EGFR-mutated non–small-cell lung cancer (NSCLC)3 have replaced cytotoxic chemo-therapy for these indications. Two new drugs have produced excel-lent results in phase 1 trials against cancers that responded poorly to standard treatments: PLX4032 yielded an 81% response rate in 38 patients with BRAF mutated melanoma (see diagram),4and crizotinib had a 57% re-sponse rate in 82 patients with the EML4-ALK fusion in NSCLC.5Rates of disease control (response

or stable disease for at least 8 weeks) exceeded 90% in both trials, with minimal toxicity. Be-cause only a fraction of patients with NSCLC and melanoma have tumors harboring the aberrant pathways, a biomarker test for the mutations was required to select appropriate trial subjects.

Drugs such as PLX4032 and crizotinib normally undergo a sec-ond phase 2 trial (involving 30 to 100 patients) to confirm their activity as second- or third-line therapy. Because of their early promise, both of these drugs en-tered phase 3 trials directly after phase 1. Patients with newly di-agnosed BRAF-mutation–positive melanoma were randomly as-signed to PLX4032 or dacarba-

zine, a standard agent with a response rate of 15%; crossover from the control group to the experimental group at the time of disease progression was not allowed, to preserve the integrity of the survival end point. (The trial has now completed enroll-ment and reached its survival end point and is being amended to allow crossover.) In its phase 3 trial for newly diagnosed NSCLC, crizotinib is being compared with standard chemotherapy, which has a 30 to 40% response rate, a limited survival benefit, and sub-stantial toxicity. Patients in the control group who have disease progression may receive crizo-tinib by entering a phase 2 trial.

Because we can now define

Accelerated Approval for Targeted Cancer Drugs

MEK

ERK

Activation of PDGFR-β and IGF-1 receptor

Mutations in NRAS

Increased expression of the P13K–AKT signaling pathway

BRAF V600E — protein from mutated BRAF Increase

in CRAF

PLX4032 inhibits BRAF

Dimerization of BRAF and CRAF

Increased expression of COT

Mechanism of Action of PLX4032 in Melanoma.

BRAF dimerizes with its partner CRAF and acts as a relay point for signals for normal cell growth and survival. The BRAF V600E mutation, found in 60% of melanomas, signals constitutively and drives the malignant behavior of the tumor. PLX4032, an inhibitor of BRAF, kills BRAF-dependent tumors and has proved to be highly effective in phase 1 trials.1 Further mutations in BRAF or activat-ing mutations in NRAS, changes in expression of mitogen-activated protein kinase kinase kinase 8 (MAP3K8, or COT) or elements of the PI3 kinase pathway, or activation of alternative signaling pathways such as platelet-derived growth factor receptor β (PDGFR-β) have led to resistance to BRAF inhibition in experimental settings (labels in red) and in clinical settings (labels in green) and form the basis for new therapeutic strategies. IGF-1 denotes insulin-like growth factor 1. The image of CRAF was produced with the use of the Chimera software and is based on the 3omv crystal in the Protein Data Bank (www.wwpdb.org).




PERSPECTIVE

1089

patient subgroups with high re-sponse rates in phase 1 trials, performance of phase 3 trials for these drugs raises important is-sues. If patients with incurable disease who have the right bio-marker for response are informed of these impressive early results, they will want and perhaps deserve access to the new drug and may not accept random assignment to a modestly effective and toxic standard agent. The phase 3 trial may lack equipoise in the eyes of both physicians and patients.

There are alternatives for ex-pediting access to new drugs for patients who are not eligible for or accepted into phase 3 trials. A phase 2 trial may be opened simultaneously for patients with progressive disease after initial therapy. Such trials are open for both PLX4032 and crizotinib, but they accrue limited numbers of patients and are conducted in only a few centers.

In addition, compassionate-use protocols allow sponsors to dis-tribute experimental drugs on a case-by-case basis. This mecha-nism was widely used to provide early access to experimental AIDS drugs, with few reports of unex-pected adverse drug reactions. But sponsors often view compas-sionate use as competing with trial accrual and relieving pres-sure on the FDA for marketing approval. Neither crizotinib nor PLX4032 was available through an expanded-access protocol when its phase 3 trial opened, though such protocols are now available in a limited number of centers.

A mechanism for early FDA approval, “accelerated approval,” was introduced in 1992. Reserved for serious or life-threatening ill-ness not effectively treated by ap-proved medications, this mecha-nism allows drugs to be registered

on the basis of surrogate end points in phase 2 — most com-monly, a clearly demonstrable tu-mor response rate (often 20 to 30%). Accelerated approval is awarded with the stipulation that definitive trials, with a survival end point, must be conducted af-ter approval. Of the 23 oncologic drugs given accelerated approval between 1993 and 2008, two were ultimately withdrawn from the U.S. market — gemtuzumab because of toxicity and gefitinib because of lack of ef ficacy.

A 2008 study of accelerated approvals, from the Government Accountability Office, concluded that the FDA hasn’t effectively enforced requirements for post-approval trials. However, my re-view of accelerated approvals reveals that most such agents are integral to standard cancer treat-ment. Indeed, the more cogent question is whether accelerated approval can take place after phase 1.

Given trialists’ ability to define patient subgroups with responsive tumors in phase 1 trials, I pro-pose that for diseases lacking therapies that meaningfully ex-tend survival, the FDA should set flexible standards permitting ac-celerated approval of new drugs after phase 1. These standards could be satisfied by the results of expanded and targeted phase 1 drug testing, as in the case of PLX4032. This strategy requires the early evaluation and validation of a companion biomarker for patient selection. Conceivably, non-targeted agents could also satisfy criteria for approval after phase 1. High response rates (>50%), high disease-control rates (>75%), and an acceptable toxic-ity profile in a biomarker-defined population of 75 to 100 subjects should be sufficient for acceler-

ated approval if there’s a clear unmet need. Randomized com-parisons with minimally effec-tive treatments or placebo should not be required. Specific end points for early approval should be maximally flexible and ad-justed according to the targeted disease and the effectiveness and toxicity of alternative thera-pies. More extensive study would be required for less effective or more toxic experimental agents.

The post–phase-1 approval of new agents meeting these goals is highly unlikely to have signifi-cant negative consequences such as ineffective treatment or un-foreseen, overwhelming toxic ef-fects. Early approval would allow rapid general access to treatment, while further evaluation focused on defining optimal doses, sched-ules, and drug combinations; long-term benefits; toxic effects; and resistance mechanisms. When striking clinical results have been demonstrated in a sizable, readi-ly identifiable patient population in phase 1, the journey to drug approval should not be prolonged.


From the Massachusetts General Hospital Cancer Center, Boston.

1. Miller FG, Joffe S. Equipose and the di-lemma of randomized clinical trials. N Engl J Med 2011;364:476-80.2. McClellan M, Benner J, Schilsky R, et al. An accelerated pathway for targeted cancer therapies. Nat Rev Drug Discov 2011;10:79-80.3. Sequist LV, Martins RG, Spigel D, et al. First-line gefitinib in patients with advanced non-small-cell lung cancer harboring somat-ic EGFR mutations. J Clin Oncol 2008;26: 2442-9. [Erratum, J Clin Oncol 2008;26:3472.]4. Flaherty KT, Puzanov I, Kim KB, et al. Inhi-bition of mutated, activated BRAF in metastat-ic melanoma. N Engl J Med 2010;363:809-19.5. Kwak EL, Bang Y-J, Camidge DR, et al. Anaplastic lymphoma kinase inhibition in non–small-cell lung cancer. N Engl J Med 2010;363:1693-703. [Erratum, N Engl J Med 2011;364:588.]Copyright © 2011 Massachusetts Medical Society.

Accelerated Approval for Targeted Cancer Drugs



PERSPECTIVE


Tapping the Unmet Potential of Health Information TechnologyAnn S. O’Malley, M.D., M.P.H.

Tapping the Unmet Potential of Health IT

Health information technolo-gy (HIT) holds promise for

facilitating vast improvements in care and, ultimately, in the health of Americans,1,2 but achieving that potential remains a daunting task. A recent article in the Los An-geles Times described the new phe-nomenon of hiring computer-savvy undergraduate “scribes” to take notes for physicians during patient encounters and enter the informa-tion into electronic health records (EHRs) — a practice that suggests how far we must go to develop EHRs that clinicians will embrace. Of course, the most highly trained professional in the room need not be the one to enter data into the computer, especially during an emergency, but the perceived need for scribes and providers’ expe-riences using EHRs3 raise impor-tant questions about both the efficiency of care processes and the usability of current EHRs.

Although EHRs laudably pro-vide immediate access to patient data and electronic messaging functions, clinicians have been frustrated by the difficulty of us-ing them to support care delivery and coordination.3 Transforming EHRs into effective clinical tools rather than a means of captur-ing information primarily for doc-umentation and billing purpos-es will require progress on multiple fronts.

Clinical processes must evolve so as to improve care and be more responsive to patients’ needs, and HIT’s capabilities must evolve along with them. HIT has par-ticular potential in such areas as coordination of care, workflow efficiency and use of teams, clin-ical decision support, and popu-lation health management — all

areas offering glimpses of both the potential and the challenges associated with improved HIT use.

Few providers today, for exam-ple, can truly coordinate care — integrating care, in consultation with patients and their relatives and caregivers, across all of a patient’s conditions, needs, clini-cians, and health care settings.2-4 Outpatient practices and inpatient facilities lack well-developed pro-cesses for exchanging informa-tion, both within their own walls and during care transitions. Poor care coordination negatively af-fects patients — particularly those with multiple chronic conditions who account for an overwhelming proportion of U.S. health care expenditures.

HIT, especially if widely imple-mented, can facilitate coordina-tion by making information elec-tronically available at the point of care. As clinical care processes become more effective and effi-cient, they can inform new HIT capabilities that will better sup-port coordination. For example, providers need to develop consis-tent notification processes to en-sure timely communication about care transitions. Medical and nurs-ing professional societies could work with HIT vendors to devel-op standardized notification pro-cedures, which could be imple-mented through refined criteria for the “meaningful use” of HIT.5 HIT can also better support care coordination through the develop-ment of referral-tracking systems, improved approaches to reconcil-ing patients’ medications, and expansion of “problem list” capa-bilities — to avoid cluttering of lists with redundant informa-tion, for example, and permit

sorting and searching of lists and linking of listed problems to rele-vant portions of progress notes, assessments, and treatment plans.3

In addition, improving care coordination and health outcomes requires teamwork. Taking into account each team member’s train-ing, skill set, and expertise when delegating tasks and defining roles is critical to improving ef-ficiency, for both primary care teams within a practice and inter-specialty teams sharing patients’ care. HIT can support team-based care with tools enabling team members to identify patient care goals and document and monitor progress using a shared care plan. In outpatient practices, providers report that electronic messaging and notification of staff about patient care tasks facilitate com-munication about delegation and task completion. Electronic mes-saging can also enable real-time communication with specialists — if they’re on the same system — to determine whether a con-sultation is necessary.2,3

Coordinating care for patients with complex health conditions who see multiple physicians can also be supported by better HIT interoperability.3 The primary care team may be in the best position to coordinate a patient’s care, but it will often need information from other providers.4 Most cur-rent EHRs don’t adequately sup-port data exchange across pro-viders and settings, so practices communicate with outsiders pri-marily on paper.3 To support infor-mation exchange, EHRs must pre-sent data in standard ways, and separate organizations providing services for the same patient need to share information securely.




PERSPECTIVE

1091

The Health Information Tech-nology for Economic and Clini-cal Health (HITECH) Act envi-sions the secure exchange of data across providers and settings oc-curring through the Nationwide Health Information Network, which will provide a common platform and protocols.5 Some states and communities are de-veloping regional health informa-tion exchanges, but most are still in their infancy. It is hoped that these local and national efforts might eventually be linked. Some have raised concerns, however, about the sustainability of these exchanges absent a stronger busi-ness case to support them.

Clearly, HIT alone cannot transform our health care system: financial incentives must be re-aligned to reward patient-centered care. Current fee-for-service reim-bursement encourages EHR use for documentation of billable events rather than for tasks im-portant to the quality of care, such as coordination. Payment in-novations such as bundled pay-ments and accountable care or-ganizations aim to encourage providers to share accountability for outcomes. Such payment re-form would offer clinicians in-centives to demand HIT capabili-ties that better support the clinical tasks required to improve pa-tients’ health and would make these activities important to the success of health care organiza-tions.

Fee-for-service reimbursement also makes it difficult for clini-cians to take the time to listen to patients’ concerns. Unfortu-nately, adding an EHR to the clinical encounter can further distract clinicians from patients (which is one of the reasons that some emergency departments use scribes). To counter this tendency,

payment reform could be accom-panied by training for clinicians, residents, and medical students in effective communication with pa-tients in the presence of an EHR.

HIT can also provide tools to help inform decision making with regard to diagnosis (with clini-cal prediction rules), prevention (reminders), disease management (registries), and treatment (elec-tronic prescribing tools).2 The use of computerized medication orders, generated with the help of decision-support tools, is asso-ciated with reductions in adverse drug events.2 But most current commercial EHRs don’t provide or link to decision-support sys-tems, particularly for managing chronic care or selecting prefer-ence-sensitive treatments. To per-mit the development of adequate decision-support tools, the evi-dence base must be expanded and actively maintained.

Finally, primary care practices increasingly must focus not just on individual patients but on whole populations, as they strive to function as medical homes.2 A new orientation and effective methods for shifting practices from reactive and acute-symptom care to approaches including pro-active, planned care for both healthy and chronically ill popu-lations will be needed. Without HIT, it’s difficult to provide effec-tive population-based care and report quality metrics, but most commercial EHRs currently can-not help identify which patients in a population may need partic-ular services. Registries are an-other critical tool for population health management and an area where HIT applications could be better developed and integrated with EHRs.

Regardless of HIT’s potential advantages, clinicians in the coun-

try’s many small primary care practices can be overwhelmed by it and will need to be convinced that EHRs are affordable, enhance efficiency, and improve care. Then, they will need extensive, ongoing support. Under HITECH, HIT Re-gional Extension Centers will pro-vide technical assistance, guid-ance, and information on EHR adoption and meaningful use, particularly to such practices.5 These centers could also compile clinicians’ feedback for policy-makers and vendors, providing an infrastructure for shared learn-ing. Unfortunately, the United States currently lacks an adequate-ly trained workforce to support practices in these areas.

Continued research on clinical care processes, the design and use of HIT, and payment reform, as well as ongoing support for clinicians, will be key to the effec-tive and meaningful use of HIT. Today’s EHRs do not sufficiently support aspects of care delivery that are vital to improving care and controlling costs.


From the Center for Studying Health Sys-tem Change, Washington, DC.

1. Goldzweig CL, Towfigh A, Maglione M, Shekelle PG. Costs and benefits of health in-formation technology: new trends from the literature. Health Aff (Millwood) 2009;28: w282-w293.2. Bates DW, Bitton A. The future of health information technology in the patient-cen-tered medical home. Health Aff (Millwood) 2010;29:614-21.3. O’Malley AS, Grossman JM, Cohen GR, Kemper NM, Pham HH. Are electronic med-ical records helpful for care coordination? Experiences of physician practices. J Gen In-tern Med 2010;25:177-85.4. Stille CJ, Jerant A, Bell D, Meltzer D, Elmore JG. Coordinating care across diseases, set-tings, and clinicians: a key role for the gener-alist in practice. Ann Intern Med 2005;142: 700-8.5. Blumenthal D, Tavenner M. The “mean-ingful use” regulation for electronic health records. N Engl J Med 2010;363:501-4.Copyright © 2011 Massachusetts Medical Society.

Tapping the Unmet Potential of Health IT




The new england journal of medicineestablished in 1812 march 24, 2011 vol. 364 no. 12

Tiotropium versus Salmeterol for the Prevention of Exacerbations of COPD

Claus Vogelmeier, M.D., Bettina Hederer, M.D., Thomas Glaab, M.D., Hendrik Schmidt, Ph.D., Maureen P.M.H. Rutten-van Mölken, Ph.D., Kai M. Beeh, M.D., Klaus F. Rabe, M.D., and Leonardo M. Fabbri, M.D.,

for the POET-COPD Investigators*

A BS TR AC T

From the Hospital of the Universities of Giessen and Marburg, Marburg (C.V.); Boehringer Ingelheim, Ingelheim (B.H., T.G., H.S.); and insaf Respiratory Research Institute, Wiesbaden (K.M.B.) — all in Germany; the Institute for Medical Tech-nology Assessment (IMTA), Erasmus Uni-versity, Rotterdam (M.P.M.H.R.-M.); and Leiden University Medical Center, Leiden (K.F.R.) — both in the Netherlands; and the University of Modena and Reggio Emilia, Modena, Italy (L.M.F.). Address reprint requests to Dr. Fabbri at the Sec-tion of Respiratory Diseases, Department of Oncology, Hematology, and Pulmonary Diseases, University of Modena and Reg-gio Emilia, Policlinico di Modena, Largo del Pozzo 71, I-41124 Modena, Italy, or at [email protected].

*The investigators in the Prevention of Exacerbations with Tiotropium in COPD (POET-COPD) trial are listed in the Supplementary Appendix, available at NEJM.org.

N Engl J Med 2011;364:1093-103.Copyright © 2011 Massachusetts Medical Society.

BACKGROUNDTreatment guidelines recommend the use of inhaled long-acting bronchodilators to alleviate symptoms and reduce the risk of exacerbations in patients with moderate-to-very-severe chronic obstructive pulmonary disease (COPD) but do not specify whether a long-acting anticholinergic drug or a β2-agonist is the preferred agent. We investi-gated whether the anticholinergic drug tiotropium is superior to the β2-agonist salmeterol in preventing exacerbations of COPD.

METHODSIn a 1-year, randomized, double-blind, double-dummy, parallel-group trial, we com-pared the effect of treatment with 18 µg of tiotropium once daily with that of 50 µg of salmeterol twice daily on the incidence of moderate or severe exacerbations in patients with moderate-to-very-severe COPD and a history of exacerbations in the preceding year.

RESULTSA total of 7376 patients were randomly assigned to and treated with tiotropium (3707 patients) or salmeterol (3669 patients). Tiotropium, as compared with salme-terol, increased the time to the first exacerbation (187 days vs. 145 days), with a 17% reduction in risk (hazard ratio, 0.83; 95% confidence interval [CI], 0.77 to 0.90; P<0.001). Tiotropium also increased the time to the first severe exacerbation (haz-ard ratio, 0.72; 95% CI, 0.61 to 0.85; P<0.001), reduced the annual number of mod-erate or severe exacerbations (0.64 vs. 0.72; rate ratio, 0.89; 95% CI, 0.83 to 0.96; P = 0.002), and reduced the annual number of severe exacerbations (0.09 vs. 0.13; rate ratio, 0.73; 95% CI, 0.66 to 0.82; P<0.001). Overall, the incidence of serious adverse events and of adverse events leading to the discontinuation of treatment was similar in the two study groups. There were 64 deaths (1.7%) in the tiotropium group and 78 (2.1%) in the salmeterol group.

CONCLUSIONSThese results show that, in patients with moderate-to-very-severe COPD, tiotropium is more effective than salmeterol in preventing exacerbations. (Funded by Boehringer Ingelheim and Pfizer; ClinicalTrials.gov number, NCT00563381.)



T h e n e w e ngl a nd j o u r na l o f m e dic i n e


Chronic obstructive pulmonary dis-ease (COPD) is a leading cause of disabil-ity and death worldwide.1-3 Exacerbations

of COPD indicate instability or worsening of the patient’s clinical status and progression of the dis-ease and have been associated with the develop-ment of complications, an increased risk of sub-sequent exacerbations, a worsening of coexisting conditions, reduced health status and physical activity, deterioration of lung function, and an increased risk of death.4-7 The prevention of ex-acerbations therefore constitutes a major goal of treatment.1,2

Therapy with a long-acting anticholinergic drug or a long-acting β2-agonist is recommended as first-line maintenance therapy in patients with moderate-to-very-severe COPD,1,2 since both of these drugs reduce symptoms, improve quality of life and lung function, and reduce the risk of exacerbations and hospitalizations.8-12 However, treatment guidelines do not specify whether a long-acting anticholinergic drug or a β2-agonist is the preferred agent.1,2

Comparative studies have indicated that tiotro-pium is associated with a greater reduction in the risk of exacerbations and exacerbation-related hospitalizations than is salmeterol, although the differences were not significant.13,14 These were short-term studies (3 to 6 months in duration) and were not designed and powered to detect a difference in the risk of exacerbations. The Preven-tion of Exacerbations with Tiotropium in COPD (POET-COPD) trial was specifically designed to directly compare the effects of tiotropium with those of salmeterol on the risk of moderate and severe exacerbations. A placebo group was not in-cluded in the study, since there is substantial evi-dence of the superiority of both tiotropium and salmeterol over placebo.8,12 Furthermore, a com-parison of two active-treatment groups is in line with the recently growing relevance of compara-tive-effectiveness research to guidance regarding treatment decisions.15,16

Me thods

Study Design and OversightWe conducted a 1-year, randomized, double-blind, double-dummy, parallel-group trial at 725 centers in 25 countries to compare the effect of tiotropi-um (Spiriva, Boehringer Ingelheim) with that of salmeterol (Serevent, GlaxoSmithKline) on moder-

ate and severe exacerbations of COPD (hereinafter called exacerbations) in patients with moderate-to-very-severe COPD.17 The study was conducted in accordance with the provisions of the Declara-tion of Helsinki (1996) and Good Clinical Prac-tice guidelines. All patients provided written in-formed consent before any study procedure was performed. The scientific steering committee (which was made up of two of the academic in-vestigators and an external clinical researcher) and three employees of Boehringer Ingelheim developed the design and concept of the study, approved the statistical plan, had full access to the data, and interpreted the data. Onsite moni-toring and site management were supported by a contract research organization (PAREXEL). The first draft of the manuscript and subsequent revi-sions were written by all the authors, and all the authors made the decision to submit the manu-script for publication. The statistical analysis was performed by an employee of the sponsor. All the authors had full access to the data and vouch for the accuracy and completeness of the data and the analyses, as well as the fidelity of the study to the protocol. (The protocol, including the statistical analysis plan, is available with the full text of this article at NEJM.org.) An independent ethics com-mittee or institutional review board at each par-ticipating center reviewed and approved the pro-tocol before commencement of the study. In addition, an independent data and safety moni-toring board and a mortality adjudication com-mittee were established (Section 10 in the Sup-plementary Appendix, available at NEJM.org).

End PointsThe primary end point was the time to the first exacerbation of COPD. The time to the first exac-erbation was selected as the primary end point because it is less likely to be affected by the intro-duction of additional therapies or by the occur-rence of multiple exacerbations in some pa-tients.17 Secondary and safety end points included time-to-event end points, number-of-event end points, serious adverse events, and death (Section 2 in the Supplementary Appendix).

An exacerbation was defined as an increase in or new onset of more than one symptom of COPD (cough, sputum, wheezing, dyspnea, or chest tightness), with at least one symptom lasting 3 days or more and leading the patient’s attending physician to initiate treatment with



Tiotropium vs. Salmeterol for COPD Exacerbations


systemic glucocorticoids, antibiotics, or both (cri-terion for moderate exacerbation) or to hospital-ize the patient (criterion for severe exacerbation). The determination of the end of the exacerbation was made on the basis of the clinical assessment of the investigator. Data on exacerbations (ac-cording to the trial definition), as well as health care resources used to treat these exacerbations, were collected by means of a questionnaire that was administered during regular clinic visits and telephone contacts. When an investigator report-ed a case of pneumonia, he or she was questioned as to whether the event had been confirmed by imaging.

PatientsPatients were eligible for inclusion in the study if they were at least 40 years of age and had a smok-ing history of 10 pack-years or more, a diagnosis of COPD, a forced expiratory volume in 1 second (FEV1) after bronchodilation of ≤70% of the pre-dicted value,18 a ratio of FEV1 to forced vital ca-pacity (FVC) of ≤70%, and a documented history of at least one exacerbation leading to treatment with systemic glucocorticoids or antibiotics or hospitalization within the previous year. Spirom-etry (FEV1 and FVC) was performed at the screen-ing visit according to the guidelines of the Amer-ican Thoracic Society19 and was used only for the assessment of the severity of COPD. Postbron-chodilator measurements were performed 30 min-utes after the patient inhaled 400 µg of albu terol. Daily peak flow was recorded over the course of 4 months in a subgroup of patients, in conjunc-tion with a genotyping analysis (for details, see Section 5 in the Supplementary Appendix); those data are not reported here. Full details regarding the exclusion criteria are provided in Section 6 in the Supplementary Appendix.

ProceduresAfter a 2-week run-in period, eligible patients were randomly assigned to receive, for 1 year, either 18 µg of tiotropium once daily, delivered through the HandiHaler inhalation device (Boeh-ringer Ingelheim), plus placebo twice daily, deliv-ered through a pressurized, metered-dose inhal-er, or 50 µg of salmeterol twice daily, delivered through a pressurized, metered-dose inhaler, plus placebo once daily, delivered through the Handi-Haler device (for details, see Section 7 in the Supplementary Appendix). All the patients were

given instruction in the use of the HandiHaler and pressurized, metered-dose inhaler devices at visits 1 (screening) and 2 (randomization). Con-comitant medication at baseline was defined as the therapy the patients were receiving at the time of the screening visit (visit 1). During the run-in period, patients who were receiving tiotropium were required to switch to 40 µg of ipratropium four times a day, and this therapy was discontin-ued at the time of randomization. Patients who were receiving a long-acting β2-agonist were per-mitted to continue the use of that medication dur-ing the run-in period. Patients receiving fixed-dose combinations of long-acting β2-agonists and inhaled glucocorticoids were instructed to switch to inhaled glucocorticoid monotherapy at the start of the treatment phase of the study. Patients were allowed to continue their usual medications for COPD, except for anticholinergic drugs and long-acting β2-agonists, during the double-blind treat-ment phase.

After randomization, clinic visits were sched-uled at months 2, 4, 8, and 12, and monthly telephone calls were scheduled between visits. Patients completed a daily diary, and records were reviewed at each study visit to assess adherence to treatment and to determine whether respiratory symptoms met the criteria for exacerbation. Ad-herence was not systematically assessed during the trial. During clinic visits and monthly tele-phone calls, a questionnaire was administered to collect details regarding exacerbations of COPD. Adverse events leading to the discontinuation of treatment and serious adverse events including fatal events were recorded at the time of each clinic visit. Patients who prematurely discontin-ued treatment were followed for vital status (i.e., whether they were alive and, if they had died, the primary cause of death) until the end of the planned treatment period of 360 days. Informa-tion on vital status was considered to be complete for patients who attended all trial visits through day 360 and for those who prematurely discontin-ued study medication but whose vital status was confirmed at day 360. Details of the randomiza-tion procedures and of the procedures for con-cealing the treatment assignments are provided in Section 8 in the Supplementary Appendix.

Statistical AnalysisWe estimated that with a sample size of approxi-mately 6800 patients (3400 in each treatment





group), the study would have 80% power to detect a 10% reduction with tiotropium as compared with salmeterol in the risk of a first exacerbation, with a two-sided test for the null hypothesis of a hazard ratio of 1 at a significance level of 0.05. A prespecified reestimation of the sample size (with the treatment assignments concealed) on the ba-sis of the predicted event rate was performed to-ward the end of the originally planned recruit-ment phase and resulted in an increase of the sample size to a total of 7350 patients (Section 9 in the Supplementary Appendix).

The efficacy and safety analyses included all the patients who underwent randomization and who received at least one dose of the study medi-cation. Primary and secondary time-to-event end points were analyzed with the use of a Cox propor-tional-hazards regression model including terms for (pooled) center and treatment; pooling was performed to account for study centers that re-cruited fewer than four patients. P values were calculated with the use of the Wald chi-square statistic. Kaplan–Meier plots were constructed, and log-rank tests were also performed.

Number-of-event end points were compared between study groups with the use of Poisson re-gression with correction for overdispersion and adjustment for treatment exposure. To allow for a clear distinction between events, individual epi-sodes of exacerbations had to be separated by a gap of at least 7 days.

In keeping with the design of the study, exac-erbations were not systematically followed up af-ter a patient’s premature discontinuation of the trial medication.17 Hence, in the efficacy analysis, only exacerbations with onset during the time a patient was receiving treatment were included.7,20 Patients who withdrew from the trial prema-turely without having had an exacerbation were considered as having had no exacerbation, and in the time-to-event analysis, their data were cen-sored at the time of withdrawal. In the analyses of secondary end points, no corrections for mul-tiple testing have been made.

Subgroup analyses were performed for time-to-event end points and for number-of-event end points with the use of the models described above, with additional terms for subgroups and for interactions of subgroups with the study treat-ment. A post hoc subgroup analysis was per-formed according to patients who received in-haled glucocorticoids on a consistent basis during

the study treatment period versus patients who received no inhaled glucocorticoids during the treatment period. Incidence rates of serious ad-verse events were calculated as the number of patients with events divided by the time at risk. The rate of death from any cause was analyzed with the use of Cox regression, with treatment as a covariate. A Kaplan–Meier analysis was also performed.

R esult s

PatientsPatients were enrolled between January 2008 and April 2009. A total of 7384 patients underwent randomization, and 7376 patients (3707 in the tiotropium group and 3669 in the salmeterol group) received at least one dose of the study medication (Fig. 1). The baseline characteristics of the patients, including coexisting conditions, were balanced between the treatment groups (Table 1, and Section 11 in the Supplementary Ap-pendix). Fewer patients in the tiotropium group than in the salmeterol group withdrew from the study prematurely: 585 patients (15.8%) vs. 648 patients (17.7%) (hazard ratio with tiotropium, 0.88; 95% confidence interval [CI], 0.78 to 0.98; P = 0.02). The Kaplan–Meier plot for the time to the discontinuation of treatment is shown in Fig-ure 2A. The collection of vital status up to day 360 was complete for 99.1% of the patients.

ExacerbationsThere were 4411 individual episodes of exacerba-tion among 2691 patients; 44% of the patients with an exacerbation had moderate COPD at the trial onset (stage II COPD, according to the clas-sification of the Global Initiative for Chronic Ob-structive Lung Disease [GOLD],1 which specifies four stages of COPD ranging from stage I, indi-cating mild disease, to stage IV, indicating very severe disease). The time to the first exacerbation (the primary end point) was increased by 42 days with tiotropium as compared with salmeterol (187 days vs. 145 days, representing the time un-til at least 25% of the patients [first quartile] had a first exacerbation), corresponding to a 17% re-duction in risk with tiotropium (hazard ratio, 0.83; 95% CI, 0.77 to 0.90; P<0.001). Figure 2B shows the Kaplan–Meier plot for the time to the first exacerbation. Given the fact that less than 50% of the patients had an exacerbation (2691 of





7376 patients [36.5%]), it was not possible to cal-culate the median time to the first exacerbation; therefore, the time to the first exacerbation in the first quartile of patients was calculated instead.

Tiotropium as compared with salmeterol sig-nificantly reduced the risk of moderate exacer-bations by 14% (hazard ratio, 0.86; 95% CI, 0.79 to 0.93; P<0.001) and of severe exacerbations by 28% (hazard ratio, 0.72; 95% CI, 0.61 to 0.85; P<0.001). The Kaplan–Meier plot for the time to a first severe exacerbation is shown in Figure 2C. In addition, tiotropium reduced the risk of exac-erbations leading to treatment with systemic glu-cocorticoids by 23% (hazard ratio, 0.77; 95% CI, 0.69 to 0.85; P<0.001), exacerbations leading to treatment with antibiotics by 15% (hazard ratio, 0.85; 95% CI, 0.78 to 0.92; P<0.001), and exacer-bations leading to treatment with both systemic

glucocorticoids and antibiotics by 24% (hazard ratio, 0.76; 95% CI, 0.68 to 0.86; P<0.001) (Section 3 in the Supplementary Appendix).

The annual rate of exacerbations was 0.64 in the tiotropium group and 0.72 in the salmeterol group, corresponding to an 11% reduction in the rate of exacerbations with tiotropium (rate ratio, 0.89; 95% CI, 0.83 to 0.96; P = 0.002). Treatment with tiotropium significantly reduced the annual rate of moderate exacerbations by 7% (0.54 vs. 0.59; rate ratio, 0.93; 95% CI, 0.86 to 1.00; P = 0.048) and the annual rate of severe exacerba-tions by 27% (0.09 vs. 0.13; rate ratio, 0.73; 95% CI, 0.66 to 0.82; P<0.001) (Section 3 in the Supple-mentary Appendix). In addition, tiotropium re-duced the rate of exacerbations leading to treat-ment with systemic glucocorticoids by 18% (0.33 vs. 0.41; rate ratio, 0.82; 95% CI, 0.76 to 0.90;

7384 Underwent randomization

8293 Patients were assessed for eligibility

909 Withdrew during screeningor did not meet entry criteria

3711 Were assigned to receivetiotropium

3673 Were assigned to receivesalmeterol

648 Discontinued salmeterol292 Had adverse event24 Had lack of efficacy

209 Withdrew consent74 Were nonadherent

to protocol15 Were lost to follow-up34 Had other reason

585 Discontinued tiotropium264 Had adverse event32 Had lack of efficacy

192 Withdrew consent66 Were nonadherent

to protocol7 Were lost to follow-up

24 Had other reason

4 Did not receive salmeterol4 Did not receive tiotropium

3122 Completed study 3021 Completed study

3707 Were included in efficacyand safety analyses

3669 Were included in efficacyand safety analyses

Figure 1. Screening, Randomization, and Follow-up.





P<0.001), exacerbations leading to treatment with antibiotics by 10% (0.53 vs. 0.59; rate ratio, 0.90; 95% CI, 0.84 to 0.97; P = 0.004), and exacerbations

leading to treatment with both systemic gluco-corticoids and antibiotics by 20% (0.23 vs. 0.28; rate ratio, 0.80; 95% CI, 0.73 to 0.88; P<0.001) (Section 3 in the Supplementary Appendix).

The effects of tiotropium as compared with salmeterol on the time to a first exacerbation and the annual rate of exacerbations per patient were consistent across prespecified subgroups accord-ing to age, sex, smoking status (current vs. non-current smoker), severity of COPD (GOLD stage), body-mass index, and use or no use of inhaled glucocorticoids at baseline (Fig. 3, and Section 4 in the Supplementary Appendix). Patients with a low body-mass index or very severe COPD seemed to benefit most from tiotropium therapy (Fig. 3). However, the P values for the tests of an interac-tion between treatment effect and subgroup were 0.17 for the subgroup according to body-mass index and 0.05 for the subgroup according to GOLD stage. In a post hoc analysis, a similar re-duction in the risk of an exacerbation with tiotro-pium as compared with salmeterol was observed among the 2932 patients who used concomitant inhaled glucocorticoids during the study-treat-ment period (hazard ratio, 0.91; 95% CI, 0.82 to 1.02), as well as among the 4046 patients who did not use inhaled glucocorticoids at any time dur-ing the study-treatment period (hazard ratio, 0.81; 95% CI, 0.72 to 0.91). In a subgroup analysis of patients who were receiving inhaled glucocor-ticoids at baseline but did not receive them during the study-treatment period versus patients who were receiving inhaled glucocorticoids at baseline and continued to receive them during the study-treatment period, the annual exacerbation rate in the tiotropium group was 0.67 (95% CI, 0.57 to 0.79) among the 395 patients who discontinued the use of inhaled glucocorticoids, as compared with 0.78 (95% CI, 0.73 to 0.85) among the 1452

Table 1. Baseline Characteristics of the Patients.*

CharacteristicTiotropium (N = 3707)

Salmeterol (N = 3669)

Male sex (%) 74.4 74.9

Age (yr) 62.9±9.0 62.8±9.0

Smoking status

Current smoker (%) 48.0 48.3

Smoking history (pack-yr) 38.8±20.0 37.8±19.2

Duration of COPD (yr)† 8.0±6.7 7.9±6.5

GOLD stage (%)‡

II 47.8 49.6

III 43.1 42.1

IV 8.9 7.9

Spirometry after bronchodilation§

FEV1 (liters) 1.41±0.47 1.41±0.45

FEV1 (% of predicted value) 49.2±13.3 49.4±13.1

FVC (liters) 2.71±0.81 2.75±0.82

Ratio of FEV1 to FVC (%) 52.5±10.8 52.4±11.2

Pulmonary medications (%)

Any 90.0 89.9

Anticholinergic drug

Tiotropium 30.5 30.3

Short-acting 29.3 29.6

β2-Agonists

Long-acting¶ 51.5 51.5

Short-acting 52.5 53.4

Glucocorticoids

Inhaled¶ 53.6 53.3

With tiotropium 18.7 18.2

With long-acting β2-agonists 43.3 43.5

Oral 2.4 2.3

Methylxanthines 23.0 21.2

* Plus–minus values are means ±SD. COPD denotes chronic obstructive pul-monary disease, FEV1 forced expiratory volume in 1 second, and FVC forced vital capacity.

† Data on duration of COPD were missing for 15 patients in the tiotropium group and 5 in the salmeterol group.

‡ The severity of COPD was defined according to the classification of the Global Initiative for Chronic Obstructive Lung Disease (GOLD), which specifies four stages of COPD ranging from stage I, indicating mild disease, to stage IV, indi-cating very severe disease. There were 23 patients with GOLD stage I COPD — 0.2% of the patients in the tiotropium group and 0.4% in the salmeterol group.

§ Pulmonary function testing was performed at the screening visit (visit 1). Data on FVC were missing for 1 patient in the tiotropium group.

¶ This medication was used either alone or in a fixed combination.

Figure 2 (facing page). Kaplan–Meier Curves for the Primary and Selected Secondary Outcomes.

Kaplan–Meier curves are shown for the probability of premature discontinuation of the study medication (Panel A), the probability of a first exacerbation of chronic obstructive pulmonary disease (COPD) (Panel B), and the probability of a first severe exacerbation of COPD leading to hospitalization (Panel C) in the tio-tropium and salmeterol groups. The hazard ratios are based on a Cox proportional-hazards regression model including terms for (pooled) center and treatment. CI denotes confidence interval.





B

A

Prob

abili

ty o

f Pre

mat

ure

Dis

cont

inua

tion

of T

rial M

edic

atio

n

1.0

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 210 240 270 300 330 36030 60 90 120 150 180

Days

Hazard ratio, 0.88 (95% CI, 0.78–0.98)P=0.02 by log-rank test

0.20

0.15

0.05

0.10

0.000 21024027030033036030 60 90 120150180

No. at RiskTiotropiumSalmeterol

37073669

32683151

32993181

33303209

33823291

34293337

35013436

35923541

32253111

31863074

31583054

31383037

28412703

Prob

abili

ty o

f CO

PD E

xace

rbat

ion

1.0

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 210 240 270 300 330 36030 60 90 120 150 180

Days

Hazard ratio, 0.83 (95% CI, 0.77–0.90)P<0.001 by log-rank test

0.45

0.30

0.10

0.20

0.350.40

0.25

0.05

0.15

0.000 21024027030033036030 60 90 120150180


37073669

24552251

25612351

26472457

27872605

29552802

31363028

33693328

23432137

22422050

21691982

21071915

18691657

C

Prob

abili

ty o

f Sev

ere

CO

PDEx

acer

batio

n

1.0

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 210 240 270 300 330 36030 60 90 120 150 180

Days

Hazard ratio, 0.72 (95% CI, 0.61–0.85)P<0.001 by log-rank test

0.15

0.05

0.10

0.000 21024027030033036030 60 90 120150180


37073669

31252982

31773032

32173080

32853172

33593244

34533362

35643502

30662921

30172870

29772834

29482806

26632489

Tiotropium

Salmeterol

TiotropiumSalmeterol

Tiotropium

Salmeterol





patients who continued to receive them; the an-nual exacerbation rate in the salmeterol group was 0.86 (95% CI, 0.74 to 0.99) among the 416 patients who discontinued the use of inhaled glucocorticoids, as compared with 0.81 (95% CI, 0.75 to 0.88) among the 1401 patients who con-tinued to receive them.

SAFETYA total of 545 patients (14.7%) in the tiotropium group and 606 (16.5%) in the salmeterol group reported a serious adverse event during the study-treatment period (Table 2). The most common serious adverse event with a frequency of 0.5% or greater was an exacerbation of COPD, which oc-curred in 270 patients (7.3%) in the tiotropium

group and in 335 (9.1%) in the salmeterol group (Section 12 in the Supplementary Appendix).

A total of 180 cases of pneumonia were re-ported, of which 158 (87.8%) were radiologically confirmed (70 in the tiotropium group and 88 in the salmeterol group). There were more patients with at least one radiologically confirmed episode of pneumonia among those who received con-comitant medication with inhaled glucocorti-coids for at least 1 day during the study-treatment period than among those who received no in-haled glucocorticoid during the study-treatment period — 89 of 3330 patients (2.7%), of whom 72 required hospitalization, as compared with 59 of 4046 patients (1.5%), of whom 46 required hospitalization.

0.6 0.8 1.0 1.41.2

SalmeterolBetter

Tiotropium Better

Age (yr)<55≥55 to <65≥65

SexMaleFemale

COPD severity stage (GOLD)Stage IIStage IIIStage IV

Smoking statusNoncurrent smokerCurrent smoker

BMI<20≥20 to <25≥25 to <30≥30

Use of inhaled glucocorticoidsat baseline

YesNo

Tiotropium Hazard Ratio (95% CI)P Value forInteractionSalmeterolSubgroup

0.82 (0.73–0.92)0.87 (0.79–0.96)

0.85 (0.72–1.00)

0.66 (0.51–0.85)0.89 (0.79–1.02)0.87 (0.76–0.99)

0.87 (0.78–0.97)0.84 (0.75–0.93)

0.64 (0.50–0.81)0.86 (0.77–0.97)0.88 (0.79–0.99)

0.84 (0.73–0.97)0.86 (0.78–0.94)

0.83 (0.74–0.93)

0.88 (0.74–1.05)

0.4

0.87 (0.77–0.98)258/665 522/1426634/1578

1016/2747398/922

635/1833627/1545152/291

746/1896668/1773

134/271 501/1254468/1284311/860

839/1955575/1714

0.76

0.83

0.05

0.64

0.17

0.41

237/655 484/1462556/1590

913/2759364/948

561/1781589/1597127/329

678/1929599/1778

105/286 455/1230424/1276293/915

785/1986492/1721

no. of patients/total no.

Figure 3. Subgroup Analysis of the Primary End Point.

The number of patients who had at least one exacerbation of chronic obstructive pulmonary disease (COPD) with onset during the study-treatment period is shown according to subgroup. Hazard ratios were calculated with the use of Cox regression with terms for treatment. Horizontal lines represent 95% confidence intervals. The size of the squares is proportional to the size of the subgroup. The severity of COPD was defined according to the classification of the Global Initiative for Chronic Obstructive Lung Disease (GOLD), which specifies four stages of COPD ranging from stage I, indicating mild disease, to stage IV, indicating very severe disease. Noncur-rent smokers included former smokers and one person who had never smoked. The body-mass index (BMI) is the weight in kilograms divided by the square of the height in meters.





There were 142 deaths during the planned treatment period of 360 days (including deaths among patients who had withdrawn from the study prematurely and whose vital status was re-corded at 360 days): 64 in the tiotropium group and 78 in the salmeterol group (hazard ratio with tiotropium, 0.81; 95% CI, 0.58 to 1.13). Addi-tional information is provided in Section 13 in the Supplementary Appendix.

Discussion

Tiotropium, as compared with salmeterol, signifi-cantly increased the time to the first moderate or severe exacerbation of COPD and significantly de-creased the annual rate of exacerbations among patients with moderate-to-very-severe COPD. The benefit with tiotropium was seen consistently in all the major subgroups that were considered in this trial and was independent of the concomi-tant use of inhaled glucocorticoids.

This 1-year study was designed and powered for the end point of moderate and severe exacer-bations, one of the most relevant patient-related outcomes, with important effects on patients’ families, caregivers, health care providers, and payers.4-6 Any exacerbation that can be avoided

would be beneficial from the perspective of both the patient and the health care system and consti-tutes a major treatment goal in COPD.1,2

Previous large, long-term trials have shown that both salmeterol and tiotropium reduce the rate of exacerbations.8,12 However, to date, there has been insufficient evidence from direct compari-sons of the two drugs; therefore, current guide-lines do not favor one long-acting agent over the other for patients with COPD.1,2

The Kaplan–Meier analyses of the time to the first exacerbation show that the benefit with tio-tropium as compared with salmeterol became evident as early as approximately 1 month after the initiation of treatment and was maintained over the entire 1-year study period. Thus, it ap-pears to be unlikely that the difference in favor of tiotropium was due to early discontinuation of treatment among patients in the salmeterol group who did not have a response to that drug. Tiotro-pium and salmeterol have been shown to reduce airflow limitation and hyperinflation but may also directly or indirectly have an effect on various aspects of lung inflammation.21,22 However, the relevance of these mechanisms to the observed differences in the end points related to exacerba-tions remains to be determined. Whether the

Table 2. Incidence Rates of Serious Adverse Events, According to System Organ Class.*

Serious Adverse Events Tiotropium (N = 3707) Salmeterol (N = 3669)

Rate Ratio for Tiotropium vs.

Salmeterol(95% CI)

no. (%)rate/100 patient-yr no. (%)

rate/100 patient-yr

Respiratory, thoracic, and mediastinal events 300 (8.1) 8.66 366 (10.0) 10.99 0.79 (0.68–0.92)

Infections 96 (2.6) 2.69 109 (3.0) 3.15 0.85 (0.65–1.12)

Cardiac events 98 (2.6) 2.73 85 (2.3) 2.44 1.12 (0.84–1.50)

Neoplasms 51 (1.4) 1.42 43 (1.2) 1.23 1.15 (0.77–1.73)

Vascular events 37 (1.0) 1.03 25 (0.7) 0.71 1.44 (0.87–2.39)

Gastrointestinal events 32 (0.9) 0.89 32 (0.9) 0.92 0.97 (0.59–1.58)

Nervous system events 28 (0.8) 0.78 29 (0.8) 0.83 0.94 (0.56–1.58)

General events† 16 (0.4) 0.44 27 (0.7) 0.77 0.57 (0.31–1.07)

Injury, poisoning, and procedural complications 22 (0.6) 0.61 19 (0.5) 0.54 1.13 (0.61–2.08)

Musculoskeletal and connective-tissue events 10 (0.3) 0.28 22 (0.6) 0.63 0.44 (0.21–0.93)

* Listed are incidence rates per 100 patient-years and incidence rate ratios of serious adverse events that occurred from the beginning of the study-treatment period until 30 days after the last dose of study drug was received and that were reported by at least 0.5% of the patients in either study group. The adverse events are categorized according to the sys-tem organ classes in the Medical Dictionary for Regulatory Activities.

† This category includes the diagnostic terms “death” and “sudden death.”





observed differences might be due to differences in the aerosolizing systems, the particle size of the aerosols, or the distribution of the drug in the lung is also unknown.

The annual exacerbation rates in this study were lower than those in large trials involving patients with COPD, such as the Trial of In-haled Steroids and Long-acting β2 Agonists (TRISTAN)23 and the Towards a Revolution in COPD Health trial (TORCH; ClinicalTrials.gov number, NCT00268216),8 were similar to those in the Understanding Potential Long-Term Impacts on Function with Tiotropium trial (UPLIFT, NCT00144339),12 and were higher than those in a recent 1-year study comparing the efficacy of two long-acting β2-agonists.24 This variability may reflect differences in inclusion criteria and in the concomitant medications, such as inhaled glucocorticoids, that patients were allowed to re-ceive. In our trial, consistent with current guide-line recommendations, concomitant therapy with inhaled glucocorticoids was allowed but was not mandatory, because the patient population in-cluded a substantial proportion of patients with moderate COPD (GOLD stage II). Approximately 40% of the patients received concomitant thera-py with inhaled glucocorticoids on a consistent basis during the study-treatment period. In a post hoc analysis, treatment with tiotropium decreased the risk of exacerbations more than did treatment with salmeterol both in patients who were receiv-ing inhaled glucocorticoids and in those who were not receiving them, suggesting that the ben-efit of tiotropium was independent of the use of inhaled glucocorticoids.

In addition, the rate of exacerbations among patients in the tiotropium group who were re-ceiving inhaled glucocorticoids at baseline but did not continue receiving them during the study-treatment period was not higher than the rate among those who were receiving inhaled gluco-corticoids at baseline and continued to receive them during the study-treatment period. This find-ing is consistent with the results of the COPD and Seretide: a Multi-Center Intervention and Characterization (COSMIC) study, which showed that withdrawal of fluticasone for 1 year after a 3-month run-in period with a fixed combination of fluticasone and salmeterol was not associated with an increase in moderate or severe exacer-bations.25

Differences between study groups in the pro-portion of patients discontinuing the study treat-

ment have been seen in other studies involving patients with COPD and are most often attrib-uted to relative differences in the efficacy, safety, or both of the agents used in the study.7,12,26,27 Similarly, we observed a significantly higher rate of premature discontinuation of treatment in the salmeterol group than in the tiotropium group. However, as compared with the between-group differences that have been seen in placebo-con-trolled studies, the absolute difference was quite small (1.9 percentage points).

Both tiotropium and salmeterol have safety profiles that have been well described in the literature.28-31 Overall, the incidence of serious adverse events, adverse events leading to treat-ment discontinuation, and fatal events were simi-lar across treatments.

In summary, among patients with moderate-to-very-severe COPD and a history of exacerba-tion, tiotropium was more effective than salme-terol in all the exacerbation end points that were assessed and across all major subgroups. The results of this large trial provide data on which to base the choice of long-acting bronchodilator therapy for maintenance treatment of COPD.

Dr. Vogelmeier reports receiving consulting fees and travel support from Boehringer Ingelheim, payment for board mem-bership from AstraZeneca, Boehringer Ingelheim, GlaxoSmith-Kline, Mundipharma, Novartis, and Nycomed, fees for expert testimony and grant support from Talecris Biotherapeutics, and lecture fees from AstraZeneca, Boehringer Ingelheim, Glaxo-SmithKline, Janssen, Merck, Novartis, Nycomed, and Talecris Biotherapeutics; Dr. Hederer, being formerly employed by Boeh-ringer Ingelheim; Drs. Glaab and Schmidt, being currently em-ployed by Boehringer Ingelheim; Dr. Rutten-van Mölken, receiv-ing grant support from Boehringer Ingelheim and having consulting fees, travel support, and fees for participation in re-view activities from Boehringer Ingelheim paid to her institution on her behalf; Dr. Beeh, having consulting fees or honoraria, grant support, and payments for development of educational materials from Boehringer Ingelheim and grant support from Almirall, Novartis, Mundipharma, Cytos Biotechnology, GlaxoSmithKline, Pfizer, and Revotar Biopharmaceuticals paid to his institution on his behalf; Dr. Rabe, receiving consulting fees and honoraria and payment for board membership from AstraZeneca, Boeh-ringer Ingelheim, Chiesi Farmaceutici, GlaxoSmithKline, Merck, Novartis, Nycomed, and Pfizer and grant support from Altana, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Novartis, and Roche; and Dr. Fabbri, receiving travel support from Boeh-ringer Ingelheim, consulting fees from AstraZeneca, Boehringer Ingelheim, Chiesi Farmaceutici, GlaxoSmithKline, Merck, No-vartis, Nycomed, Pfizer, and Sigma-Tau Pharmaceuticals, and lecture fees from AstraZeneca, Boehringer Ingelheim, Chiesi Farmaceutici, GlaxoSmithKline, Merck, Novartis, Nycomed, and Pfizer and having grant support from AstraZeneca, Boehringer Ingelheim, Chiesi Farmaceutici, GlaxoSmithKline, Menarini, Merck, Nycomed, Pfizer, Schering-Plough (now Merck), Sigma-Tau Pharmaceuticals, and Union Chimique Belge paid to his in-stitution on his behalf. No other potential conflict of interest relevant to this article was reported.

Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.





We thank the patients who participated in the trial; Dr. Klaus Fichtner, Vera Drews, and Nicole Bader for administrative trial support; Dr. Steven Kesten, Dr. Fee Rühmkorf, Dr. Harald Kögler, Dr. Susanne Stowasser, Dr. Brigitta Monz, and Dagmar Selim for scientific advice; Dr. Inge Leimer and Achim Müller for statisti-cal advice; Michael Betke-Hornfeck for technical trial support; Christine Meissner and Christina Raabe for data management

support; Bernd Damian and Rafal Falkowski for programming support; Dr. Silke Wienecke, Declan Tobin, and Karen Ryan from Parexel for onsite monitoring and site management support; Natalie Barker and Claire Scarborough, from Parexel MedCom, for editorial and technical support in the preparation of the manuscript; and Dr. Franklin Cerasoli, Dr. Idelle Weisman, and Dr. Lalitha Aiyer for review of the manuscript.

References1. Global Initiative for Chronic Obstruc-tive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: 2009 update. (http://www.goldcopd.com.)2. National Institute for Health and Clinical Excellence. Chronic obstructive pulmonary disease: management of chronic obstructive pulmonary disease in adults in primary and secondary care. London: National Clinical Guideline Cen-tre, 2010. (http://guidance.nice.org.uk/CG101/Guidance/pdf/English.)3. Niewoehner DE. Outpatient manage-ment of severe COPD. N Engl J Med 2010; 362:1407-16.4. Donaldson GC, Seemungal TAR, Bhowmik A, Wedzicha JA. Relationship between exacerbation frequency and lung function decline in chronic obstructive pulmonary disease. Thorax 2002;57:847-52. [Erratum, Thorax 2008;63:753.]5. Soler-Cataluña JJ, Martínez-García MÁ, Román Sánchez P, Salcedo E, Navar-ro M, Ochando R. Severe acute exacerba-tions and mortality in patients with chronic obstructive pulmonary disease. Thorax 2005;60:925-31.6. Wedzicha JA, Seemungal TA. COPD exacerbations: defining their cause and prevention. Lancet 2007;370:786-96.7. Hurst JR, Vestbo J, Anzueto A, et al. Susceptibility to exacerbation in chronic obstructive pulmonary disease. N Engl J Med 2010;363:1128-38.8. Calverley PMA, Anderson JA, Celli B, et al. Salmeterol and fluticasone propio-nate and survival in chronic obstructive pulmonary disease. N Engl J Med 2007; 356:775-89.9. Casaburi R, Mahler DA, Jones PW, et al. A long-term evaluation of once-daily inhaled tiotropium in chronic obstructive pulmonary disease. Eur Respir J 2002;19: 217-24.10. Mahler DA, Donohue JF, Barbee RA, et al. Efficacy of salmeterol xinafoate in the treatment of COPD. Chest 1999;115: 957-65.11. Rennard SI, Anderson W, ZuWallack R, et al. Use of a long-acting inhaled β2-adrenergic agonist, salmeterol xinafoate,

in patients with chronic obstructive pul-monary disease. Am J Respir Crit Care Med 2001;163:1087-92.12. Tashkin DP, Celli B, Senn S, et al. A 4-year trial of tiotropium in chronic obstructive pulmonary disease. N Engl J Med 2008;359:1543-54.13. Briggs DD Jr, Covelli H, Lapidus R, Bhattycharya S, Kesten S, Cassino C. Im-proved daytime spirometric efficacy of tiotropium compared with salmeterol in patients with COPD. Pulm Pharmacol Ther 2005;18:397-404.14. Brusasco V, Hodder R, Miravitlles M, Korducki L, Towse L, Kesten S. Health outcomes following treatment for six months with once daily tiotropium com-pared with twice daily salmeterol in pa-tients with COPD. Thorax 2003;58:399-404. [Erratum, Thorax 2005;60:105.]15. Mushlin AI, Ghomrawi H. Health care reform and the need for compara-tive-effectiveness research. N Engl J Med 2010;362(3):e6. (Available at NEJM.org.)16. Sawicki PT. Communal responsibility for health care — the example of benefit assessment in Germany. N Engl J Med 2009;361(20):e42. (Available at NEJM.org.)17. Beeh KM, Hederer B, Glaab T, et al. Study design considerations in a large COPD trial comparing effects of tiotropi-um with salmeterol on exacerbations. Int J Chron Obstruct Pulmon Dis 2009;4:119-25.18. Quanjer PH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. Lung volumes and forced ventilatory flows: Re-port Working Party Standardization of Lung Function Tests, European Commu-nity for Steel and Coal: official statement of the European Respiratory Society. Eur Respir J Suppl 1993;16:5-40.19. American Thoracic Society. Standard-ization of spirometry, 1994 update. Am J Respir Crit Care Med 1995;152:1107-36.20. Keene ON, Vestbo J, Anderson JA, et al. Methods for therapeutic trials in COPD: lessons from the TORCH trial. Eur Respir J 2009;34:1018-23.21. Bateman ED, Rennard S, Barnes PJ, et al. Alternative mechanisms for tiotropium. Pulm Pharmacol Ther 2009;22:533-42.

22. Johnson M, Rennard S. Alternative mechanisms for long-acting beta(2)- adrenergic agonists in COPD. Chest 2001; 120:258-70.23. Calverley P, Pauwels R, Vestbo J, et al. Combined salmeterol and fluticasone in the treatment of chronic obstructive pul-monary disease: a randomised controlled trial. Lancet 2003;361:449-56. [Erratum, Lancet 2003;361:1660.]24. Dahl R, Chung KF, Buhl R, et al. Ef-ficacy of a new once-daily long-acting in-haled beta2-agonist indacaterol versus twice-daily formoterol in COPD. Thorax 2010;65:473-9.25. Wouters EF, Postma DS, Fokkens B, et al. Withdrawal of fluticasone propionate from combined salmeterol/f luticasone treatment in patients with COPD causes immediate and sustained disease deterio-ration: a randomised controlled trial. Thorax 2005;60:480-7.26. Calverley PM, Spencer S, Willits L, Burge PS, Jones PW. Withdrawal from treatment as an outcome in the ISOLDE study of COPD. Chest 2003;124:1350-6.27. Niewoehner DE, Rice K, Cote C, et al. Prevention of exacerbations of chronic obstructive pulmonary disease with tiotropium, a once-daily inhaled anticho-linergic bronchodilator: a randomized trial. Ann Intern Med 2005;143:317-26.28. Celli B, Decramer M, Leimer I, Vogel U, Kesten S, Tashkin DP. Cardiovascular safety of tiotropium in patients with COPD. Chest 2010;137:20-30.29. Kesten S, Celli B, Decramer M, Leimer I, Tashkin D. Tiotropium HandiHaler in the treatment of COPD: a safety review. Int J Chron Obstruct Pulmon Dis 2009; 4:397-409.30. Ferguson GT, Funck-Brentano C, Fischer T, Darken P, Reisner C. Cardiovas-cular safety of salmeterol in COPD. Chest 2003;123:1817-24.31. Calverley PMA, Anderson JA, Celli B, et al. Cardiovascular events in patients with COPD: TORCH study results. Thorax 2010;65:719-25.Copyright © 2011 Massachusetts Medical Society.



original article



Pioglitazone for Diabetes Prevention in Impaired Glucose Tolerance

Ralph A. DeFronzo, M.D., Devjit Tripathy, M.D., Ph.D., Dawn C. Schwenke, Ph.D., MaryAnn Banerji, M.D., George A. Bray, M.D., Thomas A. Buchanan, M.D., Stephen C. Clement, M.D., Robert R. Henry, M.D., Howard N. Hodis, M.D.,

Abbas E. Kitabchi, M.D., Ph.D., Wendy J. Mack, Ph.D., Sunder Mudaliar, M.D., Robert E. Ratner, M.D., Ken Williams, M.Sc., Frankie B. Stentz, Ph.D.,

Nicolas Musi, M.D., and Peter D. Reaven, M.D., for the ACT NOW Study

From the Texas Diabetes Institute and University of Texas Health Science Cen-ter (R.A.D., D.T., N.M.) and KenAnCo Biostatistics (K.W.) — both in San Anto-nio; Phoenix Veterans Affairs (VA) Health Care System, Phoenix, AZ (D.C.S., P.D.R.); College of Nursing and Health Innovation, Arizona State University, Phoenix (D.C.S.); SUNY Health Science Center at Brooklyn, Brooklyn, NY (M.B.); Pennington Biomedical Research Cen-ter–Louisiana State University, Baton Rouge (G.A.B.); University of Southern California Keck School of Medicine, Los Angeles (T.A.B., H.N.H., W.J.M.); Divi-sion of Endocrinology and Metabolism, Georgetown University, Washington, DC (S.C.C.); VA San Diego Healthcare Sys-tem and University of California at San Diego, San Diego (R.R.H., S.M.); Univer-sity of Tennessee, Division of Endocrinol-ogy, Diabetes, and Metabolism, Memphis (A.E.K., F.B.S.); and Medstar Research In-stitute, Hyattsville, MD (R.E.R.). Address reprint requests to Dr. DeFronzo at the Diabetes Division, University of Texas Health Science Center, 7703 Floyd Curl Dr., San Antonio, TX 78229, or at [email protected].

Drs. DeFronzo and Tripathy contributed equally to this article.


A BS TR AC T

BACKGROUNDImpaired glucose tolerance is associated with increased rates of cardiovascular disease and conversion to type 2 diabetes mellitus. Interventions that may prevent or delay such occurrences are of great clinical importance.

METHODSWe conducted a randomized, double-blind, placebo-controlled study to examine wheth-er pioglitazone can reduce the risk of type 2 diabetes mellitus in adults with impaired glucose tolerance. A total of 602 patients were randomly assigned to receive pioglit-azone or placebo. The median follow-up period was 2.4 years. Fasting glucose was measured quarterly, and oral glucose tolerance tests were performed annually. Con-version to diabetes was confirmed on the basis of the results of repeat testing.

RESULTSAnnual incidence rates for type 2 diabetes mellitus were 2.1% in the pioglitazone group and 7.6% in the placebo group, and the hazard ratio for conversion to diabetes in the pioglitazone group was 0.28 (95% confidence interval, 0.16 to 0.49; P<0.001). Conversion to normal glucose tolerance occurred in 48% of the patients in the pio-glitazone group and 28% of those in the placebo group (P<0.001). Treatment with pioglitazone as compared with placebo was associated with significantly reduced levels of fasting glucose (a decrease of 11.7 mg per deciliter vs. 8.1 mg per deciliter [0.7 mmol per liter vs. 0.5 mmol per liter], P<0.001), 2-hour glucose (a decrease of 30.5 mg per deciliter vs. 15.6 mg per deciliter [1.6 mmol per liter vs. 0.9 mmol per liter], P<0.001), and HbA1c (a decrease of 0.04 percentage points vs. an increase of 0.20 percentage points, P<0.001). Pioglitazone therapy was also associated with a decrease in diastolic blood pressure (by 2.0 mm Hg vs. 0.0 mm Hg, P = 0.03), a re-duced rate of carotid intima–media thickening (31.5%, P = 0.047), and a greater increase in the level of high-density lipoprotein cholesterol (by 7.35 mg per deciliter vs. 4.5 mg per deciliter [0.4 mmol per liter vs. 0.3 mmol per liter], P = 0.008). Weight gain was greater with pioglitazone than with placebo (3.9 kg vs. 0.77 kg, P<0.001), and edema was more frequent (12.9% vs. 6.4%, P = 0.007).

CONCLUSIONSAs compared with placebo, pioglitazone reduced the risk of conversion of impaired glucose tolerance to type 2 diabetes mellitus by 72% but was associated with significant weight gain and edema. (Funded by Takeda Pharmaceuticals and others; ClinicalTrials.gov number, NCT00220961.)



Pioglitazone for Diabetes Prevention in IGT


Type 2 diabetes mellitus affects 21 million Americans,1 and its prevalence is increasing.2 Microvascular and macrovas-

cular complications are common in type 2 diabe-tes mellitus and are related to both the severity and the duration of hyperglycemia.3 The natural history of type 2 diabetes mellitus has been well defined,4 starting with a genetic predisposition and progression from normal glucose tolerance with insulin resistance to impaired glucose toler-ance and eventually type 2 diabetes mellitus with the superimposition of beta-cell failure.

Because hyperglycemia plays a central role in the microvascular and macrovascular complica-tions of diabetes,3,5,6 it is possible that interven-tions that prevent or delay hyperglycemia may effectively prevent or delay these long-term com-plications. Studies have shown that the rate of conversion of impaired glucose tolerance to type 2 diabetes mellitus is reduced with lifestyle mod-ification7; the use of metformin,7 thiazolidine-diones,8-11 or acarbose12; and bariatric surgery.13 The greatest reductions in conversion rates have been observed with weight loss, the use of thia-zolidinediones, and surgery. Troglitazone was reported to be associated with a 55% decrease in the rate of conversion to diabetes among women with prior gestational diabetes,11 but this agent is no longer available. In one study,9 the use of rosiglitazone decreased the risk of diabetes in adults with impaired glucose tolerance by 62%; given concerns about cardiovascular safety,14 how-ever, the Food and Drug Administration has restricted the use of rosig litazone therapy to patients in whom glycemic control cannot be achieved with other medications and who can-not take pioglitazone. We undertook the present study to examine the effect of pioglitazone on diabetes risk and cardiovascular risk factors in adults with impaired glucose tolerance.

Me thods

PatientsWe recruited male and female patients who were 18 years of age or older and had impaired glucose tolerance (defined as a 2-hour glucose level of 140 to 199 mg per deciliter [7.8 to 11.0 mmol per liter] during a single oral glucose-tolerance test)15 and a body-mass index (BMI, the weight in kilograms divided by the square of the height in meters) of 25 or more. Patients were eligible for

enrollment if they had a fasting plasma glucose level between 95 and 125 mg per deciliter (5.3 and 6.9 mmol per liter) and at least one other risk factor for diabetes.16 The complete list of inclu-sion and exclusion criteria is provided in Table 1 in the Supplementary Appendix, available with the full text of this article at NEJM.org. The crite-ria have also been published previously.16 During the course of recruitment, the glycemic inclusion criteria were modified to include patients with a fasting plasma glucose level between 90 and 125 mg per deciliter (5.0 and 6.9 mmol per liter) if their 2-hour plasma glucose level during the oral glucose-tolerance test was between 170 and 199 mg per deciliter (9.4 and 11.1 mmol per li-ter)17; the change was made in recognition of the high risk of diabetes in such persons.

The first participant was recruited in January 2004, with the screening ultimately including 1827 potentially eligible patients with impaired glucose tolerance (Fig. 1). The enrollment of 602 participants was completed in March 2006. Par-ticipants were followed until they reached the primary end point of diabetes, withdrew from the study, were lost to follow-up, or completed the end of the study.

Study DesignThe study design and protocol, which have been described previously,16 are available at NEJM.org. Eight centers participated in the study, which was approved by the institutional review board at each site. Written informed consent was obtained from all participants. The first author designed the study and, along with the coauthors, wrote the first draft and revisions and approved the final version; he also holds the data at the University of Texas Health Science Center in San Antonio. The study was conducted in accordance with the pro-tocol. All authors made the decision to submit the manuscript for publication. All results were trans-mitted to the Data Coordinating Center in Phoenix, Arizona, where they were recorded and audited and then sent to the Data Analysis Center in San Antonio. Takeda Pharmaceuticals provided finan-cial support for the study but had no access to the data.

After eligibility for the study was ascertained, participants underwent randomization accord-ing to center and sex and received 30 minutes of dietary instruction consistent with the goals of the Diabetes Prevention Program,7 which was re-





inforced on follow-up visits. Once enrolled, par-ticipants were asked to fast overnight before un-dergoing an oral glucose-tolerance test at 8 a.m. the next day. Samples were collected every 15 min-utes for 2 hours for measurements of glucose, insulin, and C-peptide.16 Additional baseline as-sessments included measurements of blood pres-sure, height, weight, waist circumference, and level of HbA1c; a lipid profile; screening blood tests; urinalysis, with calculation of the ratio of microalbumin to creatinine; and electrocardiog-raphy. At seven centers, high-resolution B-mode ultrasonography was performed at baseline, 15 to 18 months after baseline, and at the end of the study, to assess the far wall of the right distal

common carotid artery (for details, see the Meth-ods section in the Supplementary Appendix).18,19 Body-fat content was measured with the use of a dual-energy x-ray absorptiometer (DXA) (Hologic).

Participants initially received 30 mg of pioglit-azone per day or placebo. One month after ran-domization, the dose of pioglitazone was in-creased to 45 mg per day. Participants returned at 2, 4, 6, 8, 10, and 12 months during the first year of the study and once every 3 months there-after. At each visit, weight, blood pressure, and pulse were measured and the extent of edema was graded (with an increase in edema defined as an increase of two grades or more from baseline) (for details, see Table 2 in the Supplementary

602 Underwent randomization

1827 Patients were assessed for eligibility

1225 Were excluded277 Had FPG out of range187 Declined to participate118 Had medical exclusions12 Had too few risk factors4 Had behavioral or logistic problems

627 Had OGTT–FPG or 2-hr glucose out of range

299 Were assigned to receiveplacebo

303 Were assigned to receivepioglitazone

71 Did not have a final visit for thefollowing reasons:

3 Had weight gain17 Relocated1 Had work or schedule conflict9 Had lack of interest

10 Did not give reason22 Were lost to follow-up3 Had unrelated illness or adverse

events6 Had other reason

90 Did not have a final visit for thefollowing reasons:

9 Had weight gain6 Relocated

12 Had work or schedule conflict18 Had lack of interest7 Did not give reason

28 Were lost to follow-up5 Had unrelated illness or adverse

events5 Had other reason

228 Completed study213 Completed study

Figure 1. Enrollment, Randomization, and Follow-up of Study Patients.

FPG denotes fasting plasma glucose, and OGTT oral glucose-tolerance test.





Appendix). Fasting plasma glucose was also mea-sured at each follow-up visit. The levels of HbA1c, alanine aminotranferase, and aspartate amino-transferase were measured every 6 months, and the oral glucose-tolerance test was repeated an-nually. All measurements obtained at baseline were repeated at the end of the study.

Conversion of Impaired Glucose Tolerance to Diabetes

The primary outcome was the development of diabetes15 (defined as a fasting plasma glucose level ≥126 mg per deciliter [≥7.9 mmol per liter] or a 2-hour glucose level ≥200 mg per deciliter [11.1 mmol per liter]); an oral glucose-tolerance test was performed to confirm the diagnosis. If the diagnosis was not confirmed, patients con-tinued with their assigned therapy. Diabetes was not confirmed but was considered to have devel-oped in five patients receiving pioglitazone and five patients receiving placebo. At the final visit, six patients had a single oral glucose-tolerance test with results that met the diagnostic criteria for diabetes; four of the six were started on anti-diabetic medication by their physician.

Measurements and CalculationsInsulin sensitivity was derived from plasma glu-cose and insulin measurements obtained during the oral glucose-tolerance test (Matsuda index)20 and from the results of an intravenous glucose-tolerance test with frequent sampling.21 Beta-cell function was calculated as the index of insulin secretion factored by insulin resistance (∆I0–120/∆G0–120 × Matsuda index) during the oral glucose-tolerance test, where ∆I0–120/∆G0–120 represents the mean incremental concentrations of plasma insulin and glucose during the 120-minute oral glucose-tolerance test.22 Beta-cell function was also calculated as the product of insulin secretion and insulin sensitivity (∆I0–10 × SI) during the intra-venous glucose-tolerance test with frequent sam-pling. Laboratory methods are described in detail in the Methods section in the Supplementary Appendix.

Statistical AnalysisThe development of diabetes,15 the primary out-come, was assessed by means of life-table analy-sis of the time from randomization to incident diabetes. Separate life-table estimated cumula-tive incidence curves were calculated for the pio-

glitazone and placebo groups and compared with the use of the log-rank test.23 The Cox pro-portional-hazards model was used to estimate the effect of pioglitazone on the primary outcome.24 Data for patients who were lost to follow-up or who withdrew were censored at the time of the last visit. Statistical tests were two-sided, with an alpha level of 0.05. Data are reported as means ±SE. We calculated that enrollment of approxi-mately 600 patients was required to achieve 90% power, if treatment with pioglitazone decreased the rate of conversion from impaired glucose tol-erance to type 2 diabetes mellitus by 50%.16

For analyses of secondary outcomes, which included changes in levels of fasting plasma glu-cose, 2-hour glucose, and HbA1c, between-group comparisons of changes in repeated or continu-ous measures were performed with the use of general linear mixed models, with data trans-formed to logarithms when appropriate. The sta-tistical heterogeneity of treatment effects within subgroups was assessed. No adjustment was made for multiple comparisons, and subgroup analyses were not prespecified.

Two approaches were used to assess whether patients who completed the study differed from those who withdrew. The first approach involved a withdrawal-free survival analysis of time to withdrawal, with the final study visit used as the censoring variable. Data for patients who under-went an oral glucose-tolerance test at the end-of-study visit were censored at 3 years. All other patients were counted as having withdrawn as of the last study visit. On the basis of this analysis, the hazard ratio for withdrawal in the pioglita-zone group as compared with the placebo group was 1.125 (P = 0.42). In the second approach, miss-ing data (for the two study groups combined) were assessed for each continuous measure with analy-sis of variance, stratified according to whether the measure was missing at each subsequent visit. Since neither approach produced statistically sig-nificant evidence of bias due to missing data, the primary and secondary analyses were performed without data imputation.

R esult s

ParticipantsThe mean age of the 602 study participants was 52.3±0.5 years, and 58% were women (Table 1). The mean BMI was 34.5±0.4. A total of 407 pa-





Table 1. Baseline Characteristics of the Patients.*

CharacteristicPioglitazone

(N = 303)Placebo(N = 299) P Value

Isolated impaired glucose tolerance — no. 98 97 0.98

Both impaired glucose tolerance and impaired fasting glucose tolerance — no.

205 202 0.98

Ratio of women to men — % 58/42 58/42 0.96

Race or ethnic group — no.† 0.45

Hispanic 79 75

White 156 171

Black 57 44

Other 11 9

Family history of diabetes — no. (%) 177 (58.4) 186 (62.2) 0.34

History of gestational diabetes mellitus — no. (% of women) 24 (14) 37 (21) 0.07

Mean age — yr 53.0±0.4 51.5±0.7 0.12

Age group — %

18–39 yr 36 42 0.155

40–59 yr 32 28 0.33

≥60 yr 32 30 0.66

Height — cm 166±0.5 167±0.6 0.47

Mean BMI 33.0±0.4 34.5±0.4 0.44

BMI group — no. (%)

<30 79 (26.0) 76 (25.4) 0.95

30–35 109 (36.0) 100 (33.4) 0.71

>35 114 (37.6) 122 (40.8) 0.65

Waist circumference — cm

Men 110.5±1.1 112.2±1.3 0.70

Women 103.1±0.9 103.7±1.0 0.31

HbA1c — % 5.5±0.4 5.5±0.4 0.23

Fasting plasma glucose — mg/dl 105±0.4 105±0.4 0.72

2-Hr plasma glucose — mg/dl 168±1 168±1 0.80

Fasting plasma insulin — mU/liter 10.5±0.5 10.7±0.6 0.84

Lipid levels — mg/dl

Total cholesterol 169±2 173±2 0.22

LDL cholesterol 104±2 108±2 0.20

HDL cholesterol 40±1 41±1 0.57

Triglycerides 122±3 121±3 0.84

Fasting free fatty acids (µmol/liter) 551±13 528±13 0.17

Blood pressure — mm Hg

Systolic 127±0.9 128±0.9 0.57

Diastolic 74±0.6 74±0.6 0.99

* Plus–minus values are means ±SE. No intergroup differences were significant. To convert the values for glucose to mil-limoles per liter, multiply by 0.05551. To convert the values for insulin to picomoles per liter, multiply by 6.945. To con-vert the values for cholesterol to millimoles per liter, multiply by 0.02586. To convert the values for triglycerides to milli-moles per liter, multiply by 0.01129. BMI denotes body-mass index (the weight in kilograms divided by the square of the height in meters), HDL high-density lipoprotein, and LDL low-density lipoprotein.

† Race or ethnic group was self-reported.





tients had both impaired fasting glucose and im-paired glucose tolerance, and 195 had isolated impaired glucose tolerance. Baseline levels of HbA1c, fasting plasma glucose, and 2-hour glu-cose were 5.50±0.04%, 105±0.3 mg per deciliter (5.8±0.02 mmol per liter), and 168±1 mg per deciliter (9.32±0.06 mmol per liter), respectively. None of the baseline clinical, anthropometric, or laboratory variables differed significantly between the placebo group and the pioglitazone group (Table 1).

Follow-upDuring a median follow-up period of 2.4 years (mean, 2.2), diabetes developed in 50 of the 299 patients in the placebo group (16.7%) and in 15 of the 303 patients in the pioglitazone group (5.0%). The annual average incidence of diabetes, calcu-lated on the basis of person-years, was 7.6% in the placebo group and 2.1% in the pioglitazone group (P<0.001) (Fig. 2). The hazard ratio for de-velopment of diabetes in the pioglitazone group was 0.28 (95% confidence interval, 0.16 to 0.49; P<0.001). Adjustment for baseline characteristics did not alter the hazard ratio. The number of peo-ple who would need to be treated to prevent one case of diabetes was 8 for 2.2 years of the trial and 18 for 1 year. Among the patients who com-pleted the study, 103 of those in the pioglitazone group (48%) and 65 of those in the placebo group (28%) had normal glucose tolerance (P<0.001).

A total of 161 patients did not complete the study (71 in the placebo group and 90 in the pioglitazone group). The median follow-up time for these patients was 7.6 months. Baseline char-acteristics of the patients who did not complete the study were similar to those of the 441 pa-tients who completed the study (i.e., those who had conversion to type 2 diabetes mellitus dur-ing the study or who completed the oral glucose-tolerance test at the end of the trial). Reasons for not completing the study included weight gain (in 9 patients in the pioglitazone group and 3 in the placebo group); patients also left for reasons unrelated to the study medication (Fig. 1). The rate of adherence to the study regimen, assessed by means of pill counts, was greater than 80% in both groups. At the end of the study, 64% of the patients in the treatment group were taking pio-glitazone at a daily dose of 45 mg and 81% of those in the placebo group were taking the cor-responding placebo dose. The major reasons for

not increasing the dose of pioglitazone from 30 to 45 mg per day or for not maintaining the 45-mg dose were weight gain and edema.

Effects of PioglitazoneProtection from diabetes with pioglitazone was of similar magnitude (with no significant hetero-geneity) in subgroups defined by sex, age, weight, race or ethnic group, and fasting glucose level, as well as in patients with both impaired glucose tolerance and impaired fasting glucose and those with isolated impaired glucose tolerance (Fig. 3). There was no evidence of heterogeneity of the response according to the baseline level of HbA1c.

Greater reductions in fasting and 2-hour glu-cose levels were achieved in the pioglitazone group than in the placebo group (P<0.001 for both comparisons), with a between-group differ-ence of 3.5±1.1 mg per deciliter (0.2±0.06 mmol per liter) and 14±3 mg per deciliter (0.8±0.17 mmol per liter), respectively, at the end of the study (Fig. 4A and 4B). Levels of HbA1c differed between the groups throughout the study (P<0.001), increas-ing by 0.20±0.02% in the placebo group, with no change in the pioglitazone group. Body weight, BMI, and body fat increased in the placebo group (96.7±1.2 to 97.3±1.3 kg, 34.5±0.4 to 34.7±0.4, and 39.0±0.7 to 39.3±0.7%, respectively) and in the pioglitazone group (94.9±1.2 to 98.7±1.3 kg,

Cum

ulat

ive

Haz

ard

0.4

0.2

0.3

0.1

0.00 6 12 18 24 48

Months since Randomization

Hazard ratio, 0.28 (95% CI, 0.16–0.49)P<0.001

No. at RiskPlaceboPioglitazone

299303

259262

228244

204228

191218

30

134140

36

8387

42

1724

Placebo

Pioglitazone

Figure 2. Kaplan–Meier Plot of Hazard Ratios for Time to Development of Diabetes.





34.1±0.4 to 35.5±0.4, and 40.0±0.8 to 41.9±0.7%, respectively), but the increments were greater with pioglitazone (P<0.001 for all comparisons).

Systolic blood pressure declined slightly in both groups, but the difference in decline be-tween the groups was not significant. Diastolic blood pressure was consistently lower in the pio-glitazone group (P = 0.01). As compared with pla-cebo, pioglitazone reduced levels of both alanine aminotranferase and aspartate aminotransfer-ase (P<0.001). The change in high-density lipopro-tein (HDL) cholesterol was greater with pioglita-zone (40±1 to 48±1 mg per deciliter [2.2±0.06 to 2.7±0.06 mmol per liter]) than with placebo

(41±1 to 45.1±0.7 mg per deciliter [2.3±0.06 to 2.5±0.04 mmol per liter]) (P = 0.008 for the dif-ference between groups). Triglyceride levels de-clined significantly with pioglitazone (129±7.5 to 110±4.0 mg per deciliter [7.2±0.42 to 6.1±0.22 mmol per liter], P = 0.007) but not with placebo (124±4.6 to 113±4.0 mg per deciliter [6.9±0.25 to 6.30±0.22 mmol per liter], P = 0.90); the differ-ence between groups was not significant. Nei-ther pioglitazone nor placebo altered levels of low-density lipoprotein cholesterol.

Insulin sensitivity as measured with the Mat-suda index increased more with pioglitazone than with placebo (4.31±0.24 to 7.65±0.34 vs.

0.5 1.0 4.01.5 3.5

Placebo BetterPioglitazone Better

AllSex

MaleFemale

Age<50 yr50–59 yr≥60 yr

Race or ethnic groupWhiteHispanicBlack

BMI<30 30–35>35

Waist circumferenceMen, <100 cmMen, ≥100 cmWomen, <90 cmWomen, ≥90 cm

Fasting plasma glucose95–105 mg/dl106–115 mg/dl116–125 mg/dl

Glucose testIsolated IGTIFG and IGT

Hazard Ratio (95% CI)Pioglitazone PlaceboSubgroup

0.0

P Value forHeterogeneity

7.6

6.48.4

8.59.44.5

7.07.0

11.2

5.17.58.4

3.67.5

11.28.3

5.19.6

18.5

3.79.3

2.1

1.32.8

3.42.10.8

2.12.22.8

2.61.12.8

1.31.13.42.6

1.52.66.3

1.82.3

0.66

0.75

0.98

0.75

0.48

0.84

0.88

incidence rate (per 100 person-yr)

Figure 3. Incidence Rates, Hazard Ratios, and Tests for Heterogeneity for All Patients and Selected Subgroups.

The figure shows incidence rates per 100 person-years and corresponding hazard ratios and confidence intervals for the effects of pioglitazone as compared with placebo on the conversion of impaired glucose tolerance to diabetes. The x axis is interrupted to allow for better visual presentation. BMI denotes body-mass index (the weight in kilo-grams divided by the square of the height in meters), IFG impaired fasting glucose, and IGT impaired glucose toler-ance. To convert the values for glucose to millimoles per liter, multiply by 0.05551.





4.31±0.30 to 5.23±0.31, P<0.001). Insulin sensi-tivity as determined with the use of an intrave-nous glucose-tolerance test with frequent sam-pling in a subgroup of 191 patients was not altered in either group. The index of insulin secretion factored by insulin resistance, calculat-ed on the basis of the oral glucose-tolerance test (I0–120/∆G0–120 × Matsuda index), increased more with pioglitazone than with placebo (3.43±0.12 to 5.44±0.31 vs. 3.81±0.30 to 4.20±0.20, P<0.005). Similarly, the insulin secretion–insulin resistance

index, calculated with data from the frequent-sampling intravenous glucose-tolerance test, in-creased more with pioglitazone than with pla-cebo (848±65 to 1186±113 vs. 824±47 to 832±57, P<0.01).

Carotid intima–media thickening increased more slowly in the pioglitazone group than in the placebo group throughout the study (Fig. 4I). The differences between groups were 16.4% at the study midpoint and 31.5% at the end of the study (P = 0.047 for the overall difference between

A Fasting Plasma GlucoseM

ean

Perc

ent C

hang

e

0

−10

−5

−150 12 24 36

Month

P<0.001

B 2-Hr Plasma Glucose

Mea

n Pe

rcen

t Cha

nge

0

−10

−200 12 24 36

Month

P<0.001

C HbA1c

Mea

n Pe

rcen

t Cha

nge

2.5

−2.5

0.0

−5.0

−7.50 12 24 36

Month

P<0.001

D Weight

Mea

n Pe

rcen

t Cha

nge

5.0

0.0

2.5

−2.50 12 24 36

Month

P<0.001

E Systolic Blood Pressure

Mea

n Pe

rcen

t Cha

nge

1.0

−1.00.0

−2.0

−4.0−3.0

0 12 24 36

Month

P=0.16

F Diastolic Blood Pressure

Mea

n Pe

rcen

t Cha

nge

4.0

−2.0

0.02.0

−4.0−6.0

0 12 24 36

Month

P=0.01

G ALT

Mea

n Pe

rcen

t Cha

nge

1020

0

−10

−200 12 24 36

Month

P<0.001

H AST

Mea

n Pe

rcen

t Cha

nge

3030

1020

0

−10

−200 12 24 36

Month

P<0.001

I Carotid Intima–Media Thickness

Mea

n Pe

rcen

t Cha

nge

4.0

2.0

0.0

0 12 24 36

Month

P=0.047

Pioglitazone

Placebo

Figure 4. Effects of Pioglitazone as Compared with Placebo.

Over the course of the study, mean percentage changes and standard errors in continuous measures were calculated with the use of a linear, mixed-repeated-measures model fit to all available data for each measure. As compared with placebo, treatment with pioglitazone (dashed lines) had beneficial effects on fasting plasma glucose levels (Panel A), 2-hour plasma glucose levels (Panel B), and HbA1c lev-els (Panel C) and on systolic and diastolic blood pressure (Panels E and F, respectively), alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels (Panels G and H, respectively), and carotid intima–media thickness (Panel I). Weight gain was greater with pioglitazone than with placebo (Panel D). (Body-mass index was calculated at each examination with the use of height measured at baseline; as a result, the percentage change in BMI is identical to the percentage change in weight.) A total of 365 patients (placebo group, 186; pioglitazone group, 179) completed the follow-up examination at 15 to 18 months for measurement of carotid intima–media thickness, and 336 patients (placebo group, 173; pioglitazone group, 163) completed the final examination for measurement of carotid intima–media thickness. P values are shown for the interaction between time and study group, indicating whether the slopes differ sig-nificantly over time.





groups). The ratio of urinary microalbumin to creatinine was low at baseline; during the study it fell slightly and similarly in the two groups (1.85±0.2 to 1.53±0.2 mg per gram in the pioglit-azone group vs. 1.47±0.3 to 1.25±0.3 mg per gram in the placebo group, P = 0.20).

Adverse EventsAdverse events occurred in 121 patients in the pla-cebo group and 151 patients in the pioglitazone group (P = 0.03) (Table 2). Edema increased at some point during the trial in 19 patients receiv-ing placebo (6.4%) and 39 patients receiving pio-glitazone (12.9%) (P = 0.007). Events related to the cardiovascular system numbered 23 in the placebo group (7.7%) and 26 in the pioglitazone group (8.6%) (Table 3 in the Supplementary Appendix), with 1 case of congestive heart failure in each group (0.3%). One unexplained sudden death oc-

curred in the placebo group, and three deaths occurred in the pioglitazone group (one unex-plained sudden death, one death from biliary carcinoma, and one death from a carcinoid tu-mor). Nine fractures occurred in 8 of the patients receiving pioglitazone (3%) and eight fractures occurred in 7 of the patients receiving placebo (2.6%) (Table 4 in the Supplementary Appendix). All fractures were associated with trauma.

Discussion

Although they are considered to have prediabetes, patients in the upper third of the range for im-paired glucose tolerance are at or close to the maximum level of insulin resistance and have lost approximately 80% of beta-cell function.25,26 Histologic studies suggest that the beta-cell mass in patients with impaired fasting glucose is sig-nificantly reduced as compared with persons who have normal fasting glucose,27 and two thirds of the patients in our study had impaired fasting glucose. Moreover, at least 10% of patients with impaired glucose tolerance have background dia-betic retinopathy,28,29 and peripheral neuropathy is common in these patients.30 Because the charac-teristic pathophysiological defects of type 2 dia-betes mellitus and the microvascular complica-tions of diabetes are already evident in patients with impaired glucose tolerance, it is reasonable to consider interventions at this stage to prevent the development of overt diabetes.

Lifestyle interventions effectively reduce the conversion of impaired glucose tolerance to dia-betes31-34 and remain the primary approach to prevention of type 2 diabetes mellitus. However, many people remain at risk for type 2 diabetes mellitus despite attempts at lifestyle chang-es.7,31,32,34 Metformin reduces the risk of conver-sion to type 2 diabetes mellitus by 31%, without weight gain.7 Thiazolidinediones also effectively reduce the risk of development of type 2 diabetes mellitus in patients with impaired glucose toler-ance.8-11 In our study, pioglitazone decreased the rate of conversion to diabetes by 72%, a change that was slightly larger than that ob-served with other thiazolidinediones (52 to 62%) and lifestyle modification (58%). Although sub-group analyses were not prespecified in our study design, pioglitazone reduced the risk of conversion to diabetes in patients with isolated impaired glucose tolerance, in those with both

Table 2. Number and Type of Adverse Events.*

Adverse EventPioglitazone

(N = 303)Placebo (N = 299)

no. of events

Cancer 3 8

Cardiovascular system 26 23

Central nervous system 6 5

Death 3 1

Digestive system 13 12

Edema† 39 19

Elective surgery 22 16

Endocrine system 1 3

Immune system 2 4

Musculoskeletal system 12 13

Ophthalmologic system 0 1

Respiratory system 9 6

Reproductive system 4 4

Skin 6 3

Urogenital system 5 3

Weight gain‡ 205 128

Total 356 249

* For the comparison of placebo and pioglitazone regard-ing frequency of edema, cardiovascular events, and total events, P = 0.007, P = 0.80, and P = 0.03, respectively. The total number of adverse events — excluding edema — did not differ significantly between groups (P = 0.52).

† Edema was defined as an increase above baseline by two or more grades on one or more distinct study visits.

‡ Weight gain was defined as a gain of more than 1 kg.





impaired fasting glucose and impaired glucose tolerance, in both men and women, and in all age and weight groups. The proportion of patients who had a return to normal glucose tolerance was greater with pioglitazone than with placebo. The mean weight gain in patients treated with pioglitazone was 3.6 kg. However, the greater the weight gain, the greater the improvements in beta-cell function and insulin sensitivity, and thus the greater the reduction in HbA1c.26,35,36 The effect of weight gain on cardiovascular risk cannot be ascertained, but in the Prospective Pioglitazone Clinical Trial in Macrovascular Events (PROactive) (ClinicalTrials.gov number, NCT00174993),37 a weight gain of 3.6 kg was not associated with an increase in the composite cardiovascular end point.

It is not feasible to conduct studies of micro-vascular outcomes in patients with impaired glu-cose tolerance because of the large sample and long study duration required. However, if devel-opment of diabetic hyperglycemia can be delayed or prevented, it is plausible that the onset of microvascular complications might be slowed. During the course of this study, fasting and 2-hour glucose levels and HbA1c levels were sig-nificantly lower in the pioglitazone group than in the placebo group. The small difference in levels of HbA1c between the two groups is not surpris-ing, given the low baseline level of HbA1c in pa-tients with impaired glucose tolerance. However, further glycemic separation between groups over time would presumably have a beneficial effect on microvascular disease.

Pioglitazone was associated with lower dia-stolic blood pressure, higher levels of HDL choles-terol, and reduced rates of carotid intima–media thickening, as compared with placebo. Carotid intima–media thickening is highly correlated with cardiovascular events, and changes in this measure over time have predictive value beyond that of standard markers of risk.18,38 Such results suggest that pioglitazone may provide some pro-tection against the development of atheroscle-rotic cardiovascular disease, which is consistent with reports of reductions in the volume of coro-nary plaque39 and in mortality, nonfatal myocar-dial infarction, and stroke, the secondary end points in the PROactive.37

Loss to follow-up was relatively high in both study groups (24% in the placebo group and 30% in the pioglitazone group, not a significant dif-

ference). Since withdrawal rates and baseline characteristics were similar between groups, bi-ased results seem unlikely. The modest difference in levels of HbA1c between groups suggests that the reduced progression of carotid intima–media thickening with pioglitazone may reflect im-provements in other metabolic variables (Fig. 4). Although pioglitazone is a well-documented insu-lin sensitizer, insulin resistance was reduced ac-cording to the Matsuda index of insulin sensitivity measured during the oral glucose-tolerance test but not according to measurement of insulin sensitivity during the frequent-sampling intrave-nous glucose-tolerance test. These disparate re-sults may partly reflect the greater variation among centers in the results of the intravenous glucose-tolerance test with frequent sampling, which is more difficult to perform.

Edema and weight gain were greater with pio-glitazone than with placebo, as reported previ-ously,8,40 and edema largely accounted for the increase in adverse events associated with pio-glitazone. Physician-reported congestive heart failure developed in only one patient in each group. Although an increased incidence of frac-tures has been reported with the use of thiazoli-dinediones, in this study, the incidence was simi-lar in both groups, and all fractures were related to trauma.

In summary, treatment with pioglitazone in patients with impaired glucose tolerance reduced the risk of diabetes, although pioglitazone was associated with significant weight gain and ede-ma. Treatment of 18 participants for 1 year pre-vented one case of diabetes. Use of pioglitazone improved diastolic blood pressure, HDL choles-terol levels, and serum levels of alanine amino-tranferase and aspartate aminotransferase, and it slowed progression of carotid intima–media thickening. The influence of these effects on long-term diabetic complications remains to be determined.

Supported by Takeda Pharmaceuticals, grants from the Gen-eral Clinical Research Center (GCRC) at the University of Ten-nessee Health Science Center (MO1-RR-00221) and the GCRC at the University of Southern California Keck School of Medicine (MO1-RR-00043), and by the Veterans Affairs institutions in Phoenix and San Diego, which contributed their resources and the use of their facilities.

Dr. Banerji reports receiving consulting fees from BMS, No-vartis, Boehringer Ingelheim, Sanofi-Aventis, Merck, and Roche, and lecture fees from Merck and Sanofi-Aventis; Dr. Buchanan reports receiving consulting fees and lecture fees from Takeda and reports that the University of Southern California Keck





School of Medicine has received grant support from Takeda; Dr. DeFronzo reports receiving payments for board membership from Amylin, Takeda, ISIS, and Boehringer Ingelheim and reports that the University of Texas Health Science Center at San Antonio has received grant support from Takeda, Amylin, and Eli Lilly; Dr. Henry reports receiving consulting fees, lecture fees, and payment for expert testimony from Takeda; Dr. Musi reports receiving consulting fees from Merck, Daiichi-Sankyo, Takeda, and Novartis; Dr. Ratner reports that the Medstar Research In-stitute has received consulting fees from Amylin, NovoNordisk, Sanofi-Aventis, and Genentech–Roche and grant support from Amylin, NovoNordisk, GlaxoSmithKline, Bayhill, Halozyme, and Integrium; Dr. Reaven reports receiving consulting fees from BMS, lecture fees from Merck, and payment for the devel-

opment of educational presentations from Amylin; Dr. Reaven reports that the Carl T. Hayden Veterans Affairs Medical Center has received grant support from Amylin; and Dr. Tripathy re-ports receiving grant support from Takeda Pharmaceuticals. No other potential conflict of interest relevant to this article was reported.


We thank our nurses and other technical staff for their expert help, without whom this study would not have been possible; the 602 patients who participated in this study; and Joel Michalek, Ph.D., and Lee Ann Zarzabal, Ph.D., of the Department of Epide-miology and Biostatistics at the University of Texas Health Science Center at San Antonio for help in the initial statistical analyses.

References

1. Cowie CC, Rust KF, Byrd-Holt DD, et al. Prevalence of diabetes and impaired fasting glucose in adults in the U.S. popu-lation: National Health And Nutrition Ex-amination Survey 1999-2002. Diabetes Care 2006;29:1263-8.2. Diabetes programme: facts and fig-ures. Geneva: World Health Organization, 2007. (http://www.who.int/diabetes/facts/en.)3. Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with macrovas-cular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 2000;321:405-12.4. DeFronzo RA. Banting Lecture: from the triumvirate to the ominous octet: a new paradigm for the treatment of type 2 dia-betes mellitus. Diabetes 2009;58:773-95.5. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-Year follow-up of intensive glucose control in type 2 dia-betes. N Engl J Med 2008;359:1577-89.6. Nathan DM, Cleary PA, Backlund JY, et al. Intensive diabetes treatment and cardiovascular disease in patients with type 2 diabetes. N Engl J Med 2005;353: 2643-53.7. Knowler WC, Barrett-Connor E, Fowler SE, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002;346:393-403.8. Xiang AH, Peters RK, Kjos SL, et al. Effect of pioglitazone on pancreatic beta-cell function and diabetes risk in Hispan-ic women with prior gestational diabetes. Diabetes 2006;55:517-22.9. Gerstein HC, Yusuf S, Bosch J, et al. Effect of rosiglitazone on the frequency of diabetes in patients with impaired glu-cose tolerance or impaired fasting glu-cose: a randomised controlled trial. Lan-cet 2006;368:1096-105. [Erratum, Lancet 2006;368:1770.]10. Knowler WC, Hamman RF, Edelstein SL, et al. Prevention of type 2 diabetes with troglitazone in the Diabetes Preven-tion Program. Diabetes 2005;54:1150-6.11. Buchanan TA, Xiang AH, Peters RK, et al. Preservation of pancreatic beta-cell function and prevention of type 2 diabe-tes by pharmacological treatment of insu-

lin resistance in high-risk Hispanic wom-en. Diabetes 2002;51:2796-803.12. Chiasson JL, Josse RG, Gomis R, Hanefeld M, Karasik A, Laakso M. Acar-bose for prevention of type 2 diabetes mellitus: the STOP-NIDDM randomised trial. Lancet 2002;359:2072-7.13. Buchwald H, Estok R, Fahrbach K, et al. Weight and type 2 diabetes after bar-iatric surgery: systematic review and meta-analysis. Am J Med 2009;122(3):248.e5-256.e5.14. Nissen SE, Wolski K. Effect of rosiglit-azone on the risk of myocardial infarction and death from cardiovascular causes. N Engl J Med 2007;356:2457-7. [Erratum, N Engl J Med 2007;357:100.]15. American Diabetes Association. Diag-nosis and classification of diabetes melli-tus. Diabetes Care 2008;31:Suppl:S55-S60.16. DeFronzo RA, Banerji M, Bray GA, et al. Actos Now for the prevention of diabe-tes (ACT NOW) study. BMC Endocr Dis-ord 2009;9:17.17. Edelstein SL, Knowler WC, Bain RP, et al. Predictors of progression from im-paired glucose tolerance to NIDDM: an analysis of six prospective studies. Diabe-tes 1997;46:701-10.18. Hodis HN, Mack WJ, LaBree L, et al. The role of carotid arterial intima-media thickness in predicting clinical coronary events. Ann Intern Med 1998;128:262-9.19. Hodis HN, Mack WJ, Lobo RA, et al. Estrogen in the prevention of atheroscle-rosis: a randomized, double-blind, place-bo-controlled trial. Ann Intern Med 2001; 135:939-53.20. Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glu-cose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care 1999;22:1462-70.21. Bergman RN, Phillips LS, Cobelli C. Physiologic evaluation of factors control-ling glucose tolerance in man: measure-ment of insulin sensitivity and beta-cell glucose sensitivity from the response to intravenous glucose. J Clin Invest 1981; 68:1456-67.22. Abdul-Ghani MA, Jenkinson CP, Rich-ardson DK, Tripathy D, DeFronzo RA. In-sulin secretion and action in subjects with

impaired fasting glucose and impaired glucose tolerance: results from the Veter-ans Administration Genetic Epidemiolo-gy Study. Diabetes 2006;55:1430-5.23. Kalbfleisch J, Prentice RL. The statis-tical analysis of failure time data. New York: John Wiley, 1980.24. Lachin JM, Wei LJ. Estimators and tests in the analysis of multiple noninde-pendent 2 x 2 tables with partially miss-ing observations. Biometrics 1988;44:513-28. [Erratum, Biometrics 1988;44:923.]25. Ferrannini E, Gastaldelli A, Miyazaki Y, Matsuda M, Mari A, DeFronzo RA. Beta-cell function in subjects spanning the range from normal glucose tolerance to overt diabetes: a new analysis. J Clin Endocrinol Metab 2005;90:493-500.26. Gastaldelli A, Ferrannini E, Miyazaki Y, Matsuda M, Mari A, DeFronzo RA. Thia-zolidinediones improve beta-cell function in type 2 diabetic patients. Am J Physiol Endocrinol Metab 2007;292:E871-E883.27. Butler AE, Janson J, Bonner-Weir S, Ritzel R, Rizza RA, Butler PC. Beta-cell deficit and increased beta-cell apoptosis in humans with type 2 diabetes. Diabetes 2003;52:102-10.28. Diabetes Prevention Program Re-search Group. The prevalence of retinopa-thy in impaired glucose tolerance and recent-onset diabetes in the Diabetes Prevention Program. Diabet Med 2007; 24:137-44.29. Cheng YJ, Gregg EW, Geiss LS, et al. Association of A1C and fasting plasma glucose levels with diabetic retinopathy prevalence in the U.S. population: impli-cations for diabetes diagnostic thresh-olds. Diabetes Care 2009;32:2027-32.30. Ziegler D, Rathmann W, Dickhaus T, Meisinger C, Mielck A. Prevalence of poly-neuropathy in pre-diabetes and diabetes is associated with abdominal obesity and macroangiopathy: the MONICA/KORA Augsburg Surveys S2 and S3. Diabetes Care 2008;31:464-9.31. Pan XR, Li GW, Hu YH, et al. Effects of diet and exercise in preventing NIDDM in people with impaired glucose toler-ance: the Da Qing IGT and Diabetes Study. Diabetes Care 1997;20:537-44.32. Li G, Zhang P, Wang J, et al. The long-





term effect of lifestyle interventions to prevent diabetes in the China Da Qing Diabetes Prevention Study: a 20-year fol-low-up study. Lancet 2008;371:1783-9.33. Knowler WC, Fowler SE, Hamman RF, et al. 10-Year follow-up of diabetes incidence and weight loss in the Diabetes Prevention Program Outcomes Study. Lancet 2009;374:1677-86. [Erratum, Lan-cet 2009;374:2054.]34. Lindström J, Ilanne-Parikka P, Pel-tonen M, et al. Sustained reduction in the incidence of type 2 diabetes by lifestyle intervention: follow-up of the Finnish Di-abetes Prevention Study. Lancet 2006;368: 1673-9.35. Miyazaki Y, Mahankali A, Matsuda M,

et al. Improved glycemic control and en-hanced insulin sensitivity in type 2 dia-betic subjects treated with pioglitazone. Diabetes Care 2001;24:710-9.36. Miyazaki Y, De Filippis E, Bajaj M, et al. Predictors of improved glycemic con-trol with rosiglitazone therapy in type 2 diabetic patients: a practical approach for the primary care physician. Br J Diabetes Vasc Dis 2005;5:28-35.37. Dormandy JA, Charbonnel B, Eckland DJ, et al. Secondary prevention of macro-vascular events in patients with type 2 diabetes in the PROactive Study (PRO-spective pioglitAzone Clinical Trial In macroVascular Events): a randomised controlled trial. Lancet 2005;366:1279-89.

38. Crouse JR III. Thematic review series: patient-oriented research — imaging ath-erosclerosis: state of the art. J Lipid Res 2006;47:1677-99.39. Nissen SE, Nicholls SJ, Wolski K, et al. Comparison of pioglitazone vs glimepi-ride on progression of coronary athero-sclerosis in patients with type 2 diabetes: the PERISCOPE randomized controlled trial. JAMA 2008;299:1561-73.40. Mazzone T, Meyer PM, Feinstein SB, et al. Effect of pioglitazone compared with glimepiride on carotid intima-media thickness in type 2 diabetes: a random-ized trial. JAMA 2006;296:2572-81.Copyright © 2011 Massachusetts Medical Society.



original article



Mercury Exposure and Risk of Cardiovascular Disease in Two U.S. Cohorts

Dariush Mozaffarian, M.D., Dr.P.H., Peilin Shi, Ph.D., J. Steven Morris, Ph.D., Donna Spiegelman, Sc.D., Philippe Grandjean, M.D., David S. Siscovick, M.D., M.P.H.,

Walter C. Willett, M.D., Dr.P.H., and Eric B. Rimm, Sc.D.

From the Division of Cardiovascular Medi-cine (D.M.) and Channing Laboratory (D.M., W.C.W., E.B.R.), Brigham and Women’s Hospital and Harvard Medical School; and the Departments of Epide-miology (D.M, P.S., D.S., W.C.W., E.B.R.), Nutrition (D.M., W.C.W., E.B.R.), Biosta-tistics (D.S.), and Environmental Health (P.G.), Harvard School of Public Health — all in Boston; the University of Mis-souri Research Reactor, Columbia (J.S.M.); and the Cardiovascular Health Research Unit, Departments of Medicine and Epi-demiology, University of Washington, Seattle (D.S.S.). Address reprint requests to Dr. Mozaffarian, 665 Huntington Ave., Bldg. 2-319, Boston, MA 02115, or at [email protected].


A BS TR AC T

BackgroundExposure to methylmercury from fish consumption has been linked to a potentially increased risk of cardiovascular disease, but evidence from prior studies is equivo-cal. Beneficial effects of the ingestion of fish and selenium may also modify such effects.

MethodsAmong subjects from two U.S. cohorts (a total of 51,529 men and 121,700 women) whose toenail clippings had been stored, we prospectively identified incident cases of cardiovascular disease (coronary heart disease and stroke) in 3427 participants and matched them to risk-set–sampled controls according to age, sex, race, and smok-ing status. Toenail mercury and selenium concentrations were assessed with the use of neutron-activation analysis. Other demographic characteristics, cardiovascu-lar risk factors, fish consumption, and lifestyle habits were assessed by means of validated questionnaires. Associations between mercury exposure and incident car-diovascular disease were evaluated with the use of conditional logistic regression.

ResultsMedian toenail mercury concentrations were 0.23 µg per gram (interdecile range, 0.06 to 0.94) in the case participants and 0.25 µg per gram (interdecile range, 0.07 to 0.97) in the controls. In multivariate analyses, participants with higher mercury exposures did not have a higher risk of cardiovascular disease. For comparisons of the fifth quintile of mercury exposure with the first quintile, the relative risks were as follows: coronary heart disease, 0.85 (95% confidence interval [CI], 0.69 to 1.04; P = 0.10 for trend); stroke, 0.84 (95% CI, 0.62 to 1.14; P = 0.27 for trend); and total cardiovascular disease, 0.85 (95% CI, 0.72 to 1.01; P = 0.06 for trend). Findings were similar in analyses of participants with low selenium concentrations or low overall fish consumption and in several additional sensitivity analyses.

ConclusionsWe found no evidence of any clinically relevant adverse effects of mercury exposure on coronary heart disease, stroke, or total cardiovascular disease in U.S. adults at the exposure levels seen in this study. (Funded by the National Institutes of Health.)



Mercury Exposure and Risk of Cardiovascular Disease


Controversy has arisen over the risks and benefits of fish consumption in adults. Fish intake is inversely associated

with the risk of coronary heart disease, especially fatal coronary heart disease, and ischemic stroke.1 Fish are also the major source of exposure to meth-ylmercury.2,3 Chronic, low-level methylmercury exposure appears to cause subtle but measurable neurodevelopmental delay in infants, and it is rec-ommended that women of childbearing age, preg-nant or nursing mothers, and infants and young children eat no more than two servings of fish per week and also limit their intake of selected species of fish that are especially high in methyl-mercury content.4 In adults, however, the main health concern is potential cardiovascular toxicity, as suggested by results of experiments in animals and limited studies in humans.2,5

Prior clinical studies of mercury exposure and cardiovascular diseases have been relatively small, and the results have been inconsistent.6-11 Thus, government agencies, the Institute of Medicine, and risk–benefit analyses have identified the ef-fect of methylmercury exposure on cardiovascu-lar disease as an important area of uncertainty that warrants further investigation, since current data are not sufficient to quantitatively or quali-tatively determine the potential effects.1,12-15 We prospectively investigated the relationships be-tween mercury exposure and incident cardiovas-cular disease in two large U.S. cohorts. Because the trace element selenium provides protection against mercury toxicity in some experimental models,1,2 we also evaluated selenium exposure as a potential effect modifier.

Me thods

Population and Study DesignThe designs of the Health Professionals Follow-up Study (HPFS) and Nurses’ Health Study (NHS) have been described previously.16,17 The HPFS is a prospective cohort study that enrolled 51,529 male U.S. health professionals 40 to 75 years of age in 1986. The NHS is a prospective cohort study that enrolled 121,700 female U.S. registered nurses 30 to 55 years of age in 1976. Participants in both cohorts are followed by means of bien-nial questionnaires on medical history, risk fac-tors, lifestyle, and disease incidence.

We performed a nested case–control study in-volving participants from both cohorts. The study

was designed by the authors and approved by the human subjects committees of all participating institutions.

In prior analyses,18-22 we found that concen-trations of mercury and selenium in toenails are excellent biomarkers of usual methylmercury and selenium exposure. Toenail clippings were pro-vided by 68% of HPFS participants in 1987 and by 52% of NHS participants during the period from 1982 through 1984. Demographic, risk-factor, and lifestyle characteristics of these participants were similar to those of participants who did not provide clippings (data not shown). About two thirds of the HPFS participants were dentists, and they were excluded from this analysis owing to occupational exposure to inorganic mercury dur-ing dental-amalgam procedures.18 All participants provided implied consent by returning completed questionnaires and toenail samples.

Cases and ControlsParticipants with incident cardiovascular disease (defined as nonfatal myocardial infarction, fatal coronary heart disease, or stroke) were identified from among HPFS and NHS participants who had provided toenail samples. Methods for ascer-tainment of cardiovascular events in the two co-horts have been described previously.16,17,23 When cardiovascular disease outcomes were reported, we obtained permission from participants (or from relatives in cases of fatal events) to review their medical records. Physicians who were un-aware of other questionnaire information used standardized criteria to confirm and classify the events. Deaths were ascertained from relatives, postal authorities, and the National Death Index, and the cause of death was classified on the basis of medical records, death certificates, and autopsy findings. Permission to review medical records was granted for 95% of the requests.

A diagnosis of myocardial infarction was con-firmed on the basis of standardized criteria, which included typical symptoms plus either diagnostic electrocardiographic changes or elevated cardiac enzyme levels.24,25 Deaths were ascertained by contact with family members or through the Na-tional Death Index. Fatal heart disease was con-firmed on the basis of medical records or autopsy reports or, if heart disease was listed as the cause of death, on the basis of the death certifi-cates and evidence of previous heart disease in the records. Stroke was diagnosed according to





standard criteria, consisting of a constellation of neurologic deficits of sudden or rapid onset that lasted at least 24 hours or until death.23,26 Stroke subtypes were also classified as previously de-scribed23,26 (see the Supplementary Appendix, available with the full text of this article at NEJM.org).

For each case participant, a control participant was selected randomly from those with stored toenail samples who were free of cardiovascular disease at the time of the case event (risk-set sampling). Controls were matched one to one with case subjects according to age (within 1 year), sex (cohort-specific), race, smoking status (cur-rent smoker, former smoker [matched on number of years since stopping], or never smoked), and month when toenail sample was returned to us.

Mercury and Selenium ExposuresTotal mercury and selenium concentrations were assessed in the stored toenails by means of neu-tron-activation analysis (University of Missouri Research Reactor). Details of the analytic meth-ods used and information regarding validation of these measures are provided in the Supplemen-tary Appendix.

Covariate Data CollectionData on demographic characteristics, risk factors, and lifestyle habits were collected by means of validated, self-administered questionnaires, with the use of the closest preceding questionnaire ad-ministered before the collection of toenail samples from each participant. Smoking status was as-sessed, including the number of years since quit-ting in the case of former smokers. Hypertension and hypercholesterolemia were self-reported, with the validity of these reports confirmed on ran-dom sampling of medical records. A supplemen-tary questionnaire was used to confirm self-re-ported cases of diabetes according to established criteria,27 and 98% of these cases were validated on comparison with medical records. Information on weight and height was obtained; self-reported weight was validated against technician-measured weight (r = 0.96). Physical activity was assessed in terms of metabolic equivalents (METs) with the use of validated questionaires.28 Usual die-tary habits were assessed by means of validated semiquantitative food-frequency questionnaires that inquired about usual consumption of foods, beverages, and supplements during the previous year.29,30

Statistical AnalysisAssociations of mercury concentrations with in-cident cardiovascular disease were evaluated with the use of multivariate-adjusted conditional logis-tic regression. Given risk-set sampling, this model provides a direct estimation of the hazard ratio (hereafter referred to as relative risk). Mercury concentrations were evaluated in quintiles as in-dicator variables, with the use of sex-specific cut-off points among controls. Tests for trend in-volved assigning participants the median value in their quintile of exposure and evaluating this as a continuous variable. Tests for interaction involved multiplying this variable by the effect modifier of interest and using the Wald test to calculate the P value associated with the multiplicative interac-tion term. A potential nonlinear dose–response relationship was evaluated by visual inspection of relative risks across deciles of exposure. Analyses were performed separately in each cohort and then combined on the basis of the absence of significant effect modification (multiplicative in-teraction) by sex (P≥0.05). Power calculations are provided in the Supplementary Appendix.

Potential confounding was assessed with the use of multivariate models adjusted for match-ing characteristics, other major risk factors for cardiovascular disease, fish or n−3 fatty acid con-sumption, and additional dietary factors associ-ated with mercury concentrations. Multivariate modeling was guided by the principle of parsi-mony and by the clinical relevance of covariates, the observed strength of association between co-variates and exposure or outcome, and the per-cent change in the risk estimate when covariates were included. Missing data for covariates (which accounted for less than 1% of all data) were imputed by means of multiple imputation.31

We performed prespecified sensitivity analy-ses to minimize potential misclassification due to exposure changes over time, restricting analy-ses to events within 10 years of toenail sampling and to participants with no substantial change in their fish consumption (i.e., a change of no more than two quintiles in either direction) from baseline to the end of follow-up. Stratified sub-group analyses were performed with the use of unconditional logistic regression adjusted for matching factors and other covariates.

All reported P values are two-tailed, with val-ues less than 0.05 indicating statistical signifi-cance. All analyses were performed with the use of SAS software, version 9.1 (SAS Institute).





R esult s

Study PopulationWe identified 3427 participants with incident cas-es of cardiovascular disease: 1532 nonfatal myo-cardial infarctions, 831 fatal cases of coronary heart disease, and 1064 strokes. These case par-ticipants were matched with 3427 controls who had not had cardiovascular disease events during the same period of follow-up. The median follow-up interval from the time of toenail sampling to the time of a cardiovascular disease event was 11.3 years (interquartile range, 6.4 to 15.3); follow-up time was identical for controls, based on the risk-set sampling method.

Baseline characteristics are shown in Table 1. As expected, cardiovascular risk factors were more prevalent among case participants than among controls at baseline. Approximately two thirds of the study participants were women, reflecting the larger size of the NHS cohort as compared with the HPFS cohort and the exclusion of dentists in the HPFS cohort from the analysis. Mean (±SD) ages were 61.1±9.0 years for men and 53.8±6.1 years for women. Median toenail mercury con-centrations were 0.30 µg per gram (interdecile range, 0.07 to 1.26) in case participants and 0.31 µg per gram (interdecile range, 0.07 to 1.31) in controls among men and 0.21 µg per gram (interdecile range, 0.06 to 0.77) in case partici-pants and 0.23 µg per gram (interdecile range, 0.07 to 0.76) in controls among women.

Mercury Exposure and Cardiovascular Risk Factors

Mercury concentrations correlated modestly with fish consumption (r = 0.39, P<0.001) and with es-timated dietary intake of eicosapentaenoic acid and docosahexaenoic acid (EPA–DHA) (r = 0.39, P<0.001), as expected, given the predominance of seafood as a source of methylmercury exposure but also given the considerable variation in methyl-mercury and n−3 fatty acid content among fish species.1,3 Concentrations of mercury did not correlate with those of selenium (r = 0.03), a find-ing that is consistent with the multiple, varied dietary sources of selenium.

In bivariate (unadjusted) analyses at baseline among the controls, higher mercury concentra-tions were associated with a more frequent preva-lence of hypercholesterolemia, slightly lower body-mass index, modestly higher levels of physical activity, greater alcohol use, and lower total en-

ergy intake (Table 1 in the Supplementary Appen-dix). Mercury concentrations were also positively associated with dietary factors related to fish consumption and higher dietary intake of EPA–DHA, including slightly lower intakes of saturated fat, monounsaturated fat, trans fat, and dietary cholesterol and slightly higher intakes of protein and polyunsaturated fat. Mercury concentrations were not significantly associated with age, smok-ing status, family history, or presence or absence of hypertension or diabetes.

Mercury Exposure and Cardiovascular EventsAfter adjustment for matching factors, partici-pants with higher mercury exposure did not have a higher risk of cardiovascular events (Table 2). In fact, those with higher mercury concentrations had a lower incidence of coronary heart disease (P = 0.006 for trend), stroke (P = 0.09 for trend), and total cardiovascular disease (P = 0.002 for trend). These inverse associations were not sig-nificant after further adjustment for other car-diovascular disease risk factors plus estimated dietary EPA–DHA (Table 2). Further adjustment for consumption of saturated fat, monounsatu-rated fat, polyunsaturated fat, trans fat, dietary cholesterol, and total energy had little effect on the results: the adjusted relative risks for com-parison of the fifth quintile of mercury exposure with the first quintile (“extreme-quintile relative risks”) were 0.85 (95% confidence interval [CI], 0.69 to 1.06) for coronary heart disease, 0.83 (95% CI, 0.60 to 1.15) for stroke, and 0.87 (95% CI, 0.73 to 1.03) for total cardiovascular disease. Ad-justment for fish consumption instead of dietary EPA–DHA also did not alter the findings (data not shown). The results were also similar for mercury concentrations evaluated in deciles (Table 2 in the Supplementary Appendix). In separate analyses according to sex, the trend toward a lower inci-dence of cardiovascular disease with higher mer-cury concentrations was seen for women but not for men (Table 3 in the Supplementary Appendix). Interaction tests for sex, however, were not sig-nificant (P = 0.12, P = 0.14, and P = 0.05 for tests of interaction with coronary heart disease, stroke, and total cardiovascular disease, respectively).

When coronary heart disease subtypes were evaluated, mercury exposure was not associated with the risk of nonfatal myocardial infarction (extreme-quintile relative risk, 0.84 [95% CI, 0.65 to 1.08]; P = 0.10 for trend) or fatal coronary heart disease (extreme-quintile relative risk, 0.85





[95% CI, 0.59 to 1.24]; P = 0.41 for trend). Mer-cury exposure was also not associated with the risk of any stroke subtype (see the Supplemen-tary Appendix).

Sensitivity AnalysesBecause selenium above a threshold of risk may provide protection against some forms of mercury toxicity, we restricted analyses to participants with lower selenium concentrations. Mercury ex-

posure was not associated with a higher risk of total cardiovascular disease, coronary heart dis-ease, or stroke among participants with selenium levels in the lowest quartile (<0.70 µg per gram) or the lowest decile (<0.64 µg per gram) (Table 3). Mercury exposure was also not associated with a higher risk in analyses stratified accord-ing to fish consumption (Table 4). Results were also similar in analyses stratified according to the presence or absence of hypertension, high

Table 1. Baseline Characteristics of Case Participants with Incident Cardiovascular Disease and of Controls.*

Characteristic Men Women

Case Participants(N = 1211)

Controls(N = 1211) P Value

Case Participants(N = 2216)

Controls(N = 2216) P Value

Age (yr)† 61.1±9.0 61.1±9.0 0.96 53.8±6.1 53.8±6.1 0.86

Smoking status (%)†

Never smoked 40.3 42.4 0.30 35.5 35.5 1.00

Former smoker 44.7 45.9 0.54 25.2 25.7 0.70

Current smoker 11.6 10.5 0.36 39.3 38.8 0.74

Family history of MI (%) 39.0 34.1 0.01 27.4 20.6 <0.001

Hypertension (%) 36.9 21.5 <0.001 13.5 8.1 <0.001

Hypercholesterolemia (%) 13.4 12.1 0.33 6.6 4.2 <0.001

Diabetes mellitus (%) 7.0 3.4 <0.001 3.0 0.5 <0.001

Body-mass index‡ 26.3±3.3 25.5±3.0 0.89 25.9±5.7 24.6±4.7 <0.001

Physical activity (METS/wk) 15.8±21.3 19.4±26.4 <0.001 11.7±16.2 13.5±18.6 0.001

Alcohol (drinks/wk) 0.8±1.2 0.9±1.2 0.08 0.5±0.9 0.6±0.9 0.03

Toenail selenium (µg/g) 0.92±0.61 0.92±0.6 0.99 0.78±0.22 0.78±0.25 0.34

Toenail mercury (µg/g) 0.51±2.13 0.44±0.47 0.24 0.29±0.49 0.33±0.63 0.04

Dietary intake

Fish (servings/wk) 2.1±1.9 2.1±1.8 0.89 1.8±1.6 1.8±1.6 0.65

EPA and DHA (mg/wk) 270±239 264±220 0.49 184±162 184±151 0.89

Total energy intake (kcal/day) 2024±623 2063±640 0.13 1742±536 1727±530 0.38

Fat (% energy)

Total 32.5±6.4 32.6±6.3 0.72 34.8±6.4 34.6±6.4 0.22

Saturated 11.3±2.9 11.3±2.8 0.85 12.7±3.1 12.6±3.0 0.05

Monounsaturated 12.5±2.8 12.5±2.7 0.69 12.9±2.9 12.8±2.9 0.16

Polyunsaturated 5.8±1.6 5.8±1.5 0.42 6.3±1.8 6.4±1.8 0.14

Trans 1.3±0.5 1.3±0.5 0.78 1.9±0.7 1.9±0.6 0.12

Protein (% energy) 18.3±3.4 18.3±3.3 0.97 18.0±3.6 17.9±3.4 0.48

Cholesterol (mg/day) 314±153 320±159 0.32 312±138 308±141 0.40

Whole grains (g/day) 20.5±19.2 20.8±18.0 0.74 15.3±15.9 15.8±13.7 0.28

* Plus–minus values are means ±SD. DHA denotes docosahexaenoic acid, EPA eicosapentaenoic acid, METS metabolic equivalents, and MI myocardial infarction.

† Age and smoking status were matching factors.‡ The body-mass index is the weight in kilograms divided by the square of the height in meters.





cholesterol, or diabetes or, among women, use or nonuse of hormone-replacement therapy (data not shown). The results of additional sensitivity analy-ses are provided in the Supplementary Appendix.

Discussion

In our study, mercury exposure as assessed by an objective biomarker measurement was not asso-ciated with an increased risk of cardiovascular disease among men or women in two separate U.S. cohorts. An increased risk with greater mer-cury exposure was also not evident among par-ticipants with lower selenium concentrations, in

analyses restricted to the first 10 years of follow-up and analyses stratified according to the dura-tion of follow-up, or in analyses restricted to those participants without substantial changes in fish consumption over time and analyses stratified according to the level of fish consumption. These findings provide no support for clinically rele-vant adverse effects of typical levels of dietary methylmercury exposure on cardiovascular dis-ease in U.S. adults.

Higher mercury exposures were actually asso-ciated with trends toward lower cardiovascular disease risk, although these trends were not significant in the fully adjusted models. To our

Table 2. Relative Risk of Cardiovascular Disease, According to Quintile of Toenail Mercury, Among Case Participants and Matched Controls in Two Prospective Cohorts of Men and Women.*

VariableNo. of Case Participants Sex-Specific Quintile of Toenail Mercury

P Value for Trend

1 2 3 4 5

Mean mercury (µg/g) 0.09 0.17 0.25 0.38 0.95

Coronary heart disease 2363

No. of cases 542 506 446 450 419

Multivariate RR (95% CI)

Model 1† 1.00(reference)

0.97(0.81–1.15)

0.82(0.69–1.00)

0.81(0.68 – 0.97)

0.78(0.65–0.94)

0.006

Model 2‡ 1.00(reference)

1.00(0.83–1.20)

0.89(0.73–1.08)

0.87(0.72–1.06)

0.85(0.69–1.04)

0.10

Stroke 1064

No. of cases 233 226 209 209 187



0.91(0.70–1.19)

0.89(0.68–1.17)

0.94(0.72–1.23)

0.77(0.59–1.02)

0.09


0.95(0.72–1.26)

0.95(0.71–1.28)

0.98(0.73–1.31)

0.84(0.62–1.14)

0.27

Total cardiovascular disease 3427

No. of cases 775 732 655 659 606



0.95(0.82–1.10)

0.84(0.73–0.98)

0.85(0.74–0.99)

0.78(0.67–0.91)

0.002


0.98(0.84–1.15)

0.91(0.77–1.07)

0.91(0.77–1.07)

0.85(0.72–1.01)

0.06

* Values for quintiles represent mean mercury levels. Quintiles were not constructed with the data from men and women combined but were sex-specific, and the relative risks (RR) for each were then combined. CI denotes confidence interval.

† In Model 1, the RR is based on conditional logistic regression with risk-set sampling, in which the odds ratio directly estimates the hazard ratio or RR, with matching factors of age, sex, race, month of toenail receipt, and smoking status (never smoked, former smoker, or current smoker).

‡ In Model 2, the RR was further adjusted for body-mass index (quintiles), physical activity (metabolic equivalents per week, quintiles), alco-hol intake (drinks per week, quintiles), diabetes (yes or no), hypertension (yes or no), elevated cholesterol level (yes or no), and estimated dietary intake of eicosapentaenoic acid and docosahexaenoic acid (mg per week, quintiles).





knowledge, there is no biologic explanation for why mercury would induce cardiovascular bene-fits. These results plausibly reflect the extent to which mercury levels are an indirect, but none-theless objective, biomarker of fish consumption and its correlates and thus probably provide in-dependent information on how much fish a per-son consumes, even after adjustment for estimat-ed consumption. Trends toward lower risk with higher mercury exposure appeared to be con-fined to women, but this sex difference was not significant and is probably due to chance. Trends toward lower cardiovascular disease risk with higher mercury levels have also been seen in some prior studies.7,11 Of six prior studies of the relationship between mercury exposure and car-diovascular disease,6-11 only two showed positive

associations.6,7 The largest study (684 cases) in-cluded only nonfatal myocardial infarction and was retrospective,6 raising concern about possi-ble selection bias. A smaller, prospective study (282 cases) showed a positive association with total coronary events but without a clear dose–response relationship or significant associations with coronary or cardiovascular mortality.7 The remaining four studies were prospective and did not show significant associations; however, they included participants with occupational exposure to mercury vapor,8 the health effects of which differ from those of methylmercury12; they as-sessed erythrocyte mercury levels, which reflect a more recent exposure than do toenail or hair concentrations9; or they had small numbers of cases (<100).10,11 Several of the prior studies also

Table 3. Odds Ratios for Cardiovascular Disease (CVD) According to Quintile of Toenail Mercury in Case Participants with Lower Selenium Levels, for Men and Women Combined from Two Prospective Cohorts.

VariableNo. of Case Participants Sex-Specific Quintile of Toenail Mercury*

P Value for Trend

1 2 3 4 5

odds ratio (95% confidence interval)

Subjects in lowest quartile of selenium levels†

Coronary heart disease 631 1.00(reference)

0.94(0.65–1.37)

0.72(0.50–1.05)

0.71(0.48–1.05)

0.84(0.55–1.27)

0.46

Stroke 254 1.00(reference)

0.70(0.39–1.27)

0.88(0.49–1.57)

0.59(0.31–1.12)

0.40(0.20–0.79)

0.006

Total CVD 885 1.00(reference)

0.87(0.64–1.18)

0.78(0.58–1.07)

0.70(0.50–0.96)

0.68(0.48–0.96)

0.03

Subjects in lowest decile of selenium levels‡

Coronary heart disease 242 1.00(reference)

0.99(0.54–1.81)

0.74(0.40–1.36)

0.77(0.40–1.48)

0.79(0.40–1.57)

0.49

Stroke 111 1.00(reference)

1.02(0.39–2.69)

1.02(0.40–2.54)

0.81(0.28–2.32)

0.62(0.38–1.17)

0.30

Total CVD 353 1.00(reference)

0.94(0.57–1.55)

0.80(0.49–1.30)

0.78(0.46–1.34)

0.67(0.38–1.17)

0.14

* Quintile cutoff points are based on the overall control population (see Table 3 in the Supplementary Appendix). An unconditional logistic- regression model was used, as appropriate, for stratified subgroup analyses. Values were adjusted for age, sex, race, month of toenail re-ceipt, smoking status (never smoked, former smoker, or current smoker), body-mass index (quintiles), physical activity (metabolic equiva-lents per week, quintiles), alcohol use (drinks per week, quintiles), diabetes (yes or no), hypertension (yes or no), elevated cholesterol level (yes or no), and estimated dietary intake of eicosapentaenoic acid and docosahexaenoic acid (mg per week, quintiles).

† These subjects had selenium values below 0.70 µg per gram.‡ These subjects had selenium values below 0.64 µg per gram.





did not evaluate stroke6-8,11 or include women.6-8 The investigation we describe here was designed to overcome these limitations.

With respect to generalizability, it is impor-tant to consider how mercury exposures in the present study compare with those in prior studies and with average population exposures. In our highest exposure quintile, the median toenail mercury concentration was 0.68 µg per gram, and in our highest decile, 1.00 µg per gram, corresponding to hair concentrations of about 1.84 and 2.70 µg per gram, respectively, calcu-lated from a reported toenail-to-hair ratio of

mercury of about 0.37.32-35 These exposure levels are similar to those seen in two smaller studies, in which mercury levels were positively associ-ated with coronary heart disease risk,6,7 and are also similar to higher U.S. exposures (in the 95th percentile).36

Differences in population selenium levels have been hypothesized to explain discrepant find-ings of prior studies with respect to mercury and cardiovascular risk — in particular, a study from Finland.7 Before soil supplementation was begun in the 1980s, selenium levels in Finland were among the lowest in Europe (mean serum level,

Table 4. Odds Ratios for Total Cardiovascular Disease, According to Quintile of Toenail Mercury and Stratum of Fish Consumption, for Men and Women Combined from Two Prospective Cohorts.

Fish Consumption*No. of Case Participants Quintile of Toenail Mercury†

P Value for Trend

1 2 3 4 5

odds ratio (95% confidence interval)

Total

<1 serving/wk 1500 1.00(reference)

0.99(0.81–1.21)

0.91(0.73–1.13)

0.80(0.63–1.01)

0.90(0.69–1.18)

0.20

1 to <2 servings/wk 992 1.00(reference)

0.85(0.63–1.14)

0.98(0.73–1.32)

0.84(0.62–1.13)

0.74(0.54–1.02)

0.07

≥2 servings/wk 935 1.00(reference)

1.11(0.74–1.66)

0.79(0.54–1.17)

1.17(0.80–1.70)

0.96(0.66–1.39)

0.86

Tuna or other dark-meat fish

<1 serving/wk 2475 1.00(reference)

1.00(0.85–1.18)

0.93(0.78–1.10)

0.88(0.73–1.05)

0.93(0.76–1.13)

0.32

1 to <2 servings/wk 483 1.00(reference)

0.72(0.42–1.22)

0.69(0.42–1.15)

0.89(0.54–1.47)

0.58(0.35–0.95)

0.08


0.98(0.54–1.79)

0.81(0.46–1.44)

0.97(0.56–1.69)

0.81(0.47–1.39)

0.38

Other fish

<0.5 serving/wk 2121 1.00(reference)

0.99(0.83–1.19)

0.92(0.76–1.11)

0.87(0.71–1.06)

0.86(0.69–1.06)

0.10

0.5 to <1 servings/wk 932 1.00(reference)

0.91(0.66–1.25)

0.90(0.65–1.24)

0.88(0.64–1.22)

0.74(0.54–1.03)

0.06


1.05(0.54–2.64)

0.88(0.47–1.63)

1.29(0.71–2.35)

1.38(0.76–2.48)

0.08

* Total fish consumption was the sum of the consumption of tuna or other dark-meat fish and the consumption of other fish. Strata were set at logical cutoff points that provided reasonable numbers of cases per stratum.

† Quintile cutoff points are based on the overall control population (see Table 3 in the Supplementary Appendix). Thus, in every stratum of fish consumption, higher quintiles reflect subjects with similarly high mercury exposure. In the case of low fish consumption (e.g., <1 serving per week), higher quintiles would be consistent with more exclusive consumption of relatively contaminated fish (i.e., similar methylmercury exposure from fewer fish meals, indicating a greater proportion of more contaminated fish in the diet). Values are based on unconditional logistic regression, as appropriate, for stratified subgroup analyses and have been adjusted for age, sex, race, month of toenail receipt, smoking status (never, former, or current), body-mass index (quintiles), physical activity (metabolic equivalents per week, quintiles), alcohol (drinks per week, quintiles), diabetes (yes or no), hypertension (yes or no), elevated cholesterol (yes or no), and estimated dietary intake of eicosapentaenoic acid and docosahexaenoic acid (mg per week, quintiles). See Tables 5 and 6 in the Supplementary Appendix for stratified results for coronary heart disease and stroke, which were evaluated separately.





<70 µg per liter).37 In the Finnish mercury study, average serum selenium levels at baseline (from 1984 through 1989, after soil supplementation began) were higher (117 µg per liter)7 but still below average U.S. levels (138 µg per liter).38 In our study, we found no evidence of an increased risk with higher mercury levels, even among par-ticipants with selenium levels in the lowest de-cile (<0.64 µg per gram in toenails, approxi-mately equivalent to <91 µg per liter in serum39). We also found no evidence that mercury was harmful among participants in different strata of fish consumption, including those with low fish consumption, in whom higher mercury levels would suggest more exclusive consumption of mercury-contaminated fish.

Our analysis cannot exclude the possibility of mercury-related cardiovascular toxicity at higher exposures than those observed in our cohorts or in the setting of frank selenium deficiency, which would be rare in U.S. cohorts. Ecologic or small cross-sectional studies in more highly exposed populations in the Amazon,40 the Faroe Islands,32 and Asia41,42 suggest that methylmercury expo-sure may be associated with higher blood pres-sure or lower parasympathetic activity; ecologic evidence of an increased risk of clinical cardio-vascular events is lacking.43

Our analysis has potential limitations. Al-though toenail concentrations of mercury pro-vide an excellent biomarker of integrated, usual methylmercury exposure during the previous year, changes in dietary exposure over time could attenuate true relationships toward null. Toenail mercury concentration serves as a marker of fish consumption, and our findings may be partly confounded by the beneficial effects of fish in-take, despite adjustment for responses to the dietary questionnaire; this might account for

trends toward lower risk. Although the findings were similar in the two independent cohorts and there is little reason to believe that biologic effects of methylmercury in these populations would be different from those in the general population of women and men, these cohorts comprised large-ly white, educated U.S. adults, potentially limit-ing generalizability.

The absence of any association between mer-cury exposure and increased cardiovascular dis-ease risk in adults should not alter ongoing public health and policy efforts to reduce mercury con-tamination in fish and the environment, which could still have the potential to offset, at least in part, the net cardiovascular benefits of fish con-sumption. Our findings should also not alter advisories directed toward women who are or may become pregnant or who are nursing, since methyl mercury exposure from consumption of specific fish species could cause neurodevelop-mental harm, or at least partly offset the neuro-developmental benefits of fish consumption, in their children.

In summary, this prospective study of two large cohorts of men and women in the United States showed no evidence of a relationship be-tween mercury exposure and increased cardio-vascular disease risk.

Supported by grants from the National Institute of Environ-mental Health Sciences (R01-ES014433 and ES013692), the Na-tional Heart, Lung, and Blood Institute (HL34594, HL088521, and HL35464), and the National Cancer Institute (CA87969 and CA55075).


We thank the NHS and HPFS participants and our expert ad-visory panel for helpful comments on this research, including David W.K. Acheson, Managing Director, Food and Import Safety, Leavitt Partners (formerly associate commissioner for foods, Food and Drug Administration), and Rita Schoeny, senior science advisor, Office of Water, Environmental Protection Agency.

References

1. Mozaffarian D, Rimm EB. Fish in-take, contaminants, and human health: evaluating the risks and the benefits. JAMA 2006;296:1885-99.2. Mozaffarian D. Fish, mercury, seleni-um and cardiovascular risk: current evi-dence and unanswered questions. Int J Environ Res Public Health 2009;6:1894-916.3. Groth E III. Ranking the contribu-tions of commercial fish and shellfish varieties to mercury exposure in the Unit-ed States: implications for risk communi-cation. Environ Res 2010;110:226-36.

4. What you need to know about mercury in fish and shellfish: 2004 EPA and FDA advice for women who might become pregnant, women who are pregnant, nurs-ing mothers, young children. Washing-ton, DC: Food and Drug Administration, 2004. (http://www.cfsan.fda.gov/~dms/admehg3.html.)5. Yaginuma-Sakurai K, Murata K, Shi-mada M, et al. Intervention study on car-diac autonomic nervous effects of methyl-mercury from seafood. Neurotoxicol Teratol 2010;32:240-5.6. Guallar E, Sanz-Gallardo MI, van’t

Veer P, et al. Mercury, fish oils, and the risk of myocardial infarction. N Engl J Med 2002;347:1747-54.7. Virtanen JK, Voutilainen S, Rissanen TH, et al. Mercury, fish oils, and risk of acute coronary events and cardiovascular disease, coronary heart disease, and all-cause mortality in men in eastern Fin-land. Arterioscler Thromb Vasc Biol 2005; 25:228-33.8. Yoshizawa K, Rimm EB, Morris JS, et al. Mercury and the risk of coronary heart disease in men. N Engl J Med 2002; 347:1755-60.





9. Wennberg M, Bergdahl IA, Stegmayr B, et al. Fish intake, mercury, long-chain n-3 polyunsaturated fatty acids and risk of stroke in northern Sweden. Br J Nutr 2007;98:1038-45.10. Ahlqwist M, Bengtsson C, Lapidus L, Gergdahl IA, Schütz A. Serum mercury concentration in relation to survival, symptoms, and diseases: results from the prospective population study of women in Gothenburg, Sweden. Acta Odontol Scand 1999;57:168-74.11. Hallgren CG, Hallmans G, Jansson JH, et al. Markers of high fish intake are associated with decreased risk of a first myocardial infarction. Br J Nutr 2001;86: 397-404.12. Mercury Study Report to Congress. Washington, DC: Environmental Protec-tion Agency, 1997. (http://www.epa.gov/mercury/report.htm.)13. Rice DC. The US EPA reference dose for methylmercury: sources of uncertainty. Environ Res 2004;95:406-13.14. König A, Bouzan C, Cohen JT, et al. A quantitative analysis of fish consump-tion and coronary heart disease mortality. Am J Prev Med 2005;29:335-46.15. Joint FAO/WHO expert consultation on the risks and benefits of fish consump-tion — executive summary. Geneva: Food and Agriculture Organization of the United Nations, World Health Organization, 2010. (http://www.who.int/foodsafety/chem/meetings/jan2010/en/index.html.)16. Hu FB, Bronner L, Willett WC, et al. Fish and omega-3 fatty acid intake and risk of coronary heart disease in women. JAMA 2002;287:1815-21.17. Mozaffarian D, Ascherio A, Hu FB, et al. Interplay between different polyunsatu-rated fatty acids and risk of coronary heart disease in men. Circulation 2005; 111:157-64.18. Joshi A, Douglass CW, Kim HD, et al. The relationship between amalgam resto-rations and mercury levels in male den-tists and nondental health professionals. J Public Health Dent 2003;63:52-60.19. Garland M, Morris JS, Rosner BA, et al. Toenail trace element levels as biomark-ers: reproducibility over a 6-year period. Cancer Epidemiol Biomarkers Prev 1993; 2:493-7.20. MacIntosh DL, Williams PL, Hunter DJ, et al. Evaluation of a food frequency questionnaire–food composition approach

for estimating dietary intake of inorganic arsenic and methylmercury. Cancer Epi-demiol Biomarkers Prev 1997;6:1043-50.21. Longnecker MP, Stampfer MJ, Morris JS, et al. A 1-y trial of the effect of high-selenium bread on selenium concentra-tions in blood and toenails. Am J Clin Nutr 1993;57:408-13.22. Longnecker MP, Stram DO, Taylor PR, et al. Use of selenium concentration in whole blood, serum, toenails, or urine as a surrogate measure of selenium intake. Epidemiology 1996;7:384-90.23. Iso H, Rexrode KM, Stampfer MJ, et al. Intake of fish and omega-3 fatty acids and risk of stroke in women. JAMA 2001; 285:304-12.24. Rose GA, Blackburn H. Cardiovascu-lar survey methods. 2nd ed. World Health Organization monograph series no. 56. Geneva: World Health Organization, 1982.25. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial infarction rede-fined — a consensus document of The Joint European Society of Cardiology/American College of Cardiology Com-mittee for the redefinition of myocardial infarction. J Am Coll Cardiol 2000;36: 959-69.26. Walker AE, Robins M, Weinfeld FD. The National Survey of Stroke: clinical findings. Stroke 1981;12:I13-I44.27. National Diabetes Data Group. Classi-fication and diagnosis of diabetes melli-tus and other categories of glucose intol-erance. Diabetes 1979;28:1039-57.28. Hu FB, Stampfer MJ, Colditz GA, et al. Physical activity and risk of stroke in women. JAMA 2000;283:2961-7.29. Feskanich D, Rimm EB, Giovannucci EL, et al. Reproducibility and validity of food intake measurements from a semi-quantitative food frequency questionnaire. J Am Diet Assoc 1993;93:790-6.30. Salvini S, Hunter DJ, Sampson L, et al. Food-based validation of a dietary ques-tionnaire: the effects of week-to-week variation in food consumption. Int J Epi-demiol 1989;18:858-67.31. Schafer JL. Analysis of incomplete multivariate data. New York: Chapman and Hall, 1997.32. Choi AL, Weihe P, Budtz-Jørgensen E, et al. Methylmercury exposure and ad-verse cardiovascular effects in Faroese whaling men. Environ Health Perspect 2009;117:367-72.

33. Suzuki T, Watanabe S, Matsuo N. Comparison of hair with nail as index media for biological monitoring of mer-cury. Sangyo Igaku 1989;31:235-8.34. Morton J, Mason HJ, Ritchie KA, White M. Comparison of hair, nails and urine for biological monitoring of low level inorganic mercury exposure in den-tal workers. Biomarkers 2004;9:47-55.35. Ohno T, Sakamoto M, Kurosawa T, Dakeishi M, Iwata T, Murata K. Total mer-cury levels in hair, toenail, and urine among women free from occupational ex-posure and their relations to renal tubular function. Environ Res 2007;103:191-7.36. McDowell MA, Dillon CF, Osterloh J, et al. Hair mercury levels in U.S. children and women of childbearing age: reference range data from NHANES 1999-2000. En-viron Health Perspect 2004;112:1165-71.37. Varo P, Alfthan G, Ekholm P, Aro A, Koivistoinen P. Selenium intake and se-rum selenium in Finland: effects of soil fertilization with selenium. Am J Clin Nutr 1988;48:324-9.38. Bleys J, Navas-Acien A, Laclaustra M, et al. Serum selenium and peripheral arte-rial disease: results from the National Health and Nutrition Examination Survey, 2003-2004. Am J Epidemiol 2009;169:996-1003.39. Mason MM, Morris JS, Spate VL, et al. Comparison of whole blood, plasma and nails as monitors for the dietary intake of selenium. J Radioanal Nucl Chem 1998; 236:29-34.40. Fillion M, Mergler D, Sousa Passos CJ, Larribe F, Lemire M, Guimarães JR. A pre-liminary study of mercury exposure and blood pressure in the Brazilian Amazon. Environ Health 2006;5:29.41. Yorifuji T, Tsuda T, Kashima S, Takao S, Harada M. Long-term exposure to methylmercury and its effects on hyper-tension in Minamata. Environ Res 2010; 110:40-6.42. Lim S, Chung HU, Paek D. Low dose mercury and heart rate variability among community residents nearby to an indus-trial complex in Korea. Neurotoxicology 2010;31:10-6.43. Chan HM, Egeland GM. Fish con-sumption, mercury exposure, and heart diseases. Nutr Rev 2004;62:68-72.Copyright © 2011 Massachusetts Medical Society.



original article



Carbamazepine-Induced Toxic Effects and HLA-B*1502 Screening in Taiwan

Pei Chen, Ph.D., Juei-Jueng Lin, M.D., Chin-Song Lu, M.D., Cheung-Ter Ong, M.D., Peiyuan F. Hsieh, M.D., Chih-Chao Yang, M.D.,

Chih-Ta Tai, M.D., Shey-Lin Wu, M.D., Cheng-Hsien Lu, M.D., Yung-Chu Hsu, M.D., Hsiang-Yu Yu, M.D., Long-Sun Ro, M.D., Chung-Ta Lu, M.D., Chun-Che Chu, M.D.,

Jing-Jane Tsai, M.D., Yu-Hsiang Su, M.D., Sheng-Hsing Lan, M.D., Sheng-Feng Sung, M.D., Shu-Yi Lin, M.S., Hui-Ping Chuang, B.S.,

Li-Chen Huang, B.S., Ying-Ju Chen, M.S., Pei-Joung Tsai, M.S., Hung-Ting Liao, M.S., Yu-Hsuan Lin, M.S., Chien-Hsiun Chen, Ph.D.,

Wen-Hung Chung, M.D., Ph.D., Shuen-Iu Hung, Ph.D., Jer-Yuarn Wu, Ph.D., Chi-Feng Chang, Ph.D., Luke Chen, Ph.D., Yuan-Tsong Chen, M.D., Ph.D.,

and Chen-Yang Shen, Ph.D., for the Taiwan SJS Consortium*

The authors’ affiliations are listed in the Appendix. Address reprint requests to Dr. Shen or Dr. Y.-T. Chen at the Institute of Biomedical Sciences, Academia Sinica, 128 Academia Rd., Section 2, Nankang, Taipei 115, Taiwan, or at bmcys@ibms .sinica.edu.tw or [email protected] .edu.tw.

*Other members of the Taiwan Stevens–Johnson Syndrome (SJS) Consortium are listed in the Supplementary Appen-dix, available at NEJM.org.


A bs tr ac t

BackgroundCarbamazepine, an anticonvulsant and a mood-stabilizing drug, is the main cause of the Stevens–Johnson syndrome (SJS) and its related disease, toxic epidermal necrolysis (TEN), in Southeast Asian countries. Carbamazepine-induced SJS–TEN is strongly associated with the HLA B*1502 allele. We sought to prevent carbamaze-pine-induced SJS–TEN by using HLA-B*1502 screening to prospectively identify subjects at genetic risk for the condition.

MethodsFrom 23 hospitals in Taiwan, we recruited 4877 candidate subjects who had not taken carbamazepine. We genotyped DNA purified from the subjects’ peripheral blood to determine whether they carried the HLA-B*1502 allele. Those testing posi-tive for HLA-B*1502 (7.7% of the total) were advised not to take carbamazepine and were given an alternative medication or advised to continue taking their prestudy medication; those testing negative (92.3%) were advised to take carbamazepine. We interviewed the subjects by telephone once a week for 2 months to monitor them for symptoms. We used the estimated historical incidence of SJS–TEN as a control.

ResultsMild, transient rash developed in 4.3% of subjects; more widespread rash developed in 0.1% of subjects, who were hospitalized. SJS–TEN did not develop in any of the HLA-B*1502–negative subjects receiving carbamazepine. In contrast, the estimated historical incidence of carbamazepine-induced SJS–TEN (0.23%) would translate into approximately 10 cases among study subjects (P<0.001).

ConclusionsThe identification of subjects carrying the HLA-B*1502 allele and the avoidance of carbamazepine therapy in these subjects was strongly associated with a decrease in the incidence of carbamazepine-induced SJS–TEN. (Funded by the National Science Council of Taiwan and the Taiwan Drug Relief Foundation.)



HLA-B*1502 Screening and Carbamazepine ther apy


The Stevens–Johnson syndrome (SJS) and its related disease, toxic epidermal necrolysis (TEN), are two of the most seri-

ous adverse reactions caused by drugs. SJS is characterized by high fever, malaise, and a rap-idly developing, blistering exanthema of macular papules and target-like lesions, accompanied by mucosal involvement. This condition is associat-ed with a rate of death of approximately 5%. TEN has a similar presentation, with even more exten-sive skin detachment and a death rate of 25 to 35%.1 Carbamazepine, an anticonvulsant and specific analgesic agent for trigeminal neuralgia, is the most common cause of SJS–TEN in South-east Asian countries.2 We previously reported that carbamazepine-induced SJS–TEN is strongly asso-ciated with the HLA-B*1502 allele in Han Chinese populations.3 This association was subsequently confirmed in persons from Hong Kong, Malaysia, Thailand, and India and in descendants of im-migrants from Southeast Asia, regions in which the HLA-B*1502 allele is prevalent.4-9 Another HLA allele, HLA-A*3101, has been associated with carbamazepine-induced hypersensitivity re-actions. We have observed its association with the relatively mild maculopapular exanthema.10 More recently, a genomewide association study has shown an association between the HLA-A*3101 allele and SJS–TEN in Japanese persons,11 and a report in this issue of the Journal shows the association between this allele and a range of hypersensitivity reactions, including SJS–TEN, in persons of European descent.12

Among persons of Han Chinese descent, carbamazepine-induced SJS–TEN almost never occurs in noncarriers of the HLA-B*1502 allele, evidence that this allele is directly involved in the pathogenesis of the condition. Carbamazepine directly binds to HLA-B molecules on antigen-presenting T cells and contributes to cell death mediated by cytotoxic T cells in persons with SJS–TEN.13 HLA-B*1502 can directly present car-bamazepine to cytotoxic T cells without antigen processing. More important, carbamazepine-specific T-cell–mediated cytotoxicity is restricted to HLA-B*1502.14

The risk of carbamazepine-induced SJS–TEN is significantly higher among persons of Chi-nese origin who carry the HLA-B*1502 allele than among those who do not carry the allele (odds ratio, 1357; 95% confidence interval, 193 to

8838; P = 1.6×10−41).10 If HLA-B*1502 were used as a marker to predict carbamazepine-induced SJS–TEN, the test would have a high sensitivity (98.3%) and specificity (95.8%). On the basis of an incidence of carbamazepine-induced SJS–TEN of 0.25%, this allele would have a negative pre-dictive value of 99.9% and a positive predictive value of 5.6%. The use of HLA-B*1502 genotyp-ing to prevent carbamazepine-induced SJS–TEN in routine clinical practice thus seems warrant-ed. We conducted a study to determine whether prospective screening by means of HLA-B*1502 genotyping before deciding on carbamazepine treatment reduces the incidence of carbamaze-pine-induced SJS–TEN.

Me thods

Study DesignWe recruited subjects from 23 participating hos-pitals throughout Taiwan (see the Supplementary Appendix, available with the full text of this ar-ticle at NEJM.org). To ensure that all investiga-tional sites complied with all applicable regula-tions and protocol requirements, an independent contract research organization (CRO Service Divi-sion, Formosa Biomedical Technology) monitored the study.

We had 9 points of interaction with HLA-B*1502–negative subjects and 10 points of inter-action with HLA-B*1502 carriers: the first screen-ing visit, a second clinic visit for HLA-B*1502 carriers only, and weekly telephone interviews during the 2-month follow-up period. Subjects between the ages of 6 months and 99 years who had not previously received carbamazepine and who would normally have received it at the time of screening were invited to participate in the study.

We excluded subjects who had a history of carbamazepine allergy, those who had undergone bone marrow transplantation, and those who were not of Han Chinese descent. Status with respect to Han Chinese descent was determined with the use of a multiple-choice questionnaire that asked the subjects to identify the ethnic group of their parents and grandparents.

We prescribed and provided carbamazepine to all subjects at the time of the screening visit, but we requested that they defer taking the drug until we had obtained and relayed to them their





genetic test results. Blood samples were collect-ed and transferred to a central laboratory (Insti-tute of Biomedical Sciences, Academia Sinica) for HLA-B*1502 genotyping, and the results were reported to the participating physicians within 2 to 3 days.

We asked HLA-B*1502–positive subjects to return to the clinic within 3 days, at which point we explained to them the risk of carbamazepine-induced SJS–TEN and recommended alternative drugs or therapy. HLA-B*1502–negative subjects (who also were counseled about the risk of SJS–TEN) were started on carbamazepine. Because the onset of SJS–TEN occurs within 2 months after the initiation of carbamazepine therapy,10 we interviewed all subjects by telephone during the 2-month period after the screening visit (for HLA-B*1502–negative subjects) or after the sec-ond clinic visit (for HLA-B*1502 carriers) to moni-tor them for symptoms of adverse drug reac-tions, including SJS–TEN. We asked subjects to return to the hospital immediately to be evaluated by a dermatologist in the event that early symp-toms of SJS–TEN developed. We monitored all subjects (aside from those who were lost to fol-low-up) throughout the study.

We performed the study in accordance with Good Clinical Practice Standards and the provi-sions of the Declaration of Helsinki. The study protocol was approved by the research ethics committee at Academia Sinica and by the insti-tutional review board at each participating hos-pital. Written informed consent was obtained from all subjects or from parents or guardians for subjects who were under 21 years of age. The study was conducted in accordance with the pro-tocol.

Genotyping of HLA-B*1502We obtained 2 ml of whole blood from each sub-ject in a Monovettes tube, stored the sample at 4 to 12°C, and sent it to Academia Sinica on the same day that we obtained it. Genomic DNA was isolated with the use of the QIAamp DNA purifi-cation system (Qiagen). The presence or absence of the HLA-B*1502 allele was determined with the use of the PG1502 DNA detection kit (Pharmi-gene), according to the manufacturer’s instruc-tions. The kits are based on real-time polymerase-chain-reaction assay, with sequence-specific primers for HLAB*1502. To validate the results of the PG1502 DNA detection kit, the first 2000

samples were also tested in parallel by HLA sequence-specific oligonucleotide reverse line blot (Dynal Biotech). In each of these 2000 samples, the results that were obtained with the use of the PG1502 DNA kit were consistent with those ob-tained with the use of the line-blot assay.

Annual IncidenceThe estimated number of SJS–TEN cases was based on diagnostic code 695.1 for erythema multi-forme in the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). We obtained the number of subjects with this code from the National Health Insurance Research Database (NHIRD), provided by the Bureau of Na-tional Health Insurance of the Department of Health. We calculated the annual incidence of carbamazepine-induced SJS–TEN in Taiwan as the annual number of cases of SJS–TEN caused by carbamazepine divided by the annual number of new carbamazepine recipients. A previous 5-year retrospective study (1997–2002), in which a der-matologist reviewed the medical records of 700 subjects with ICD-9-CM code 695.1 in the Chang-Gung Memorial Hospital system, suggested that among 230 subjects with SJS–TEN, 60 had taken carbamazepine; SJS–TEN had developed in the remaining 170 subjects after they had received other drugs.2 We therefore assumed that 32.9% of subjects with ICD-9-CM diagnostic code 695.1 had SJS–TEN and that the disease would be caused by carbamazepine in 26.1% of these sub-jects. Thus, we calculated the annual number of cases of carbamazepine-induced SJS–TEN as the number of subjects with ICD-9-CM diagnostic code 695.1 multiplied by 32.9% (the estimated proportion of subjects in Taiwan with ICD-9-CM diagnostic code 695.1 who would have SJS or TEN) multiplied by 26.1% (the estimated propor-tion of cases of carbamazepine-induced SJS–TEN among all those with SJS–TEN in Taiwan).

Since it usually takes at least 14 days for SJS–TEN to develop after the initiation of carbamaz-epine treatment, we obtained data for years 2001, 2002, 2003, and 2004 from the NHIRD regard-ing the number of persons who received new prescriptions for at least 14 days of carbamaz-epine each year.

Statistical AnalysisOn the basis of the prevalence of the HLA-B*1502 allele (8%) in the Taiwanese population,3 we de-





4877 Patients were enrolled

4855 Were included in the study

22 Had a protocol violation

4483 (92.3%) Were HLA-B*1502–negative 372 (7.7%) Were HLA-B*1502–positive

283 Did not take carba-mazepine

5 Were lost to follow-up

125 Were lost to follow-up

4120 Took carbamazepine 367 Did not take carbamazepine(215 were given alternative medicines)

AUTHOR:

FIGURE:

RETAKE:

SIZE

4-C H/TLine Combo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

1st2nd3rd

Shen

1 of 1

ARTIST:

TYPE:

ts

03-24-11JOB: 36412 ISSUE:

6 col33p9

Figure 1. Enrollment and Outcomes.

Carbamazepine was prescribed and provided for all subjects at the time of the screening visit, but they were asked to defer taking the drug until the results of genetic testing were available to them. All subjects, regardless of HLA-B status, were followed for 2 months, with weekly telephone interviews.

termined that 4419 subjects would provide a power of 99% to detect a reduction in the inci-dence of carbamazepine-induced SJS–TEN from 0.25% (i.e., 25 cases per 10,000 new recipients2) to 0.03%. We used Fisher’s exact to compare the rate of carbamazepine-induced SJS–TEN in the prospective screening population with the his-torical incidence. All reported P values are two-tailed, and a P value of less than 0.05 was consid-ered to indicate statistical significance.

R esult s

SubjectsFrom July 2007 through April 2010, we enrolled 4877 subjects, of whom 4855 underwent geno-typing and were included in the 2-month follow-up (Fig. 1). There were about equal numbers of men and women, with a mean age of 56.5 years (range, 0.6 to 98.2) for all subjects (Table 1). In-dications for carbamazepine treatment included epilepsy (14.2% of subjects), neuralgia (54.1%),

diabetes-related neuropathic pain (11.7%), tinni-tus (3.6%), and bipolar disorder or other psychi-atric disorders (2.8%) (Table 1).

Screening for HLA-B*1502Of the 4855 subjects who were included in the study, 372 (7.7%) were found to carry the HLA-B*1502 allele and were advised not to take carba-mazepine. These subjects were prescribed alterna-tive medications or advised to continue taking their prestudy medication; all were monitored for adverse events. Of these subjects, 5 were lost to follow-up, 215 took an alternative medication, and 152 continued taking their pre study medi-cation (Fig. 1). Alternative medications included gabapentin, valproic acid, oxcarbazepine, imipra-mine, clonazepam, and lamotrigine. The remaining 4483 subjects (92.3%) were HLA-B*1502–nega-tive. Of these subjects, 238 did not take carbam-azepine and 125 were lost to follow-up, leaving 4120 HLA-B*1502–negative subjects who took the drug and were monitored (Fig. 1).





Adverse Events

Among all 4855 subjects, mild and transient rash and itching developed in 211 (4.3%), including 21 subjects who had a combination of rash, itching, and localized blisters and 26 subjects who had limited oral ulcers (Table 2). Of the 211 subjects with rash or itching, 5 were HLA-B*1502–posi-tive and presented with symptoms after taking alternative drugs (gabapentin, lamotrigine, naprox-en, imipramine, or prednisolone) (Table 2). In addition, 7 subjects had more severe cutaneous symptoms — maculopapular eruption (in 3 sub-jects), hypersensitivity syndrome (in 2 subjects), and urticaria (in 2 subjects) — for which they were hospitalized. One of the two subjects with urticaria was HLA-B*1502–positive and had tak-en oxcarbazepine before study enrollment. None of the subjects received a diagnosis of SJS–TEN.

Other adverse events included fever, sore throat, fatigue, dizziness, insomnia, and gastro-intestinal symptoms. These events occurred in both HLA-B*1502–positive and HLA-B*1502–negative subjects.

Historical IncidenceThe NHIRD data showed that carbamazepine was prescribed for at least 14 days for 50,917 per-sons in 2002, 48,522 in 2003, and 49,670 in 2004 who had not previously received carbamazepine, at least dating back to the beginning of the previ-ous calendar year (Table 3). We therefore calcu-

lated that the incidence of SJS–TEN in 2002, 2003, and 2004 in Taiwan was 0.24%, 0.22%, and 0.23%, respectively. We used the mean (0.23%) as the historical incidence for further analysis.

Incidence of SJS–TEN after Genetic ScreeningOn the basis of the estimated historical incidence of 0.23%, 10 cases of SJS or TEN would be ex-pected among the 4120 subjects who took carba-mazepine in our study. However, there were no cases of either SJS or TEN in any of the subjects, a result that differed significantly from the his-torical incidence (P<0.001 by Fisher’s exact test) (Table 3).

Discussion

Our findings suggest that screening patients for the HLA-B*1502 allele before the initiation of carbamazepine treatment and withholding car-bamazepine from HLA-B*1502–positive patients can reduce the incidence of carbamazepine- induced SJS–TEN among Han Chinese. In esti-mating the historical incidence of this condition, we defined new carbamazepine recipients as those who had not received carbamazepine dur-ing the previous year and who were prescribed carbamazepine for at least 14 days in the year of interest, because carbamazepine-induced SJS–TEN is a delayed hypersensitivity reaction that usually takes at least 14 days to develop. However, even if

Table 1. Characteristics of the Subjects.

CharacteristicHLA-B*1502–Positive

(N = 372)HLA-B*1502–Negative

(N = 4483)Total

(N = 4855)

Sex — no. (%)

Male 193 (51.9) 2132 (47.6) 2325 (47.9)

Female 179 (48.1) 2351 (52.4) 2530 (52.1)

Age — yr

Mean 55.7 56.5 56.5

Range 4.3–91.4 0.6–98.2 0.6–98.2

Indication for carbamazepine — no. (%)

Epilepsy 57 (15.3) 632 (14.1) 689 (14.2)

Neuralgia 195 (52.4) 2430 (54.2) 2625 (54.1)

Diabetes-related neuropathic pain 53 (14.2) 515 (11.5) 568 (11.7)

Tinnitus 8 (2.2) 168 (3.7) 176 (3.6)

Bipolar or other psychiatric disorder 12 (3.2) 122 (2.7) 134 (2.8)

Other conditions* 47 (12.6) 647 (14.4) 694 (14.3)

* These conditions include herpes zoster, unspecified cerebral-artery occlusion, meralgia paresthetica, and multiple sclerosis.





Table 2. Adverse Events during the 2-Month Follow-up.

Adverse Event

HLA-B*1502–Positive with Alternative Medication

(N = 215)

HLA-B*1502–Negative with Carbamazepine

(N = 4120) Total

number of events

Mild cutaneous events

Rash and itching 5* 206 211

Rash, itching, and blisters 1† 20 21

Rash, itching, and oral ulcers 0 14 14

Rash, itching, blisters, and oral ulcers 0 7 7

Itching, blisters, and oral ulcers 0 2 2

Blisters and oral ulcers 0 3 3

Severe cutaneous events

Maculopapular eruption 0 3 3

Hypersensitivity syndrome 0 2 2

Urticaria 1‡ 1 2

Stevens–Johnson syndrome or toxic epidermal necrolysis

0 0 0

Other adverse events§

Fever 1 92 93

Sore throat 4 126 130

Fatigue 16 818 834

Dizziness 10 497 507

Insomnia 5 197 202

Gastrointestinal symptoms 4 185 189

* Among these 5 subjects, the alternative drugs were gabapentin, lamotrigine, naproxen, imipramine, and prednisolone.† This subject had rash, itching, and blisters after taking gabapentin as an alternative treatment. These symptoms were

mild and disappeared in 7 days.‡ This subject had taken oxcarbazepine before study enrollment.§ Subjects may have had more than one adverse event. Adverse events with a low frequency are not listed.

we had included all new carbamazepine recipi-ents, regardless of the duration of treatment, as historical controls, the difference in the incidence of SJS–TEN would still be significant (P = 0.01, P = 0.02, and P = 0.02 for 2002, 2003, and 2004, respectively).

Because we estimated the historical incidence of carbamazepine-induced SJS–TEN on the basis of data obtained from the NHIRD, the reliability of these data is critical for the validity of our esti-mation. The NHIRD was established in Taiwan when the government launched the National Health Insurance system in 1995. This manda-tory single-payer health insurance system, which is administered by the Taiwanese government, provides health care for almost all people in Taiwan, with enrollment of 99.5% of the popula-tion in 2008. Of the health care facilities in Tai-wan, 92.5% have been contracted by the Na-

tional Health Insurance system. The NHIRD data were therefore likely to be comprehensive. To estimate both the percentage of subjects with SJS–TEN among those with ICD-9-CM diagnos-tic code 695.1 (indicating erythema multiforme) and the percentage of cases of carbamazepine-induced SJS–TEN in Taiwan, we based our review on the medical records of 700 cases with diag-nostic code 695.1 during a 5-year period at the Chang-Gung Memorial Hospital, the largest hos-pital system in Taiwan, with several regional centers. This hospital provides health care for about 12% of the Taiwanese population, and its patients are thought to be representative of the general population.

It is possible that some of the drug-related adverse reactions we observed were early SJS le-sions or that early withdrawal of carbamazepine may have prevented a more severe SJS–TEN or





TEN-like reaction. However, we think that this is unlikely, since once patients are sensitized by car-bamazepine and have early blisters or ulcers, SJS–TEN progresses, even after the withdrawal of the drug.15-17 SJS–TEN did not develop in any of the subjects who completed the 2-month follow-up.

Adverse cutaneous reactions, including blis-ters, oral lesions, and rash, that occurred in the subjects were mild, localized, and transient. In some subjects, blisters developed only after the rash subsided (i.e, blisters were sporadic and tended not to occur at the same time as rash). Furthermore, many HLA-B*1502–negative sub-jects resumed taking carbamazepine without a recurrence of skin lesions. More important, SJS–TEN did not develop in these subjects, a finding that is consistent with the concept that the inci-dence of carbamazepine-induced SJS–TEN in HLA-B*1502–negative persons is very low.

Our results suggest the value of HLA-B*1502 screening to prevent carbamazepine-induced SJS–TEN. However, as for any new pharmacoge-nomic test, it is important to document the use and safety of the alternative medications.18 Of the 367 HLA-B*1502 carriers, 215 (58.6%) were given an alternative medication, such as gabap-entin, valproic acid, oxcarbazepine, imipramine,

clonazepam, or lamotrigine; the remainder con-tinued to take their prestudy medication. Among the 215 HLA-B*1502 carriers who took alterna-tive drugs, the only symptom seen during the 2-month follow-up was mild, transient rash in 5 subjects (2.3%).

A strong association between HLA-B*1502 and carbamazepine-induced SJS–TEN has been found in Asian populations other than the Han Chinese,4-9 including Malay, Thai, and South Asian Indians. In Malaysia, Thailand, and India, studies have shown that carbamazepine was the major cause of drug-induced SJS–TEN. Since the contribution of HLA-B*1502 to carbamaze-pine-induced SJS–TEN has been proved to be causal,10,13,14,17 we speculate that in these coun-tries, in which HLA-B*1502 is relatively prev-alent, HLA-B*1502 screening could provide a benefit.

Supported by grants from the National Research Program for Genomic Medicine, National Science Council, Taiwan (for the National Clinical Core and the National Genotyping Core) and from the Academy–Industry Collaboration Program (NSC95-3112-B-001-023 and NSC97-3112-B-001-007); and the Taiwan Drug Relief Foundation (TDRF9707 and TDRF9809).


We thank the subjects who participated in this study and the research and nursing staff for their meticulous data collection.

AppendixThe authors’ affiliations are as follows: the Institute of Biomedical Sciences, Academia Sinica, Taipei (P.C., S.-Y.L., H.-P.C., L-.C.H., Y.-J.C., P-.J.T., H.-T.L., Y.-H.L., C.-H.C., S.-I.H., J.-Y.W., Y.-T.C., C.-Y.S.); Chu Shang Show Chwan Hospital, Nantou (J.-J.L.); the De-partments of Neurology (C.-S.L, L.-S.R., C.-C.C.) and Dermatology (W.-H.C.), Chang-Gung Memorial Hospital at Linkou and College of Medicine, Chang-Gung University, Kweishan; Chia-Yi Christian Hospital, Chia-Yi (C.-T.O., Y.-C.H., Y.-H.S., S.-F.S.); Taichung Vet-erans General Hospital, Taichung, and National Chi Nan University, Nantou (P.-F.H.); National Taiwan University Hospital, Taipei (C.-C.Y.); Kaohsiung Medical University Chung-Ho Memorial Hospital and Kaohsiung Medical University, Kaohsiung (C.-T.T., S.-H.L.); Changhua Christian Hospital, Changhua (S.-L.W.); Chang-Gung Memorial Hospital at Kaohsiung, Kaohsiung (C-.H.L.); Neurology Department, Taipei Veterans General Hospital and National Yang-Ming University, Taipei (H.-Y.Y.); College of Public Health (C.-Y.S.) and China Medical University Hospital (C.-T.L.), China Medical University, Taichung; National Cheng Kung University Hospital, Tainan (J.-J.T.); Pharmigene (C.-F.C., L.C.); and the Department of Pediatrics, Duke University Medical Center, Durham, NC (Y.-T.C.).

Table 3. Historical Incidence of Carbamazepine-Induced SJS–TEN in 2002, 2003, and 2004, as Compared with the Incidence among Study Subjects.*

Variable 2002 2003 2004

New recipients of carbamazepine (no.) 50,917 48,522 49,670

Subjects with ICD-9-CM diagnostic code 695.1 (no.) 1441 1261 1354

Carbamazepine-induced SJS–TEN (no.) 123 108 116

Incidence of carbamazepine-induced SJS–TEN (%) 0.24 0.22 0.23

P value for comparison between historical incidence and incidence among study subjects†

<0.001 <0.001 <0.001

* ICD-9-CM denotes International Classification of Diseases, 9th Revision, Clinical Modification, and SJS–TEN Stevens–Johnson syndrome and toxic epidermal necrolysis.

† All P values were calculated with the use of Fisher’s exact test.





References

MY NEJM IN THE JOURNAL ONLINE

Individual subscribers can store articles and searches using a feature on the Journal’s Web site (NEJM.org) called “My NEJM.”

Each article and search result links to this feature. Users can create personal folders and move articles into them for convenient retrieval later.

1. Roujeau JC, Stern RS. Severe adverse cutaneous reactions to drugs. N Engl J Med 1994;331:1272-85.2. Hung SI, Chung WH, Chen YT. HLA-B genotyping to detect carbamazepine- induced Stevens-Johnson syndrome: im-plications for personalized medicine. Per-sonalized Medicine 2005;2:225-37.3. Chung WH, Hung SI, Hong HS, et al. Medical genetics: a marker for Stevens-Johnson syndrome. Nature 2004;428:486.4. Man CB, Kwan P, Baum L, et al. As-sociation between HLA-B*1502 allele and antiepileptic drug-induced cutaneous re-actions in Han Chinese. Epilepsia 2007;48: 1015-8. [Erratum, Epilepsia 2008;49:941.]5. Lim KS, Kwan P, Tan CT. Association of HLA-B*1502 allele and carbamazepine-induced severe adverse cutaneous drug reaction among Asians, a review. Neurol Asia 2008;13:15-21.6. Locharernkul C, Loplumlert J, Limo-tai C, et al. Carbamazepine and phenytoin induced Stevens-Johnson syndrome is as-sociated with HLA-B*1502 allele in Thai population. Epilepsia 2008;49:2087-91. [Erratum, Epilepsia 2009;50:971.]7. Tassaneeyakul W, Tiamkao S, Janta-raroungtong T, et al. Association between HLA-B*1502 and carbamazepine-induced

severe cutaneous adverse drug reactions in a Thai population. Epilepsia 2010;51: 926-30.8. Mehta TY, Prajapati LM, Mittal B, et al. Association of HLA-B*1502 allele and carbamazepine-induced Stevens-Johnson syndrome among Indians. Indian J Der-matol Venereol Leprol 2009;75:579-82.9. Lonjou C, Thomas L, Borot N, et al. A marker for Stevens-Johnson syndrome: ethnicity matters. Pharmacogenomics J 2006;6:265-8.10. Hung SI, Chung WH, Jee SH, et al. Ge-netic susceptibility to carbamazepine-induced cutaneous adverse drug reactions. Pharmacogenet Genomics 2006;16:297-306.11. Ozeki T, Mushiroda T, Yowang A, et al. Genome-wide association study identi-fies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutane-ous adverse drug reactions in Japanese pop-ulation. Hum Mol Genet 2011;20:1034-41.12. McCormack M, Alfirevic A, Bourgeois S, et al. HLA-A*3101 and carbamazepine-induced hypersensitivity reactions in Eu-ropeans. N Engl J Med 2011;364:1134-43.13. Yang CW, Hung SI, Juo CG, et al. Hu-man leukocyte antigen-B*1502-bound pep-tides: implications for the pathogenesis

of carbamazepine-induced Stevens-John-son syndrome. J Allergy Clin Immunol 2007;120:870-7.14. Wei CY, Huang HW, Chen YT, Hung SI. The role of HLA-B*1502 in carbamaze-pine-induced Stevens-Johnson syndrome. In: Programs and abstracts of the Taiwan-ACGA 2010 International Conference on Genetic and Genomic Medicine, Taipei, May 2–5, 2010.15. Garcia-Doval I, LeCleach L, Bocquet H, Otero XL, Roujeau JC. Toxic epidermal necrolysis and Stevens-Johnson syndrome: does early withdrawal of causative drugs decrease the risk of death? Arch Dermatol 2000;136:323-7.16. Fritsch PO, Sidoroff A. Drug-induced Stevens-Johnson syndrome/toxic epider-mal necrolysis. Am J Clin Dermatol 2000; 1:349-60.17. Chung WH, Hung SI, Yang JY, et al. Granulysin is a key mediator for dissemi-nated keratinocyte death in Stevens-John-son syndrome and toxic epidermal necrol-ysis. Nat Med 2008;14:1343-50.18. Flockhart DA, Skaar T, Berlin DS, Klein TE, Nguyen AT. Clinically available pharmacogenomics tests. Clin Pharmacol Ther 2009;86:109-13.Copyright © 2011 Massachusetts Medical Society.



original article



HLA-A*3101 and Carbamazepine-Induced Hypersensitivity Reactions in Europeans

Mark McCormack, B.A., Ana Alfirevic, M.D., Ph.D., Stephane Bourgeois, Ph.D., John J. Farrell, M.S., Dalia Kasperavi i t , Ph.D., Mary Carrington, Ph.D.,

Graeme J. Sills, Ph.D., Tony Marson, M.B., Ch.B,, M.D., Xiaoming Jia, M.Eng., Paul I.W. de Bakker, Ph.D., Krishna Chinthapalli M.B., B.S.,

Mariam Molokhia, M.B., Ch.B., Ph.D., Michael R. Johnson, D.Phil., Gerard D. O’Connor, M.R.C.P.I., Elijah Chaila, M.R.C.P.I.,

Saud Alhusaini, M.B., Kevin V. Shianna, Ph.D., Rodney A. Radtke, M.D., Erin L. Heinzen, Ph.D., Nicole Walley, B.S., Massimo Pandolfo, M.D., Ph.D., Werner Pichler, M.D., B. Kevin Park, Ph.D., Chantal Depondt, M.D., Ph.D.,

Sanjay M. Sisodiya, M.D., Ph.D., David B. Goldstein, Ph.D., Panos Deloukas, Ph.D., Norman Delanty, B.M., Gianpiero L. Cavalleri, Ph.D.,

and Munir Pirmohamed, Ph.D., F.R.C.P.

The authors’ affiliations are listed in the Appendix. Address reprint requests to Dr. Alfirevic at the Wolfson Centre for Personalised Medicine, Department of Molecular and Clinical Pharmacology, In-stitute of Translational Medicine, Univer-sity of Liverpool, Block A: Waterhouse Bldgs., 1-5 Brownlow St., Liverpool L69 3GL, United Kingdom, or at [email protected].

Mr. McCormack, Dr. Alfirevic, Dr. Cavalleri, and Dr. Pirmohamed contributed equally to this article.


A bs tr ac t

BackgroundCarbamazepine causes various forms of hypersensitivity reactions, ranging from maculopapular exanthema to severe blistering reactions. The HLA-B*1502 allele has been shown to be strongly correlated with carbamazepine-induced Stevens–Johnson syndrome and toxic epidermal necrolysis (SJS–TEN) in the Han Chinese and other Asian populations but not in European populations.

MethodsWe performed a genomewide association study of samples obtained from 22 subjects with carbamazepine-induced hypersensitivity syndrome, 43 subjects with carbamaze-pine-induced maculopapular exanthema, and 3987 control subjects, all of European descent. We tested for an association between disease and HLA alleles through proxy single-nucleotide polymorphisms and imputation, confirming associations by high-resolution sequence-based HLA typing. We replicated the associations in samples from 145 subjects with carbamazepine-induced hypersensitivity reactions.

ResultsThe HLA-A*3101 allele, which has a prevalence of 2 to 5% in Northern European populations, was significantly associated with the hypersensitivity syndrome (P = 3.5×10−8). An independent genomewide association study of samples from sub-jects with maculopapular exanthema also showed an association with the HLA-A*3101 allele (P = 1.1×10−6). Follow-up genotyping confirmed the variant as a risk factor for the hypersensitivity syndrome (odds ratio, 12.41; 95% confidence interval [CI], 1.27 to 121.03), maculopapular exanthema (odds ratio, 8.33; 95% CI, 3.59 to 19.36), and SJS–TEN (odds ratio, 25.93; 95% CI, 4.93 to 116.18).

ConclusionsThe presence of the HLA-A*3101 allele was associated with carbamazepine-induced hypersensitivity reactions among subjects of Northern European ancestry. The presence of the allele increased the risk from 5.0% to 26.0%, whereas its absence reduced the risk from 5.0% to 3.8%. (Funded by the U.K. Department of Health and others.)



HLA-A*3101 and Carbamazepine-Induced Hypersensitivity


Carbamazepine is one of the most commonly prescribed drugs for the treat-ment of epilepsy, as well as trigeminal neu-

ralgia and bipolar disorder. A minority of treated persons have hypersensitivity reactions that vary in prevalence and severity,1 with some forms as-sociated with substantial morbidity and mortal-ity. The mildest form, maculopapular exanthema, occurs in 5 to 10% of treated persons of Euro-pean ancestry and resolves spontaneously after drug discontinuation. More severe reactions, such as the hypersensitivity syndrome, are associated with mortality of up to 10%2 and include symp-toms such as rash, fever, eosinophilia, hepatitis, and nephritis. The most severe reactions, such as the Stevens–Johnson syndrome (SJS) and toxic epi-dermal necrolysis (TEN), are characterized by a blistering rash affecting a variable percentage of the body-surface area. The rate of death increases with the degree of epidermal detachment. TEN is the rarest of these phenotypes and is associated with mortality of up to 30%. According to the labeling of carbamazepine, as mandated by the Food and Drug Administration (FDA), the estimat-ed incidence of SJS–TEN is 1 to 6 cases in 10,000 persons of European ancestry who are exposed to the drug.

Genomewide approaches are increasingly used to identify genetic predisposing factors for drug-induced hypersensitivity reactions and drug-induced liver injury.3 For example, the HLA-B*1502 allele has been identified as a clinically important predictor of SJS–TEN in Asians of Han Chinese descent who are candidates for treat-ment with carbamaz e pine.4 This finding led the FDA to require a warning label for carbamazepine to indicate the need for genotyping for HLA-B*1502 before the drug is prescribed. This change seems warranted, given the prospective study described in this issue of the Journal,5 which shows that withholding carbamazepine from subjects of Han Chinese ancestry who carry the HLA-B*1502 allele substantially decreased the rate of the development of SJS–TEN. The value of HLA-B*1502 in predicting the risk of SJS–TEN has been shown in other Asian populations, in-cluding those of Thailand, Malaysia, and India.4,6-8 This effect is specific for SJS–TEN, as compared with the full spectrum of carbamazepine-related hypersensitivity reactions, which led to the sug-gestion that different phenotypes of carbamaze-

pine-induced hypersensitivity may have distinct genetic predictors. Furthermore, HLA-B*1502 is rare (with a prevalence of less than 2%) in popu-lations of European descent, in which carbamaz-epine-induced SJS–TEN occurs at a lower inci-dence than in Asian populations.9-12 Using a candidate-gene approach, previous studies have shown weak or moderate signals of association between maculopapular exanthema or the DRESS (drug reaction with eosinophilia and systemic symptoms) syndrome and single-nucleotide poly-morphisms (SNPs) in the major histocompatibil-ity complex (MHC) region,13-15 and these asso-ciations have not been consistently replicated.13,14

We present data from two independent studies that suggest the relevance of an HLA variant to the clinical spectrum of carbamazepine-related hypersensitivity reactions, including maculopap-ular exanthema, the hypersensitivity syndrome, and SJS–TEN.

Me thods

Case SubjectsRecruitmentWe recruited case subjects at centers collaborat-ing with the University of Liverpool and Walton Centre for Neurology (both hereinafter referred to as “the Liverpool collaborators”) or at centers affiliated with the EPIGEN consortium (for de-tails, see the Supplementary Appendix, available with the full text of this article at NEJM.org). All case subjects and control subjects were of Euro-pean ancestry as determined by either self-report or genetic-marker analysis.13 Many of these sub-jects have been described previously in the con-text of hypersensitivity reactions.9,13,16

Hypersensitivity SyndromeThe group with the hypersensitivity syndrome consisted of 26 subjects: 23 who were recruited through the Liverpool collaborators and 3 who were recruited through the EPIGEN consortium. The hypersensitivity syndrome was defined as the presence of rash or liver involvement within 3 months after the initiation of carbamazepine treatment, accompanied by a minimum of two of the following manifestations: a prolonged recov-ery phase, despite withdrawal of the drug; fever; involvement of other internal organs (liver, kidney, lung, central nervous system, heart, muscle, thy-





roid, or lymphoid system); or the presence of he-matologic abnormalities, such as eosinophilia and atypical lymphocytosis.17 We considered the DRESS syndrome to be the same as the hyper-sensitivity syndrome.

Acute Generalized Exanthematous PustulosisWe recruited one subject with acute generalized exanthematous pustulosis (AGEP), who had a typical pustular reaction with high fever,18 from a collaborating clinic in Switzerland. We then analyzed the data from this subject together with data from the subjects with the hypersensitivity syndrome.

Maculopapular ExanthemaThe group with maculopapular exanthema con-sisted of 106 subjects: 57 who were recruited through the Liverpool collaborators and 49 who were recruited through the EPIGEN consortium. Maculopapular exanthema was defined as rash without systemic symptoms that required the dis-continuation of carbamazepine within 3 months after the initiation of drug therapy.

SJS–TENThe SJS–TEN group consisted of 12 subjects: 8 who were recruited through the Liverpool collaborators and 4 who were recruited through the EPIGEN consortium. The diagnostic criteria suggested by the RegiSCAR Project (a research network for se-vere cutaneous adverse drug reactions)11,19 were used for this group (see the Supplementary Ap-pendix).

Control SubjectsPopulation ControlsGenotype data from the Wellcome Trust Case Control Consortium (WTCCC) (www.wtccc.org .uk) were used as a population control group. The Liverpool collaborators used data from a genome-wide association study of samples from 2691 subjects in the U.K. National Blood Services Col-lection. EPIGEN used samples from a homoge-neous subgroup of 1296 subjects in the 1958 Brit-ish Birth Cohort that were selected by principal component analysis. We did not screen either set of population controls for carbamaze pine-related adverse drug reactions. (For details about our access to these data, see the Supplementary Ap-pendix.)

Clinical ControlsThe clinical control group consisted of 257 sub-jects with epilepsy who had been taking carba-mazepine for at least 3 months with no clinical or biochemical signs of hypersensitivity. Of these subjects, 44 were recruited by the Liverpool col-laborators and 213 were recruited from clinics affiliated with EPIGEN.

Genomewide GenotypingGenomewide analysis of samples that were ini-tially available from the 22 Liverpool subjects with carbamazepine-induced hypersensitivity syn-drome was performed with the use of the Illu-mina Infinium 1.2M chip at the Wellcome Trust Sanger Institute. Samples from 43 case subjects and 200 control subjects from the EPIGEN con-sortium were genotyped with the use of the Illu-mina 610K Quad platform at the Duke University Center for Human Genome Variation. (Details regarding quality controls and principal-compo-nents analysis are provided in the Supplementary Appendix.)

Imputation of HLA AllelesWe based the imputation of classic HLA alleles in samples from the 22 Liverpool subjects with the hypersensitivity syndrome and control subjects from the U.K. National Blood Service on a dense reference panel of SNP data and four-digit HLA types in 2767 unrelated subjects of European de-scent, according to a method described recently.20 We imputed classic HLA alleles in the samples from the EPIGEN consortium, using MACH 1.0 on the Web site of the Center for Statistical Genetics (www.sph.umich.edu/csg/abecasis/MACH).21 The 1458 subjects in the HLA-typed 1958 British Birth Cohort constituted the reference population. (See the Supplementary Appendix for further details.) We used different approaches to impute the HLA alleles in the Liverpool and EPIGEN subjects, because the two studies were originally inde-pendent.

HLA-A High-Resolution GenotypingTo confirm the accuracy of imputation, high-resolution, sequence-based HLA-A typing of sam-ples from 22 subjects with the hypersensitivity syndrome and from a subgroup of 44 clinical control subjects (i.e., subjects who did not have adverse reactions to carbamazepine) was per-





formed by Histogenetics. We determined the HLA-A alleles in samples from 49 subjects with maculopapular exanthema and a subgroup of 213 clinical control subjects from the EPIGEN consor-tium, using sequence-specific polymerase-chain-reaction assay primers (details available on request). Direct genotyping showed 100% concordance with the imputed alleles.

HLA-A*3101 Proxy SNP GenotypingGenotyping of the rs1061235 SNP in samples from all Liverpool subjects (including case sub-jects with maculopapular exanthema, the hyper-sensitivity syndrome, or SJS–TEN and clinical control subjects) was performed by KBioscience. This allele has previously been shown to be in complete linkage disequilibrium (r2 = 1) with the HLA-A*3101 allele in populations of European descent.22

Statistical AnalysisStatistical analyses were performed with the use of PLINK (version 1.05),23 RevMan (version 5) (http://ims.cochrane.org/revman), and Haploview24 software packages. For the genomewide associa-tion studies, both the Liverpool collaborators and the EPIGEN consortium used logistic regression with an additive model of inheritance. They ad-justed for population stratification by including significant principal components as covariates in the logistic-regression model and used a P value of less than 5.0×10−8 to indicate genomewide sig-nificance. Pooled analysis of the different groups of subjects was performed to estimate odds ratios and confidence intervals from summary data with the use of a random-effects model. Between-study heterogeneity was tested by calculating I2 and Tau.2 The z-score directionality or estimated risk was checked for consistency with observed data.

R esult s

Association with the Hypersensitivity SyndromeWe first conducted an association analysis in-volving 1.2M SNPs in samples from the 22 Liver-pool subjects with carbamazepine-induced hy-persensitivity syndrome and from 2691 healthy controls from the U.K. National Blood Services Collection. (The full set of results can be obtained from the European Genotype Archive, access num-

ber EGAS00000000037.) We identified a strong signal in the MHC region on chromosome 6, with several SNPs around HLA-A reaching genome-wide significance (P = 3.5×10−8) (Fig. 1A and 1B). This variant was seen in 40.0% of case subjects but in only 4.9% of control subjects. The top hit (rs1061235) had previously been shown to be a proxy for the HLA-A*3101 allele in persons of European descent.22 Therefore, we focused on this locus as a predictor of response through high-resolution, sequence-based HLA-A typing, which confirmed the absolute correlation (r2 = 1) between rs1061235 and HLA-A*3101 that had been observed previously.22 No other classic HLA-A alleles reached nominal significance (Table 1 in the Supplementary Appendix).

During the course of this study, samples from 4 additional subjects with the hypersensitivity syndrome and from 1 additional subject with AGEP became available from the Liverpool and EPIGEN centers. Of these 5 subjects, 2 (including the single subject with AGEP) tested positive for HLA-A*3101. A comparison of all 27 subjects with the 257 control subjects without adverse drug reactions produced an odds ratio of 12.41 (95% confidence interval [CI], 1.27 to 121.03; P = 0.03) (Fig. 2).

Association with Maculopapular ExanthemaEPIGEN investigators performed a genomewide association study of samples from 43 subjects with maculopapular exanthema and 1296 healthy control subjects from the 1958 British Birth Co-hort. (The full set of results can be obtained from dbGAP, access number EGAS00000000036.) HLA-A*3101 was the most strongly associated allele (P = 1.11×10−6) and was observed in 27% of case subjects and 4% of control subjects, although the effect was not significant genomewide (Fig. 3 in the Supplementary Appendix). However, we had ac-cess to another 63 case subjects with maculopap-ular exanthema and 257 control subjects without adverse drug reactions. Our combined analysis of all 106 subjects with maculopapular exanthema and the 257 clinical control subjects showed a significant allelic association with HLA-A*3101 (P = 8.0×10−7), generating an odds ratio of 8.33 (95% CI, 3.59 to 19.36) (Fig. 2). Although the analyses of samples from subjects with maculo-papular exanthema and those with the hypersen-sitivity syndrome were carried out in the Nether-





HLA-A

HLA-A*3101

P Va

lue

(−lo

g 10)

7

6

01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17181920

2122

Chromosome

B SNPs in the HLA-A Region

A Carbamazepine-Induced Hypersensitivity

P Va

lue

(−lo

g 10)

Reco

mbi

natio

n Ra

te (c

M/M

b)6

7

8

10

9

5

2

1

0

4

3

60

80

0

40

20

29.7 30.0 30.3

Chromosome 6 (positions in Mb)

TypedLD

0.80 to 1.00

0.50 to <0.80

0.20 to <0.50

<0.20

0.80 to 1.00

0.50 to <0.80

0.20 to <0.50

<0.20

Imputed

Figure 1. Results of a Genomewide Association Study of Samples from Case Subjects with Carbamazepine-Induced Hypersensitivity Syndrome and Control Subjects.

Panel A shows a Manhattan plot for logistic regression with a strong signal in the HLA-A region on chromosome 6. Each dot represents a P value for the comparison of 22 case subjects with the hypersensitivity syndrome and 2691 healthy control subjects from the Wellcome Trust Case Control Consortium. Panel B shows P values for single-nu-cleotide polymorphisms (SNPs) in the HLA-A region. There is a strong association between carbamazepine-induced hypersensitivity syndrome and the HLA-A*3101 allele, with identification of the allele in 40.0% of case subjects and only 4.9% of control subjects. The x axis shows the SNP position on chromosome 6 (National Center for Biotechnol-ogy Information build 36). The left y axis shows the negative log10 of P values for the comparison between case sub-jects and control subjects, as calculated with logistic regression. The right y axis shows the recombination rate on chromosome 6 between 29 and 31 Mb. The diamonds show the degree of linkage disequilibrium (LD) in the samples.





lands and Liverpool, respectively, the effect size of HLA-A*3101 on risk was consistent across the EPIGEN and U.K. groups.

Association with SJS–TENOn the basis of our hypothesis that the HLA-A*3101 allele is associated with carbamazepine-induced SJS–TEN, we genotyped HLA-A*3101 in a group of 12 subjects with SJS–TEN who were recruited from both the Liverpool and EPIGEN-affiliated centers. Of these 12 subjects, 5 (42%) carried the allele, as compared with 10 (4%) of the 257 clinical control subjects (odds ratio, 25.93; 95% CI, 4.93 to 116.18; P = 8.0×10−5) (Fig. 2).

A pooled analysis of all 145 subjects with carbamazepine-induced hypersensitivity and 257 control subjects without adverse drug reactions showed a strong association between hypersen-sitivity and the HLA-A*3101 allele (odds ratio, 9.12; 95% CI, 4.03 to 20.65; P = 1.0×10−7). On the basis of this pooled collection, we calculated that the presence of HLA-A*3101 had a sensitiv-ity of 26% and a specificity of 96% as a predic-tor of carbamazepine-associated hypersensitivity. Given that carbamazepine hypersensitivity has a prevalence of 5% (1:20), application of the test criteria from our pooled analysis would increase the post-test probability to 26% (approximately 1:4) on the basis of the positive likelihood ratio of 6.74 (95% CI, 3.51 to 13.00). Since the nega-tive likelihood ratio was 0.77 (95% CI, 0.67 to 0.84), a negative HLA-A*3101 test would reduce the probability of hypersensitivity from 5.0% to 3.8% (1:26 ratio).

Discussion

We found that a variation in the presence of the HLA-A*3101 allele is an important predictor of the full spectrum of carbamazepine-induced hy-persensitivity reactions in persons of European descent. Such reactions range from the relatively mild but nevertheless troublesome maculopapu-lar exanthema to more severe reactions, such as the hypersensitivity syndrome and SJS–TEN. Al-though the presence of the HLA-A*3101 allele is neither necessary nor sufficient for the develop-ment of hypersensitivity to carbamazepine, it is associated with a significantly increased risk.

Our data add to growing evidence of the role of different HLA alleles in predisposing patients

to immune-mediated adverse reactions, such as drug-induced hypersensitivity 4,25-27 and liver in-jury.3,28 The most prominent example thus far is the strong predictive value of the HLA-B*5701 allele for hypersensitivity to the drug abacavir, used for the treatment of human immunodefi-ciency virus infection.25,26,29 FDA guidelines now recommend HLA-B*5701 testing in advance of abacavir prescription. Testing results in a reduc-tion in the incidence of abacavir hypersensitivi-ty27,30 and is cost-effective.29,31

Our study now provides strong evidence sug-gesting that HLA-A*3101 is a predictor of carba-mazepine hypersensitivity in Europeans. First, despite the small number of subjects who were included in the analysis, this signal had genome-wide significance. Second, the signal was ob-served across independent groups of case subjects with different phenotypes of carbamazepine hy-persensitivity, with the use of independent sets of control subjects. Third, HLA-A*3101 has pre-viously been shown to be associated with multiple carbamazepine-hypersensitivity phenotypes, in-cluding SJS–TEN in Japanese subjects15,32 and with carbamazepine-induced maculopapular ex-anthema in subjects of Han Chinese ancestry.14 The one subject in our study who had AGEP, a reaction characterized by sterile pustules, neutro-philic inflammation, and high levels of interleu-kin-8 in skin,33 also was found to carry the HLA-A*3101 allele. The prevalence of this allele is 2 to 5% in Northern European populations, 2% in Han Chinese populations, and 9% in Japanese populations.32,34,35 The finding that HLA-A*3101 is associated with carbamazepine-induced hyper-sensitivity in populations of different ancestries parallels the finding of HLA-B*5701-associated abacavir hypersensitivity.30,36

HLA-B*1502, a predictor of carbamazepine hypersensitivity reactions in Asian populations, seems to be phenotype-specific in that it pre-dicts the development of SJS–TEN but not the development of the hypersensitivity syndrome or maculopapular exanthema.14 This finding has led to the suggestion that the pathogenesis of carba-mazepine-induced SJS–TEN is distinct from that of the hypersensitivity syndrome. This hypothe-sis is based on the fact that class I alleles are usually associated with specific CD8+ cytotoxic T-cell reactions, which is consistent with the known pathologic features of SJS–TEN.37 How-





ever, our data suggest that the class I allele HLA-A*3101 may be important for an increased range of phenotypes, including maculopapular exan-thema and the hypersensitivity syndrome, at least in European populations. Persons with carba-mazepine-induced hypersensitivity syndrome have been shown to have several types of drug-specif-ic T cells in peripheral blood, including CD4+, CD8+, and CD4−CD8+ cells.38 Although there is

no obvious relationship between HLA-A*3101 and HLA-B*1502, reports have indicated a possi-ble overlap in peptide-binding specificity.39 Giv-en the degree of linkage disequilibrium across the extended MHC region, we cannot be certain that either HLA-B*1502 or HLA-A*3101 is causal.

We used samples from healthy volunteers in the WTCCC as control subjects for both ge-nomewide association studies. Given that the

0.1 1.0 10.0 500.0

Hypersensitivity syndromeEPIGENUKSubtotal Heterogeneity: tau2=1.03, chi2=1.58; 1 df, P=0.21; I2=37%.Test for overall effect: z=2.17; P=0.03

Odds Ratio (95% CI)Carbamazepine

Reaction Controls Study Weighting (%)Subgroup

3.08 (0.15–63.89)

12.41 (1.27–121.03)30.71 (3.61–261.65)

Maculopapular exanthemaEPIGENUKSubtotal Heterogeneity: tau2=0.00, chi2=0.01; 1 df, P=0.92; I2=0%.Test for overall effect: z=4.93; P=8.0×10−7

8.19 (3.26–20.56)

8.33 (3.59–19.36)9.15 (1.12–74.47)

Stevens–Johnson syndromeEPIGENUKSubtotal Heterogeneity: tau2=0.00, chi2=0.01; 1 df, P=0.93; I2=0%.Test for overall effect: z=3.94; P=8.0×10−5

22.67 (2.86–179.73)

25.93 (4.93–116.18)25.80 (2.24–297.58)

All phenotypesEPIGENUKSubtotal Heterogeneity: tau2=0.00, chi2=0.35; 1 df, P=0.55; I2=0%.Test for overall effect: z=5.30; P=1.0×10−7

8.29 (3.40–20.23)

9.12 (4.03–20.65)

0.002

14.98 (1.95–115.09)

No. Positivefor the

HLA-A*3101Allele

01010

131023

235

152338

TotalNo.

32427

4957

106

48

12

5689

145

39.460.6

100

83.816.2

100

58.241.8

100

83.916.1

100

91

10

No. Positivefor the

HLA-A*3101Allele

91

10

91

10

91

10

21344

257

TotalNo.

21344

257

21344

257

21344

257

HLA-A*3101 Less Likely

HLA-A*3101 More Likely

Figure 2. Distribution of the HLA-A*3101 Allele across the Spectrum of Clinical Phenotypes of Case Subjects with Carbamazepine- Induced Hypersensitivity and Control Subjects without Adverse Reactions to Carbamazepine.

Subjects were recruited at centers collaborating with the University of Liverpool and Walton Centre for Neurology (UK) or centers affili-ated with the EPIGEN consortium. The sample obtained from one patient with acute generalized exanthematous pustulosis (AGEP) was analyzed with the samples for the group of case subjects with the hypersensitivity syndrome. Since there were no observations of the HLA-A*3101 allele in the three case subjects with the hypersensitivity syndrome in the EPIGEN cohort, 0.5 was added to each value in a two-by-two contingency table to estimate an odds ratio. One patient with the Stevens–Johnson syndrome and toxic epidermal necrolysis (SJS–TEN) was of mixed European and Thai ancestry. Study weighting (indicated by different sizes of squares) refers to the proportion of subjects who were recruited from each study cohort. Diamonds indicate pooled odds ratios. The horizontal lines indicate 95% confi-dence intervals. The abbreviation df denotes degrees of freedom, and I2 is the percentage of total variation that is due to heterogeneity rather than chance.





hypersensitivity syndrome occurs in 1 in 5000 to 1 in 10,000 persons, it is unlikely that more than one or two population control subjects would be at risk for hypersensitivity with exposure to car-bamazepine. The larger WTCCC collections there-fore afforded increased power over that provided by comparison of case subjects with the clinical control subjects. Nevertheless, in analyses involv-ing the 257 clinical control subjects, the associa-tion with HLA-A*3101 remained strong for both the hypersensitivity syndrome (odds ratio, 12.41; 95% CI, 1.27 to 121.03) and SJS (odds ratio, 25.93; 95% CI, 4.93 to 116.18), although the ef-fect size was smaller with maculopapular exan-thema (odds ratio, 8.33; 95% CI, 3.59 to 19.36).

Additional studies in populations of Europe-an descent and other ancestry are required to further characterize this association. The Inter-national Serious Adverse Event Consortium is coordinating a global effort to recruit subjects with drug-induced severe adverse events (www .saeconsortium.org). Efforts such as this will fa-cilitate the analysis of larger groups of subjects, which may well uncover other genetic variants associated with a smaller effect size or popula-tion-specific variants. Additional work is also required to determine whether the effect of the HLA-A*3101 allele is specific to carbamazepine or whether it also applies to other drugs. A pro-spective study that is based on the results we present here would facilitate the routine clinical use of HLA-A*3101 testing in carbamazepine treatment. In this issue of the Journal, a report describes how prospective HLA-B*1502 typing can prevent carbamazepine-induced SJS–TEN in Han Chinese populations.5

We propose that HLA-A*3101 is clinically rele-vant as a marker to predict hypersensitivity reac-tions. On the assumption that the prevalence of carbamazepine-induced hypersensitivity is 5.0%, the presence of the HLA-A*3101 allele increases the risk of hypersensitivity to 26.0%, whereas its absence reduces the risk to 3.8%. Among Japa-nese patients, the estimated reduction is from 2.9% to 1.1%. The sensitivity and specificity values estimated here for Europeans (26% and 96%) and by others for Japanese populations (61% and 88%)32 vary. Among a number of pos-sible reasons for these differences are differences in study size, in the frequency of HLA-A*3101 (and potential interacting variation), and in the

effect size of the allele across the two populations. Nonetheless, these numbers are similar to those underlying the test criteria calculated for the presence of the HLA-B*5701 allele and abacavir hypersensitivity (sensitivity, 45.5%; specificity, 97.6%; and a risk reduction from 7.8% to 2.7%).

On the basis of our data, we have calculated that 83 patients would need to be screened to prevent one case of carbamazepine hypersensi-tivity; the number in the Japanese population is 56. However, it is important to note that this calculation is based on a conservative estimate of the prevalence of carbamazepine hypersensitiv-ity (5%). The prevalence of carbamazepine-in-duced hypersensitivity that was determined in the context of the Standard and New Antiepileptic Drugs (SANAD) trial (Current Controlled Trials number, ISRCTN38354748) (also involving subjects of European descent) was 10%.40 On the basis of this prevalence, the number of persons who would need to be screened to prevent one in-stance of carbamazepine-induced hypersensitiv-ity is 39. We therefore suggest that consideration be given to adding the association with HLA-A*3101 to the drug label for carbamazepine.

Supported by grants to the Liverpool collaborators from the Department of Health, the National Health Service Chair of Pharmacogenetics, the Medical Research Council Centre for Drug Safety Science, the Wolfson Foundation, the Wellcome Trust Sanger Institute, and the National Institute for Health Research (to Dr. Pirmohamed); by grants to the EPIGEN consor-tium from the Medical Research Council (G0400126), the Well-come Trust (084730), University College London Hospitals Char-ity, Clinical Research and Development Committee (F136), and the National Institute for Health Research (08-08-SCC); by an award (2009/001) from Brainwave–the Irish Epilepsy Associa-tion; by the Medical Research Charities Group of Ireland, the Health Research Board, the National Society for Epilepsy, Fonds National de la Recherche Scientifique, and Fonds Erasme pour la Recherche Médicale, Université Libre de Bruxelles (Belgium); by a Translational Research Scholars award from the Health Re-search Board of Ireland (to Mr. McCormack); by the Department of Health National Institute for Health Research Biomedical Research Centres; and by a contract from the National Cancer Institute (HHS-N261200800001E) and the Intramural Research Program at the National Cancer Institute, National Institutes of Health.


We thank the subjects and the physicians who recruited them; U.K. Eudragene co-coordinator Paul McKeigue; Drs. Ines Salado, Alfonso Carvajal, Luisa Ibanez, Jean-Louis Montastruc, Maryse Lapeyre-Mestre, Emmanuelle Bondon-Guitton, Aatif M. Husain, William B. Gallantine, and Mohamad Mikati for their help in recruiting subjects; the pharmacovigilance centers involved in case ascertainment in France and Spain, the Association Fran-çaise des Centres Régionaux de Pharmacovigilance and Sistema Español de Farmacovigilancia; and Agencia Española de Medica-mentos y Productos Sanitarios.





AppendixThe authors’ affiliations are as follows: Molecular and Cellular Therapeutics, the Royal College of Surgeons in Ireland (M. McCormack, S.A., N.D., G.L.C.), and the Division of Neurology, Beaumont Hospital (G.D.O., E.C., N.D.) — both in Dublin; the Department of Molecular and Clinical Pharmacology, University of Liverpool (A.A., G.J.S., T.M., B.K.P., M. Pirmohamed), and Walton Centre for Neurology (T.M.) — both in Liverpool; Wellcome Trust Sanger Institute, Hinxton (S.B., P.D.); and the National Society for Epilepsy, Chalfont-St-Peter, Buckinghamshire (K.C., S.M.S.) — all in the United Kingdom; the Department of Medicine, Boston University (J.F.F.), Ragon Institute of Massachusetts General Hospital, Massachusetts Institute of Technology, and Harvard University (M.C.), Harvard–MIT Division of Health Sciences and Technology (X.J.), and the Division of Genetics, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School (P.I.W.B.) — all in Boston; the Department of Clinical and Experimental Epilepsy, Univer-sity College London, Institute of Neurology, Queen Square (D.K., K.C., S.M.S.), the Department of Primary Care and Public Health Sciences, Division of Health and Social Care Research, King’s College (M. Molokhia), and the Centre for Neuroscience, Department of Medicine, Imperial College London (M.R.J.) — all in London; the Cancer and Inflammation Program, Laboratory of Experimental Im-munology, SAIC–Frederick, National Cancer Institute at Frederick, Frederick, MD (M.C.); the Program in Medical and Population Ge-netics, Broad Institute of Harvard and MIT, Cambridge, MA (P.I.W.B.); Julius Center for Health Sciences and Primary Care and the Department of Medical Genetics, Division of Biomedical Genetics, University Medical Center, Utrecht, the Netherlands (P.I.W.B.); the Center for Human Genome Variation (K.V.S., E.L.H., N.W., D.B.G.) and the Department of Medicine (Neurology) (R.A.R.), Duke Uni-versity Medical School, Durham, NC; the Department of Neurology, Hôpital Erasme, Université Libre de Bruxelles, Brussels (M. Pan-dolfo, C.D.); and the Department of Rheumatology, Clinical Immunology, and Allergology, University of Bern, Bern, Switzerland (W.P.).

References

1. Roujeau JC. Clinical heterogeneity of drug hypersensitivity. Toxicology 2005; 209:123-9.2. Syn WK, Naisbitt DJ, Holt AP, Pirmo-hamed M, Mutimer DJ. Carbamazepine-induced acute liver failure as part of the DRESS syndrome. Int J Clin Pract 2005;59: 988-91. [Erratum, Int J Clin Pract 2005;59: 1371.]3. Daly AK, Donaldson PT, Bhatnagar P, et al. HLA-B*5701 genotype is a major de-terminant of drug-induced liver injury due to flucloxacillin. Nat Genet 2009;41: 816-9.4. Chung WH, Hung SI, Hong HS, et al. Medical genetics: a marker for Stevens-Johnson syndrome. Nature 2004;428:486.5. Chen P, Lin J-J, Lu C-S, et al. Carba-mazepine-induced toxic effects and HLA-B*1502 screening in Taiwan. N Engl J Med 2011;364:1126-33.6. Ding WY, Lee CK, Choon SE. Cutane-ous adverse drug reactions seen in a ter-tiary hospital in Johor, Malaysia. Int J Dermatol 2010;49:834-41.7. Locharernkul C, Loplumlert J, Limo-tai C, et al. Carbamazepine and phenytoin induced Stevens-Johnson syndrome is as-sociated with HLA-B*1502 allele in Thai population. Epilepsia 2008;49:2087-91. [Erratum, Epilepsia 2009;50:971.]8. Mehta TY, Prajapati LM, Mittal B, et al. Association of HLA-B*1502 allele and carbamazepine-induced Stevens-Johnson syndrome among Indians. Indian J Der-matol Venereol Leprol 2009;75:579-82.9. Alfirevic A, Jorgensen AL, Williamson PR, Chadwick DW, Park BK, Pirmohamed M. HLA-B locus in Caucasian patients with carbamazepine hypersensitivity. Pharmacogenomics 2006;7:813-8.10. Lonjou C, Borot N, Sekula P, et al. A European study of HLA-B in Stevens-Johnson syndrome and toxic epidermal necrolysis related to five high-risk drugs. Pharmacogenet Genomics 2008;18:99-107.11. Lonjou C, Thomas L, Borot N, et al. A marker for Stevens-Johnson syndrome:

ethnicity matters. Pharmacogenomics J 2006;6:265-8.12. Ferrell PB Jr, McLeod HL. Carbamaze-pine, HLA-B*1502 and risk of Stevens-Johnson syndrome and toxic epidermal necrolysis: US FDA recommendations. Pharmacogenomics 2008;9:1543-6.13. Alfirevic A, Mills T, Harrington P, et al. Serious carbamazepine-induced hy-persensitivity reactions associated with the HSP70 gene cluster. Pharmacogenet Genomics 2006;16:287-96.14. Hung SI, Chung WH, Jee SH, et al. Ge-netic susceptibility to carbamazepine-induced cutaneous adverse drug reactions. Pharmacogenet Genomics 2006;16:297-306.15. Kashiwagi M, Aihara M, Takahashi Y, et al. Human leukocyte antigen genotypes in carbamazepine-induced severe cutane-ous adverse drug response in Japanese patients. J Dermatol 2008;35:683-5.16. Pirmohamed M, Lin K, Chadwick D, Park BK. TNFalpha promoter region gene polymorphisms in carbamazepine-hyper-sensitive patients. Neurology 2001;56: 890-6.17. Kardaun SH, Sidoroff A, Valeyrie-Al-lanore L, et al. Variability in the clinical pattern of cutaneous side-effects of drugs with systemic symptoms: does a DRESS syndrome really exist? Br J Dermatol 2007;156:609-11.18. Speeckaert MM, Speeckaert R, Lam-bert J, Brochez L. Acute generalized exan-thematous pustulosis: an overview of the clinical, immunological and diagnostic concepts. Eur J Dermatol 2010;20:425-33.19. Auquier-Dunant A, Mockenhaupt M, Naldi L, Correia O, Schröder W, Roujeau JC. Correlations between clinical patterns and causes of erythema multiforme ma-jus, Stevens-Johnson syndrome, and toxic epidermal necrolysis: results of an inter-national prospective study. Arch Derma-tol 2002;138:1019-24.20. The International HIV Controllers Study (2010). The major genetic deter-

minants of HIV-1 control affect HLA class I peptide presentation. Science 2010; 330(6010):1551-7.21. Li Y, Willer CJ, Ding J, Scheet P, Abe-casis GR. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol 2010:34:816-34.22. de Bakker PI, McVean G, Sabeti PC, et al. A high-resolution HLA and SNP haplo-type map for disease association studies in the extended human MHC. Nat Genet 2006;38:1166-72.23. Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome as-sociation and population-based linkage analyses. Am J Hum Genet 2007;81:559-75.24. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005;21:263-5.25. Hetherington S, Hughes AR, Mostell-er M, et al. Genetic variations in HLA-B region and hypersensitivity reactions to abacavir. Lancet 2002;359:1121-2.26. Mallal S, Nolan D, Witt C, et al. Asso-ciation between presence of HLA-B*5701, HLA-DR7, and HLA-DQ3 and hypersensi-tivity to HIV-1 reverse-transcriptase in-hibitor abacavir. Lancet 2002;359:727-32.27. Mallal S, Phillips E, Carosi G, et al. HLA-B*5701 screening for hypersensitiv-ity to abacavir. N Engl J Med 2008;358:568-79.28. Donaldson PT, Daly AK, Henderson J, et al. Human leucocyte antigen class II genotype in susceptibility and resistance to co-amoxiclav-induced liver injury. J Hepatol 2010;53:1049-53.29. Hughes DA, Vilar FJ, Ward CC, Al-firevic A, Park BK, Pirmohamed M. Cost-effectiveness analysis of HLA B*5701 genotyping in preventing abacavir hyper-sensitivity. Pharmacogenetics 2004;14: 335-42.30. Saag M, Balu R, Phillips E, et al. High sensitivity of human leukocyte antigen-B*5701 as a marker for immunologically





confirmed abacavir hypersensitivity in white and black patients. Clin Infect Dis 2008;46:1111-8.31. Schackman BR, Scott CA, Walensky RP, Losina E, Freedberg KA, Sax PE. The cost-effectiveness of HLA-B*5701 genetic screening to guide initial antiretroviral therapy for HIV. AIDS 2008;22:2025-33.32. Ozeki T, Mushiroda T, Yowang A, et al. Genome-wide association study identi-fies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutane-ous adverse drug reactions in Japanese population. Hum Mol Genet 2011;20: 1034-41.33. Britschgi M, Steiner UC, Schmid S, et al. T-cell involvement in drug-induced acute generalized exanthematous pustu-losis. J Clin Invest 2001;107:1433-41.34. Schmidt AH, Baier D, Solloch UV, et al. Estimation of high-resolution HLA-A,

-B, -C, -DRB1 allele and haplotype fre-quencies based on 8862 German stem cell donors and implications for strategic do-nor registry planning. Hum Immunol 2009;70:895-902.35. Wen SH, Lai MJ, Yang KL. Human leu-kocyte antigen-A, -B, and -DRB1 haplo-types of cord blood units in the Tzu Chi Taiwan Cord Blood Bank. Hum Immunol 2008;69:430-6.36. Zucman D, Truchis P, Majerholc C, Stegman S, Caillat-Zucman S. Prospective screening for human leukocyte antigen-B*5701 avoids abacavir hypersensitivity reaction in the ethnically mixed French HIV population. J Acquir Immune Defic Syndr 2007;45:1-3.37. Lerch M, Pichler WJ. The immuno-logical and clinical spectrum of delayed drug-induced exanthems. Curr Opin Al-lergy Clin Immunol 2004;4:411-9.

38. Wu Y, Farrell J, Pirmohamed M, Park BK, Naisbitt DJ. Generation and charac-terization of antigen-specific CD4+, CD8+, and CD4+CD8+ T-cell clones from patients with carbamazepine hypersensi-tivity. J Allergy Clin Immunol 2007;119: 973-81.39. Chelvanayagam G. A roadmap for HLA-A, HLA-B, and HLA-C peptide bind-ing specificities. Immunogenetics 1996; 45:15-26.40. Marson AG, Al-Kharusi AM, Alwaidh M, et al. The SANAD study of effective-ness of carbamazepine, gabapentin, la-motrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an un-blinded randomised controlled trial. Lan-cet 2007;369:1000-15.Copyright © 2011 Massachusetts Medical Society.

SPECIALTIES AND TOPICS AT NEJM.ORG

Specialty pages at the Journal’s Web site (NEJM.org) feature articles in cardiology, endocrinology, genetics, infectious disease, nephrology, pediatrics, and many other medical specialties. These pages, along with collections of articles on clinical and nonclinical topics, offer links to interactive and multimedia content and feature

recently published articles as well as material from the NEJM archive (1812–1989).



Review article


Genomic MedicineW. Gregory Feero, M.D., Ph.D., and Alan E. Guttmacher, M.D., Editors

Genomics and Drug ResponseLiewei Wang, M.D., Ph.D., Howard L. McLeod, Pharm.D.,

and Richard M. Weinshilboum, M.D.

From the Division of Clinical Pharmacol-ogy, Department of Molecular Pharma-cology and Experimental Therapeutics, Mayo Medical School, Mayo Clinic, Roch-ester, MN (L.W., R.M.W.); and the UNC Institute for Pharmacogenomics and In-dividualized Therapy, University of North Carolina, Chapel Hill (H.L.M). Address reprint requests to Dr. Weinshilboum at the Division of Clinical Pharmacology, Department of Molecular Pharmacology and Experimental Therapeutics, Mayo Clin-ic, 200 First St. SW, Rochester, MN 55905, or at [email protected].


Pharmacogenomics is the study of the role of inherited and ac-quired genetic variation in drug response.1 Clinically relevant pharmacoge-netic examples, mainly involving drug metabolism, have been known for

decades, but recently, the field of pharmacogenetics has evolved into “pharmacoge-nomics,” involving a shift from a focus on individual candidate genes to genome-wide association studies. Such studies are based on a rapid scan of markers across the genome of persons affected by a particular disorder or drug-response pheno-type and persons who are not affected, with tests for association that compare ge-netic variation in a case–control setting.2 An example is provided in this issue of the Journal: McCormack and colleagues, testing for genomewide association, identified an HLA allele that is associated with hypersensitivity reactions to the anticonvulsant and mood-stabilizing drug carbamazepine in persons of European descent.3 Phar-macogenomics facilitates the identification of biomarkers that can help physicians optimize drug selection, dose, and treatment duration and avert adverse drug reac-tions. In addition, pharmacogenomics can provide new insights into mechanisms of drug action and as a result can contribute to the development of new therapeutic agents.

In 2003, two reviews of pharmacogenetics were published the Journal.4,5 Since then, both genomic science and its application to drug response have undergone major advances.6 Here we review some of those advances, with an emphasis on discovery through genomewide association studies. We describe examples that highlight principles of pharmacogenomics that are relevant to a wide variety of drugs. The Food and Drug Administration (FDA) has altered drug labels and issued warn-ings about pharmacogenomic variation affecting drug response, raising the issue of the level of evidence required to show clinical utility7 and the respective roles of regulatory agencies such as the FDA and of academic and professional societies in the evaluation of pharmacogenetic analyses for the clinic.

C a r diova scul a r Drugs

Many drugs have proven efficacy in the treatment and prevention of cardiovascular disease. Not uncommonly, these drugs have narrow therapeutic indexes that are influenced by genetic variation — a hallmark of drugs for which pharmacoge-nomic approaches are likely to provide substantial clinical benefit. The anticoagu-lant agents warfarin and clopidogrel are high on the list of widely prescribed car-diovascular drugs with narrow therapeutic indexes. The pharmacogenomic features of these drugs illustrate the rapid evolution of our understanding of the role of in-heritance in the variation in drug efficacy and the risk of adverse drug reactions. In the case of both agents, the application of classic candidate-gene pharmacogenetics



genomic medicine


has identified important genomic markers of variation in efficacy and adverse reactions, ob-servations that were subsequently confirmed in genomewide association studies. The FDA acted quickly on these data by relabeling warfarin and adding a warning box on the labeling for clopi-dogrel. Data supporting the clinical utility of routine use of pharmacogenetic testing for both these drugs are evolving.8-10

Warfarin is the most widely prescribed oral anticoagulant in North America and much of Eu-rope.11 Despite the availability of the international normalized ratio (INR), a laboratory test that is universally used to measure the anticoagulant ef-fect of warfarin, serious adverse responses, includ-ing hemorrhage and undesired coagulation, con-tinue to complicate therapy, making warfarin one of the drugs most often responsible for emergency room visits.12,13 Chemically, warfarin is a racemic mixture (i.e., one that is composed of two enantio-morphic isomers). S-warfarin is three to five times as potent as R-warfarin as an anticoagulant, has a shorter half-life, and is metabolized predomi-nantly by a cytochrome P-450 enzyme, CYP2C9.11 Two common CYP2C9 allozymes (see Glossary) have only a fraction of the level of enzyme activ-ity of the wild-type allozyme CYP2C9*1: 12% for CYP2C9*2 and 5% for CYP2C9*3.11,14 More than a decade ago, it was reported that patients who required a low final dose of warfarin on the basis of INR values often carried one or two of these two common CYP2C9 variant alleles and were at increased risk for hemorrhage during warfarin therapy, presumably because they metabolized the drug more slowly.14 Those observations were con-firmed, but it quickly became clear that the pres-ence of CYP2C9 polymorphisms did not explain most of the variation in the final warfarin dose.

Pharmacogenetic studies of warfarin changed dramatically in 2004 when the target for warfarin-based anticoagulants, vitamin K epoxide reductase complex subunit 1 (VKORC1), was identified,15,16 and single-nucleotide polymorphisms (SNPs) in VKORC1 were shown to be associated with the dose of warfarin required to achieve a target INR val-ue.17 In 2009, a genomewide association study looked for associations between several hundred thousand SNPs and warfarin dose in about 1000 Swedish patients who were taking warfarin. The results showed two major signals in and around CYP2C9 and VKORC1 (Fig. 1A).18 When the authors removed the effects of those signals through

multiple regression adjustment, they observed an additional signal, implicating another cytochrome P450 gene (CYP4F2) (Fig. 1B). CYP4F2 was subse-quently shown to catalyze vitamin K oxidation.19 The variant CYP4F2 allozyme shows decreased ability to catalyze the reaction, and as a result persons who carry the relevant genetic variant in CYP4F2 might require an increase in the warfarin dose (Fig. 1C). CYP2C9, VKORC1, and CYP4F2 have also been implicated in a genomewide association study of the administration of acenocoumarol, an anticoagulant related to warfarin.20

Taken together, CYP2C9 and VKORC1 genotypes explain about 30 to 40% of the total variation in the final warfarin dose.21 These observations raise the possibility that testing patients for variations in CYP2C9 and VKORC1 might provide information that could enhance clinical algorithms currently used to guide the administration of warfarin. To examine the potential clinical utility of testing for CYP2C9 and VKORC1 genotypes, in addition to INR monitoring and routine use of clinical algo-rithms, the International Warfarin Pharmacoge-netics Consortium recently investigated the anti-coagulant response to warfarin, as well as CYP2C9 and VKORC1 genotype data, for about 4000 per-sons of various ancestral origins. The investigators compared therapeutic outcomes with the applica-tion of standard clinical algorithms that included age, sex, and INR values and outcomes with the use of an algorithm that included CYP2C9 and VKORC1 genotype information and concluded that the addition of genotype information enhanced outcomes, especially for patients who required unusually high or low warfarin doses.22 CYP4F2 was not included in this algorithm but has been included in several algorithms developed more recently.23,24 Consistent with this conclusion are the results of a study comparing nearly 900 pa-

Glossary

Allozyme: Alternate versions of an enzyme determined by genetic variants (alleles) present at a genetic locus.

Gene cluster: Two or more genes in close physical proximity in the genome that encode similar gene products.

Genomewide association study: An approach used in genetics research to look for associations between large numbers (typically hundreds of thou-sands) of specific genetic variations (most commonly single-nucleotide polymorphisms) and particular diseases.

Single-nucleotide polymorphism: A single-nucleotide variation in a genetic sequence; a common form of variation in the human genome.





tients for whom genetic information on CYP2C9 and VKORC1 was made available to prescribing physicians with a matched historical control group of patients who were started on warfarin therapy without genetic information.25 Six months after the initiation of warfarin therapy, hospitalizations for hemorrhage were 28% less common in the group of patients for whom genetic information on CYP2C9 and VKORC1 had been supplied to prescribing physicians than in the control group (Fig. 2).

The FDA revised the label on warfarin in Feb-ruary 2010, providing genotype-specific ranges of doses and suggesting that genotypes be taken into consideration when the drug is prescribed. The wide availability of CYP2C9 and VKORC1 genotyp-ing and the release of both Web-based and per-sonal decision-support tools have facilitated the clinical use of this information. Nevertheless, the clinical adoption of genotype-guided administra-tion of warfarin has been slow, even though the evidence supporting such adoption is similar to

32 composite

3

2−L

og10

P V

alue

78

17

31

14

7

01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

2021

22X

YM

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1920

2122

XY

M

Chromosome

B Association with CYP4F2

C Warfarin and the Vitamin K Cycle

A Association with CYP2C9 and VKORC1CYP2C9 VKORC1

−Log

10 P

Val

ue 14

7

0

Chromosome

CYP4F2

Vitamin Kdihydroquinone

Vitamin K epoxide

Clotting protein

Carbon dioxide

γ-Carboxy-glutamylclottingfactors

Vitamin K–dependent

carboxylationVitamin Kquinone

Hydroxy-vitamin K

VKORC1and quinone reductase

VKORC1

Warfarin7-OH-warfarin

7-OH-warfarin Warfarin

CYP2C9

CYP4F2

CYP2C9

* **

*

Figure 1. Warfarin Pharmacogenomics.

Panels A and B show Manhattan plots of P values (negative log10) for the association between single-nucleotide-polymorphisms (SNPs) across the genome and the final warfarin dose. The horizontal line indicates a P value of 1.5×10−7, which is the level of genomewide statistical significance. In Panel A, the results of univariate regression analysis highlight SNP signals in or near CYP2C9 and VKORC1. In Panel B, the results of multivariate regression analysis with adjustment for the contributions of CYP2C9 and VKORC1 show the CYP4F2 signal on chromosome 19. (Data are from Takeuchi et al.18) The label *2 indicates the nonsynonymous SNP rs1799853, *3 indicates the non-synonymous SNP rs1057910, and the *2*3 composite indicates the SNP rs4917639. M denotes mitochondrial SNPs. Panel C shows the sites of action of warfarin in the vitamin K cycle, as well as the roles of CYP2C9, CYP4F2, and VKORC1 in this process.



genomic medicine


the evidence supporting currently used clinical variables, such as age, drug interactions, and an-cestral origin. Some observers have expressed a need for prospective assessment of the value of this genetic information in warfarin therapy, and several prospective clinical trials are ongoing.26 Alternative anticoagulant therapies are also be-ing developed that might replace warfarin, per-haps in patients with genotypes associated with extreme variation in warfarin response.27

Clopidogrel inhibits adenosine diphosphate (ADP)–stimulated platelet activation by binding irreversibly to a specific platelet receptor of ADP, P2Y12, thus inhibiting platelet aggregation.28,29 Dual antiplatelet therapy — clopidogrel and as-pirin — has been shown to decrease the risk of subsequent ischemic vascular events.30-32 How-ever, clopidogrel is a prodrug that requires meta-bolic activation in a reaction catalyzed by another cytochrome P-450 enzyme, CYP2C19. Like CYP2C9, CYP2C19 is genetically polymorphic with a com-mon SNP that results in a truncated protein prod-uct with little enzymatic activity.33 Several studies have shown that genetic variation in CYP2C19 re-sulting in a paucity of activity is associated with decreased clopidogrel metabolic activation, a de-creased antiplatelet effect, and an increased like-lihood of a cardiovascular event.34,35 These ob-servations have been confirmed in a genomewide association study.34

Early in 2010, the FDA added a boxed warning to prescribing information for clopidogrel, stat-ing that persons with a CYP2C19 variant encod-ing a form of the enzyme associated with a low rate of metabolism might require dose adjust-ment or the use of a different drug.36 After this FDA action, the American Heart Association and the American College of Cardiology issued a joint endorsement of CYP2C19 genotyping for patients at moderate or high risk for cardiovascular events who are treated with clopidogrel.37 This genetic test is widely available in the United States. How-ever, enthusiasm for its use has been muted, ow-ing to a lack of clarity with regard to the optimal treatment of patients who carry a CYP2C19 vari-ant, as shown by data from two large, random-ized trials in which CYP2C19 genotyping did not have a significant effect on the incidence of car-diovascular events among patients with acute coronary syndromes or atrial fibrillation.38,39 On the other hand, in a recent meta-analysis of data from nine pharmacogenetic studies of clopido-

grel involving 9685 patients who had an acute coronary syndrome or were undergoing percuta-neous coronary intervention, there was a signifi-cant association between homozygosity or hetero-zygosity for CYP2C19 reduced-function alleles and an increased risk of death from cardiovascular causes, myocardial infarction, or stroke.9 At pres-ent, it is unclear whether genotyping to predict the response to clopidogrel is clinically useful. Several studies are under way to assess the effect of dose adjustment for clopidogrel in patients who carry CYP2C19 variant alleles.8

Agen t s Used for Infec tious Dise a ses

Genomewide association studies have confirmed the identity of genetic variants in previously im-

Risk

of H

ospi

taliz

atio

n (%

)

30

25

20

15

10

5

00 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Days after Onset of Treatment

B Hospitalization for Bleeding or Thromboembolism

A Hospitalization for Any Cause

Hazard ratio, 0.69 (95% CI, 0.58–0.82)P<0.001

Intervention group

Intervention group

Historical control

Historical control

Risk

of H

ospi

taliz

atio

n (%

)

10

6

8

4

2

00 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Days after Onset of Treatment

Hazard ratio, 0.72 (95% CI, 0.53–0.97)P=0.03

AUTHOR:

FIGURE:

RETAKE:

SIZE

4-C H/TLine Combo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

1st2nd3rd

Weinshilbuom (Wang)

2 of 3

ARTIST:

TYPE:

ts

03-24-11JOB: 36412 ISSUE:

4 col22p3

Figure 2. Risk of Hospitalization among Patients Who Underwent VKORC1 and CYP2C9 Genotyping, as Compared with a Historical Control Group, 6 Months after the Initiation of Warfarin Therapy.

Shown are the rates of hospitalization for any cause (Panel A) and for bleed-ing or thromboembolism (Panel B). There was a significant benefit for pa-tients who had undergone genotyping for the presence of VKORC1 and CYP2C9 variants that have been significantly associated with the risk of over-anticoagulation. Data are from Epstein et al.25





plicated candidate genes that contribute to clini-cally important outcomes, including severe idio-syncratic adverse reactions and variation in drug efficacy. In the next set of examples, the results of pharmacogenomic studies were unanticipated.

Hepatotoxicity is the most common reason for the termination of clinical trials investigating the efficacy of new drugs, accounting for approxi-mately 33% of such terminations, and is a major reason for postmarketing drug withdrawal.40 Floxacillin, an antibiotic used in Europe and Aus-tralia to treat staphylococcal infections, has been associated with an unusual form of cholestatic hepatitis, with an estimated incidence of approxi-mately 8.5 cases per 100,000 patients.41-44 A mul-ticenter genomewide association study, reported in 2009, analyzed the genotypes of 51 persons with floxacillin-induced hepatic injury and 282 matched controls.45 A SNP in the major histo-compatibility complex and closely linked with HLA-B*5701 showed very strong association with hepatic injury. The association between the pres-ence of HLA-B*5701 and hypersensitivity reactions to abacavir, a nucleoside analogue used to treat human immunodeficiency virus type 1 infection, had already been reported,46-48 which resulted in the FDA modification of the abacavir label to include a recommendation that patients undergo genotyping for HLA-B*5701 before the initiation of therapy.49 Rare but severe adverse events rep-resent a major reason why drugs are withdrawn after FDA approval. Although it was possible to attempt a replication of the association between the variant in HLA-B*5701 and floxacillin-induced hepatitis,45 it is often difficult to gather enough cases of rare adverse drug reactions to apply ge-nomewide techniques.

This situation presents a challenge for regula-tors. To date, the FDA has generally chosen to include pharmacogenetic information relevant to rare severe adverse events on drug labels — even when the association between the variant and drug response has not been replicated — so as to warn prescribers of potential risk.50 This ap-proach places a burden on clinicians to use their own judgment regarding the need for pharmaco-genetic testing before prescribing a drug. In con-trast with unreplicated tests for association are prospective trials of genotyping to avoid adverse pharmacogenetic effects. One such study is re-ported in this issue of the Journal,51 in which in-vestigators observed no instances of the Stevens–Johnson syndrome or toxic epidermal necrolysis in

a sample of nearly 5000 Taiwanese candidates for carbamazepine therapy, among whom carbam-azepine had been withheld from carriers of the HLA-B*1502 allele, which has been reported to be associated with the Stevens–Johnson syndrome in Han Chinese.52

Another pharmacogenomic example involving agents used to treat infectious diseases concerns the treatment of chronic infection with hepatitis C virus (HCV), which develops in approximately 80% of patients who are infected with the virus and is a major cause of liver failure.53,54 Success-ful treatment of chronic HCV infection involves a sustained virologic response, which is defined by an undetectable level of HCV RNA in plasma. Unfortunately, only 40 to 50% of patients who are infected with HCV genotype 1 have a sustained virologic response when receiving the current standard of care for the treatment of chronic HCV infection — injections of pegylated interferon alfa together with oral ribavirin for 48 weeks.53,54

The ability to identify patients with a differ-ential response to pegylated interferon alfa is important in the current era of new anti-HCV drugs because pegylated interferon alfa remains the backbone of therapy, to which many of these new agents are added. Recently, in three inde-pendent genomewide association studies55-57 in-volving patients with chronic HCV infection who were treated with pegylated interferon alfa and ribavirin, there was an association between a variant in IL28B, the gene encoding interleukin-28B, and the drug response. In one of these studies, peripheral-blood mononuclear cells from patients carrying the variant allele that was as-sociated with a poor response had comparatively low levels of IL28B expression.56 IL28B encodes a protein that is thought to be involved in suppress-ing the replication of a number of viruses, in-cluding HCV.55-58 This example shows how phar-macogenomic genomewide association studies not only have identified biomarkers of response to pegylated interferon alfa but also have provided insights that might be used to determine therapeu-tic approaches to this chronic infection and to se-lect a drug target for therapeutic development.

A n tineopl a s tic Drugs

The field of cancer pharmacogenomics is compli-cated by the fact that two genomes are involved: the germline genome of the patient and the so-matic genome of the tumor. Obviously, the tumor



genomic medicine


genome plays a critical role in the variation in response to antineoplastic therapy. Prominent ex-amples include HER2 overexpression or amplifi-cation in patients with breast cancer and the re-sponse of these tumors to trastuzumab59,60 and increased sensitivity to the epidermal growth fac-tor receptor (EGFR) antagonist gefitinib among patients with non–small-cell lung cancer who have activating mutations in the gene encoding EGFR.61,62 A recent example involves melanoma and a mutation in BRAF encoding a serine–threo-nine protein kinase. Since a specific inhibitor, PLX4032, targets the mutant activated kinase, there is a pharmacogenetic effect in that PLX4032 pro-longs survival in patients carrying the mutation. This clinical finding was based on the discovery of a BRAF mutation through the sequencing of a large number of kinase genes in tumors.63,64 On the other hand, germline SNPs in the gene encod-ing the enzyme thiopurine S-methyltransferase (TPMT) can result in increased sensitivity to mer-captopurine as a result of decreased metabo-lism,1,4,5 whereas the number of TA dinucleotide repeats in the promoter of UGT1A1 in germline DNA can increase the toxic effects of irinotecan, also as a result of decreased metabolism.1,65 There are now many examples of pharmacogenetic tests paired with anticancer drugs that are considered part of routine oncologic care (Table 1). The fact that clinically relevant pharmacogenomic varia-tion in both the tumor genome and the patient’s germline genome can influence the response to antineoplastic therapy is illustrated in Figure 3, with gefitinib and irinotecan as examples.

A rom ata se Inhibi t or s

Genetic polymorphisms in a patient’s germline genome can also play an important role in varia-tion in the response to cancer therapy. Endocrine therapy of breast cancer 66,67 offers a striking ex-ample of how a genomewide association study has lead to the identification of a mechanism that would seem to be responsible for a serious drug-induced adverse reaction that limits thera-peutic options for some patients.

The tumors of approximately 70% of post-menopausal women with breast cancer express the estrogen receptor. The blockade of this receptor with tamoxifen or the blockade of estrogen syn-thesis through the inhibition of aromatase (which catalyzes estrogen synthesis) halves the recur-

rence rate.66-68 However, the administration of an aromatase inhibitor can also result in severe mus-culoskeletal pain that leads women (10 to 20% in some studies) to terminate therapy.67 In a genome-wide association study that used DNA samples from a large clinical trial of aromatase inhibitors to treat women with breast cancer (called MA.27) (ClinicalTrials.gov number, NCT00968214), there was an association between musculoskeletal pain and variants in the gene cluster encoding T-cell leukemia–lymphoma (TCL) proteins. The marker showing the strongest (although not significant) association created a new estrogen-response ele-ment close to TCL1A.69 Functional studies showed that the markers that were associated with sus-ceptibility to musculoskeletal pain were also as-

Table 1. Anticancer Drugs Approved by the Food and Drug Administration (FDA) with Labeling Regarding Pharmacogenomic Biomarkers.*

Type of Biomarker and Associated Drug

Biomarker with pharmacokinetic effectTPMT

MercaptopurineThioguanine

UGT1A1IrinotecanNilotinib

Biomarker with pharmacodynamic effectEGFR

CetuximabErlotinibGefitinibPanitumumab

KRASCetuximabPanitumumab

ABLImatinibDasatinibNilotinib

C-Kit (KIT)Imatinib

HER2/neu (ERBB2)LapatinibTrastuzumab

Estrogen receptorTamoxifen

* Data are from the FDA’s pharmacogenetics Web site (www.fda.gov/Drugs/ScienceResearch/ResearchAreas/Pharmacogenetics/ucm083378.htm). The biomarkers have been separated into pharmacokinetic effect (drug metabolism) and pharmacodynamic effect (drug target). Biomarkers for cytogenetic alterations have been excluded.





sociated with increased TCL1A expression after es-trogen exposure. TCL1A regulates the expression of interleukin-17 receptor A, an experimental target for the treatment of patients with rheuma-toid arthritis.70 These observations, if confirmed, may provide new insight into the relationship between estrogens and joint pain.

This example illustrates several challenges and opportunities associated with pharmacogenomic studies and their application to clinical practice. First, associations that are uncovered by genome-wide association studies require replication if there are appropriate sample sets. However, MA.27 is a large clinical trial of aromatase inhibitors, span-

A Tumor genome

Gefitinib

Lung cancer

Tumor genome

EGFR activatingmutations

Increased tumorsensitivity to gefitinib

Gefitinib(kinase inhibitor)

EGF receptorEGF

Irinotecan

Irinotecan(prodrug)

SN-38(active drug)

SN-38glucuronide

(inactive metabolite)

SN-38(active drug)

Low expression of UGT1A1 andlow level of glucuronidation

High expression of UGT1A1 andhigh level of glucuronidation

B Germline genome

Healthy liver

Coloncancer

Germline genome

UGT1A1*28(TA)7TAA

UGT1A1(TA)6TAA

Germline genome

03/04/11

AUTHOR PLEASE NOTE:Figure has been redrawn and type has been reset

Please check carefully

AuthorFig #TitleME

DEArtist

Issue date

COLOR FIGURE

Rev5Dr. Weinshilboum

03/24/2011

3

PhimisterDaniel Muller

Figure 3. Cancer Pharmacogenomics and Tumor and Germline Genomes.

Both the tumor genome (e.g., in the case of gefitinib therapy) and the patient’s germline genome (e.g., in the case of irinotecan therapy) can contribute to pharmacogenomic variation in response to antineoplastic drugs. The tumor genome plays a critical role in the response to gefitinib (Panel A), since the sensitivity of non–small-cell lung cancer to this drug is enhanced by activating mutations in the kinase domain of the gene encoding epidermal growth factor receptor (EGFR).58,59 Tumor EGFR encoding activating mutations within the kinase domain results in enhanced tu-mor sensitivity to gefitinib. The rate of toxic effects associated with irinotecan (diarrhea and myelosuppression) is increased in patients with seven TA dinucleotide repeats rather than the more common six repeats in the promoter region of UGT1A1 encoding a UDP-glucuronosyltransferase in germline DNA, resulting in lower enzyme activity and a decreased rate of drug metabolism (Panel B).1,62



genomic medicine


ning 8 years at a cost of more than $35 million. Therefore, identifying a large and appropriate sample to test for replication will be difficult. In cases in which replication samples are not avail-able or are difficult to obtain, pharmacogenomic studies may benefit from the use of functional validation to help verify the results of genomewide studies. For example, the biologic plausibility that is provided by the functional data (i.e., the asso-ciation between phenotype-associated markers and TCL1A expression) increases confidence that the genetic association is driven by biology rather than chance. A final consideration is the clinical context. Because aromatase inhibitors have only a slight benefit over tamoxifen in the treatment of breast cancer, and tamoxifen is much less ex-pensive than aromatase inhibitors, a clear thera-peutic alternative is available for patients at in-creased risk for musculoskeletal pain. Therefore, a genetic test with sufficient predictive power to identify such patients might be clinically useful.

Clinic a l Tr a nsl ation

The use of genotyping to inform clinical decisions about drug use is not widely practiced. The slow pace of the clinical application of pharmacoge-nomics has many causes. Obviously, the most im-portant issue is the need to establish clinical util-ity in order to support the value of genotyping. In the absence of such evidence, payers will be un-likely to provide reimbursement for routine use of pharmacogenetic testing, and tests will remain inaccessible to the majority of patients. There seems to be little consensus on the level or nature of data required to establish clinical utility.7

No matter what level of evidence is required for each situation, it will be necessary to develop simple clinical algorithms to aid physicians in their interpretation and use of genetic data. This goal may be best achieved through the develop-ment of point-of-care tools embedded in elec-tronic medical record systems. Even with such tools, physicians and other health care providers need to be aware of this area of biomedical sci-ence in order to apply the information clinically. A major effort will be required to educate all members of the health care team about clinical genomics.

In recent years, the FDA has aggressively pur-sued drug-label modification when excess risk can be convincingly linked to a genetic marker. Several of the examples have been described here; many more are listed in the FDA’s Table of Pharma-cogenomic Biomarkers in Drug Labels (www .fda.gov/Drugs/ScienceResearch/ResearchAreas/Pharmacogenetics/ucm083378.htm). Warnings that the FDA has issued about the prescription of clopidogrel and abacavir without testing of the relevant genotype are examples of the agency’s increasingly activist stance.

Conclusions

There has been a good deal of comment in the scientific literature71-74 and the popular press75 about the slow pace of the application of genom-ics to clinical medicine. We hope that we have provided some reassurance that advances result-ing from the application of genomic science to drug therapy may be helpful in drug selection and administration and reduce the odds of adverse drug reactions. Challenges that are associated with the replication of study findings and the development of proof of the clinical significance of implicated variants underscore the importance of functional experiments to test for biologic plausibility and to extend our understanding of drug mechanisms. Finally, a blend of scientific, regulatory, and psychological factors must be ad-dressed if pharmacogenomic tests are to become a routine part of clinical practice. The FDA-man-dated incorporation of pharmacogenomic infor-mation in drug labeling will remain an impor-tant step in the acceptance of pharmacogenomics in clinical practice. Perhaps equally important will be the willingness of physicians to reexam-ine suboptimal pharmacologic management pro-grams.

Supported in part by grants (K22-CA130828, R01-CA138461, R01-GM28157, R01-CA132780, U01-HG005137, U19-GM61388, U01-GM63340, and UL1RR025747) from the National Institutes of Health, a PhRMA Foundation Center of Excellence in Clinical Pharmacology Award, and an American Society for Pharmacol-ogy and Experimental Therapeutics (ASPET)–Astellas Award.


We thank Luanne Wussow, Linda Pelleymounter, and Katar-zyna Ellsworth for their assistance in the preparation of the manuscript.





References

1. Weinshilboum RM, Wang L. Pharma-cogenetics and pharmacogenomics: devel-opment, science, and translation. Annu Rev Genomics Hum Genet 2006;7:223-45.2. Manolio TA. Genomewide association studies and assessment of the risk of dis-ease. N Engl J Med 2010;363:166-76.3. McCormack M, Alfirevic A, Bourgeois S, et al. HLA-A*3101 and carbamazepine-induced hypersensitivity reactions in Euro-peans. N Engl J Med 2011;364:1134-43.4. Weinshilboum R. Inheritance and drug response. N Engl J Med 2003;348: 529-37.5. Evans WE, McLeod HL. Pharmacoge-nomics — drug disposition, drug targets, and side effects. N Engl J Med 2003;348: 538-49.6. Feero WG, Guttmacher AE, Collins FS. Genomic medicine — an updated primer. N Engl J Med 2010;362:2001-11.7. Woodcock J. Assessing the clinical utility of diagnostics used in drug thera-py. Clin Pharmacol Ther 2010;88:765-73.8. Fuster V, Sweeny JM. Clopidogrel and the reduced-function CYP2C19 genetic variant: a limited piece of the overall ther-apeutic puzzle. JAMA 2010;304:1839-40.9. Mega JL, Simon T, Collet J-P, et al. Reduced-function CYP2C19 genotype and risk of adverse clinical outcomes among patients treated with clopidogrel predom-inantly for PCI: a meta-analysis. JAMA 2010;304:1821-30.10. Paré G, Mehta SR, Yusuf S, et al. Ef-fects of CYP2C19 genotype on outcomes of clopidogrel treatment. N Engl J Med 2010;363:1704-14.11. Daly AK, King BP. Pharmacogenetics of oral anticoagulants. Pharmacogenetics 2003;13:247-52.12. Budnitz DS, Shehab N, Kegler SR, Richards CL. Medication use leading to emergency department visits for adverse drug events in older adults. Ann Intern Med 2007;147:755-65.13. Wysowski DK, Nourjah P, Swartz L. Bleeding complications with warfarin use: a prevalent adverse effect resulting in regulatory action. Arch Intern Med 2007; 167:1414-9.14. Aithal GP, Day CP, Kesteven PJ, Daly AK. Association of polymorphisms in the cytochrome P450 CYP2C9 with warfarin dose requirement and risk of bleeding complications. Lancet 1999;353:717-9.15. Li T, Chang CY, Jin DY, Lin PJ, Khvo-rova A, Stafford DW. Identification of the gene for vitamin K epoxide reductase. Na-ture 2004;427:541-4.16. Rost S, Fregin A, Ivaskevicius V, et al. Mutations in VKORC1 cause warfarin re-sistance and multiple coagulation factor deficiency type 2. Nature 2004;427:537-41.17. Rieder MJ, Reiner AP, Gage BF, et al. Effect of VKORC1 haplotypes on tran-

scriptional regulation and warfarin dose. N Engl J Med 2005;352:2285-93.18. Takeuchi F, McGinnis R, Bourgeois S, et al. A genome-wide association study con-firms VKORC1, CYP2C9, and CYP4F2 as principal genetic determinants of warfa-rin dose. PLoS Genet 2009;5(3):e1000433.19. McDonald MG, Rieder MJ, Nakano M, Hsia CK, Rettie AE. CYP4F2 is a vitamin K1 oxidase: an explanation for altered warfarin dose in carriers of the V433M variant. Mol Pharmacol 2009;75:1337-46.20. Teichert M, Eijgelsheim M, Rivadeneira F, et al. A genome-wide association study of acenocoumarol maintenance dosage. Hum Mol Genet 2009;18:3758-68.21. Manolopoulos VG, Ragia G, Tavridou A. Pharmacogenetics of coumarinic oral anticoagulants. Pharmacogenomics 2010; 11:493-6.22. The International Warfarin Pharma-cogenetics Consortium. Estimation of the warfarin dose with clinical and pharma-cogenetic data. N Engl J Med 2009;360: 753-64. [Erratum, N Engl J Med 2009;361: 1613.]23. Sagreiya H, Berube C, Wen A, et al. Extending and evaluating a warfarin dos-ing algorithm that includes CYP4F2 and pooled rare variants of CYP2C9. Pharma-cogenet Genomics 2010;20:407-13.24. Zambon C-F, Pengo V, Padrini R, et al. VKORC1, CYP2C9 and CYP4F2 genetic-based algorithm for warfarin dosing: an Italian retrospective study. Pharmacoge-nomics 2010;12:15-25.25. Epstein RS, Moyer TP, Aubert RE, et al. Warfarin genotyping reduces hospital-ization rates: results from the MM-WES (Medco-Mayo Warfarin Effectiveness Study). J Am Coll Cardiol 2010;55:2804-12.26. Ginsburg GS, Voora D. The long and winding road to warfarin pharmacoge-netic testing. J Am Coll Cardiol 2010;55: 2813-5.27. Kanagasabapathy P, Chowdary P, Gatt A. Alternatives to warfarin — the next generation of anticoagulants. Cardiovasc Ther 2010 July 8 (Epub ahead of print).28. Hollopeter G, Jantzen HM, Vincent D, et al. Identification of the platelet ADP receptor targeted by antithrombotic drugs. Nature 2001;409:202-7.29. Savi P, Pereillo JM, Uzabiaga MF, et al. Identification and biological activity of the active metabolite of clopidogrel. Thromb Haemost 2000;84:891-6.30. Yusuf S, Zhao F, Mehta SR, et al. Ef-fects of clopidogrel in addition to aspirin in patients with acute coronary syndromes without ST-segment elevation. N Engl J Med 2001;345:494-502. [Errata, N Engl J Med 2001;345:1506, 1716.]31. Sabatine MS, Cannon CP, Gibson CM, et al. Addition of clopidogrel to aspirin and fibrinolytic therapy for myocardial in-

farction with ST-segment elevation. N Engl J Med 2005;352:1179-89.32. King SB III, Smith SC Jr, Hirshfeld JW Jr, et al. 2007 Focused Update of the ACC/AHA/SCAI 2005 Guideline Update for Per-cutaneous Coronary Intervention: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines: 2007 Writing Group to Review New Evidence and Update the ACC/AHA/SCAI 2005 Guideline Update for Percutaneous Coronary Intervention, writing on behalf of the 2005 Writing Committee. Circulation 2008;117:261-95. [Erratum, Circulation 2008;117(6):e161.]33. de Morais SM, Wilkinson GR, Blais-dell J, Nakamura K, Meyer UA, Goldstein JA. The major genetic defect responsible for the polymorphism of S-mephenytoin metabolism in humans. J Biol Chem 1994; 269:15419-22.34. Shuldiner AR, O’Connell JR, Bliden KP, et al. Association of cytochrome P450 2C19 genotype with the antiplatelet effect and clinical efficacy of clopidogrel ther-apy. JAMA 2009;302:849-57.35. Mega JL, Close SL, Wiviott SD, et al. Cytochrome P-450 polymorphisms and re-sponse to clopidogrel. N Engl J Med 2009; 360:354-62.36. FDA drug safety communication: re-duced effectiveness of Plavix (clopido-grel) in patients who are poor metaboliz-ers of the drug. (http://www.fda.gov/drugs/drugsafety/PostmarketDrugSafety InformationforPatientsandProviders/ucm203888.htm.)37. Holmes DR Jr, Dehmer GJ, Kaul S, Leifer D, O’Gara PT, Stein CM. ACCF/AHA clopidogrel clinical alert: approaches to the FDA “boxed warning”: a report of the American College of Cardiology Founda-tion Task Force on clinical expert consen-sus documents and the American Heart Association endorsed by the Society for Cardiovascular Angiography and Interven-tions and the Society of Thoracic Sur-geons. J Am Coll Cardiol 2010;56:321-41.38. Wallentin L, James S, Storey RF, et al. Effect of CYP2C19 and ABCB1 single nu-cleotide polymorphisms on outcomes of treatment with ticagrelor versus clopido-grel for acute coronary syndromes: a ge-netic substudy of the PLATO trial. Lancet 2010;376:1320-8.39. Paré G, Mehta SR, Yusuf S, et al. Ef-fects of CYP2C19 genotype on outcomes of clopidogrel treatment. N Engl J Med 2010;363:1704-14.40. Mann R, Andrews E, eds. Pharmaco-vigilance. 2nd ed. Chichester, England: John Wiley and Sons, 2006.41. Olsson R, Wiholm BE, Sand C, Zetter-gren L, Hultcrantz R, Myrhed M. Liver damage from flucloxacillin, cloxacillin and dicloxacillin. J Hepatol 1992;15:154-61.42. Fairley CK, McNeil JJ, Desmond P, et



genomic medicine


al. Risk factors for development of flu-cloxacillin associated jaundice. BMJ 1993; 306:233-5. [Erratum, BMJ 1993;307:1179.]43. Russmann S, Kaye JA, Jick SS, Jick H. Risk of cholestatic liver disease associated with flucloxacillin and flucloxacillin pre-scribing habits in the UK: cohort study using data from the UK General Practice Research Database. Br J Clin Pharmacol 2005;60:76-82.44. Pachkoria K, Lucena MI, Crespo E, et al. Analysis of IL-10, IL-4 and TNF-alpha polymorphisms in drug-induced liver in-jury (DILI) and its outcome. J Hepatol 2008;49:107-14. [Erratum, J Hepatol 2009; 50:636.]45. Daly AK, Donaldson PT, Bhatnagar P, et al. HLA-B*5701 genotype is a major determinant of drug-induced liver injury due to flucloxacillin. Nat Genet 2009;41: 816-9.46. Hetherington S, McGuirk S, Powell G, et al. Hypersensitivity reactions during therapy with the nucleoside reverse tran-scriptase inhibitor abacavir. Clin Ther 2001;23:1603-14.47. Mallal S, Nolan D, Witt C, et al. As-sociation between presence of HLA-B*5701, HLA-DR7, and HLA-DQ3 and hypersensitivity to HIV-1 reverse-transcrip-tase inhibitor abacavir. Lancet 2002;359: 727-32.48. Mallal S, Phillips E, Carosi G, et al. HLA-B*5701 screening for hypersensitiv-ity to abacavir. N Engl J Med 2008;358: 568-79.49. Information for healthcare profes-sionals: abacavir (marketed as Ziagen) and abacavir-containing medications. (http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationfor PatientsandProviders/ucm123927.htm.)50. Ferrell PB Jr, McLeod HL. Carbamaze-pine, HLA-B*1502 and risk of Stevens-Johnson syndrome and toxic epidermal necrolysis: US FDA recommendations. Pharmacogenomics 2008;9:1543-6.51. Chen P, Lin J-J, Lu C-S, et al. HLA-

B*1502 screening to prevent carbamaze-pine-induced Stevens–Johnson syndrome. N Engl J Med 2011;364:1126-33.52. Chung WH, Hung SI, Hong HS, et al. Medical genetics: a marker for Stevens-Johnson syndrome. Nature 2004;428:486.53. NIH consensus statement on manage-ment of hepatitis C. NIH Consens State Sci Statements 2002;19(3):1-46.54. Shepard CW, Finelli L, Alter MJ. Global epidemiology of hepatitis C virus infec-tion. Lancet Infect Dis 2005;5:558-67.55. Suppiah V, Moldovan M, Ahlenstiel G, et al. IL28B is associated with response to chronic hepatitis C interferon-alpha and ribavirin therapy. Nat Genet 2009;41: 1100-4.56. Tanaka Y, Nishida N, Sugiyama M, et al. Genome-wide association of IL28B with response to pegylated interferon-alpha and ribavirin therapy for chronic hepatitis C. Nat Genet 2009;41:1105-9.57. Ge D, Fellay J, Thompson AJ, et al. Ge-netic variation in IL28B predicts hepatitis C treatment-induced viral clearance. Na-ture 2009;461:399-401.58. O’Brien TR. Interferon-alfa, interferon-lambda and hepatitis C. Nat Genet 2009; 41:1048-50.59. Eisenhauer EA. From the molecule to the clinic — inhibiting HER2 to treat breast cancer. N Engl J Med 2001;344: 841-2.60. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a mono-clonal antibody against HER2 for meta-static breast cancer that overexpresses HER2. N Engl J Med 2001;344:783-92.61. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth factor receptor underlying respon-siveness of non–small-cell lung cancer to gefitinib. N Engl J Med 2004;350:2129-39.62. Paez JG, Jänne PA, Lee JC, et al. EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy. Sci-ence 2004;304:1497-500.63. Davies H, Bignell GR, Cox C, et al.

Mutations of the BRAF gene in human cancer. Nature 2002;417:949-54.64. Flaherty KT, Puzanov I, Kim KB, et al. Inhibition of mutated, activated BRAF in metastatic melanoma. N Engl J Med 2010; 363:809-19.65. Iyer L, King CD, Whitington PF, et al. Genetic predisposition to the metabolism of irinotecan (CPT-11): role of uridine di-phosphate glucuronosyltransferase iso-form 1A1 in the glucuronidation of its active metabolite (SN-38) in human liver microsomes. J Clin Invest 1998;101:847-54.66. Freedman RA, Winer EP. Adjuvant therapy for postmenopausal women with endocrine-sensitive breast cancer. Breast 2010;19:69-75.67. Janni W, Hepp P. Adjuvant aromatase inhibitor therapy: outcomes and safety. Cancer Treat Rev 2010;36:249-61.68. Early Breast Cancer Trialists’ Collab-orative Group (EBCTCG). Effects of che-motherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the ran-domised trials. Lancet 2005;365:1687-717.69. Ingle JN, Schaid DJ, Goss PE, et al. Genome-wide associations and functional genomic studies of musculoskeletal ad-verse events in women receiving aroma-tase inhibitors. J Clin Oncol 2010;28: 4674-82.70. Miossec P, Korn T, Kuchroo VK. Inter-leukin-17 and type 17 helper T cells. N Engl J Med 2009;361:888-98.71. The human genome at ten. Nature 2010;464:649-50.72. Collins F. Has the revolution arrived? Nature 2010;464:674-5.73. Venter JC. Multiple personal genomes await. Nature 2010;464:676-7.74. Varmus H. Ten years on — the human genome and medicine. N Engl J Med 2010;362:2028-9.75. Wade N. A decade later, genetic map yields few new cures. New York Times. June 12, 2010.Copyright © 2011 Massachusetts Medical Society.

SPECIALTIES AND TOPICS AT NEJM.ORG

Specialty pages at the Journal’s Web site (NEJM.org) feature articles in cardiology, endocrinology, genetics, infectious disease, nephrology, pediatrics, and many other medical specialties. These pages, along with collections of articles on clinical and nonclinical topics, offer links to interactive and multimedia content and feature

recently published articles as well as material from the NEJM archive (1812–1989).



images in clinical medicine



A 44-year-old man with diabetes mellitus presented to our hospi-tal after 4 days of fever and abdominal pain. The initial evaluation revealed tachycardia (heart rate, 137 beats per minute), hypotension (blood pressure,

81/44 mm Hg), and abdominal discomfort in the right upper quadrant. There was no rebound tenderness. A lesion with heterogeneous radiodensity was noted in the right upper abdomen on chest radiography (Panel A, arrowheads). Computed tomo-graphic imaging revealed an intrahepatic lesion containing gas and fluid (Panel B, arrowheads). A pyogenic liver abscess was suspected. Blood cultures ultimately grew Klebsiella pneumoniae, which can be gas-producing. Diabetes is an important risk factor for this condition. Despite fluid resuscitation and treatment with ino-tropic agents and antibiotics, the patient’s clinical condition deteriorated, and he died within 48 hours after admission.Copyright © 2011 Massachusetts Medical Society.

Pyogenic Liver Abscess

Chin-Wei Yu, M.D.Ching-Hsing Lee, M.D.

Chang Gung Memorial HospitalTaoyuan, Taiwan

A B



case records of the massachusetts general hospital



Founded by Richard C. CabotNancy Lee Harris, m.d., Editor Eric S. Rosenberg, m.d., Associate EditorJo-Anne O. Shepard, m.d., Associate Editor Alice M. Cort, m.d., Associate EditorSally H. Ebeling, Assistant Editor Christine C. Peters, Assistant Editor

From the Divisions of Rheumatology, Allergy, and Immunology (M.R.M.) and Cardiology (D.M.D.); and the Depart-ments of Radiology (J.Y.S.) and Pathology (R.P.H.), Massachusetts General Hospital; the Division of Rheumatology, Allergy, and Immunology, Brigham and Women’s Hospital (M.C.C.); and the Departments of Medicine (M.R.M., M.C.C., D.M.D.), Radiology (J.Y.S.), and Pathology (R.P.H.), Harvard Medical School — all in Boston.


Pr esen tation of C a se

Dr. Leana S. Wen (Emergency Medicine): A 37-year-old man was admitted to the hos-pital because of flushing and hypotension. That morning, sneezing, rhinorrhea, scratchy throat, and subjective fever had developed. After lunch, he took an over-the-counter cold preparation that included acetylsalicylic acid, dextromethorphan hy-drobromide, and phenylephrine bitartrate. Thirty minutes later, fatigue, weakness, nausea, epigastric pain, facial flushing, and “beefy red” ears developed, accompa-nied by two episodes of nonbloody emesis. Light-headedness, diaphoresis, and near-syncope developed. He was caught by a coworker as he fell to the ground, without head trauma or loss of consciousness. Emergency medical services were called. On evaluation, his skin was flushed; the systolic blood pressure was 50 mm Hg, the pulse 56 beats per minute and regular, the respiratory rate 16 breaths per minute, and the oxygen saturation 100% while he was breathing supplemental oxygen by means of a nonrebreather face mask. An electrocardiogram (ECG) showed sinus bra-dycardia with T-wave inversions. Ondansetron and normal saline were administered intravenously. He was transported to the emergency department at this hospital.

On arrival, the patient reported pleuritic chest pain and worsening diffuse abdominal discomfort. He had had multiple similar episodes during the previous 12 years, with flushing, conjunctival injection, vomiting, and diarrhea. These epi-sodes had increased in frequency in the past year, from approximately twice a year to once every 2 months. The symptoms, which were usually provoked by physical exertion, mental stress, or intense emotion, lasted up to 12 hours and were fol-lowed by weakness of 3 to 4 days’ duration. Evaluations at other hospitals had shown systolic pressures as low as 60 mm Hg. The episodes were attributed to de-hydration and stress, and they were treated with intravenous normal saline (≤5 liters). The patient had no other illnesses and no known allergies to medications and took no other medications. Kiwifruit had caused throat constriction and vomiting. He worked in a health-related field and participated in triathlons. He drank alcohol occasionally, had smoked in the past, and did not use illicit drugs or over-the-counter herbal preparations. There were no recent exposures to ill persons, seafood,

Case 9-2011: A 37-Year-Old Man with Flushing and Hypotension

Mandakolathur R. Murali, M.D., Mariana C. Castells, M.D., Ph.D., James Y. Song, M.D., David M. Dudzinski, M.D., J.D.,

and Robert P. Hasserjian, M.D.





or unusual foods. He was of mixed European and Caribbean ancestry. There was no family history of renal disease, anaphylaxis, or rheuma-tologic diseases.

On examination, the patient was alert and ori-ented. The temperature was 36.2°C (over the tem-poral artery), the blood pressure 57/33 mm Hg, the pulse up to 116 beats per minute and irregu-lar, the respiratory rate 26 breaths per minute, and the oxygen saturation 93% while he was breathing 4 liters of oxygen by nasal cannula. There was scleral injection and diffuse blanching erythema over the upper body; the extremities were warm and well perfused, with normal cap-illary refill. The examination was otherwise nor-mal. Levels of sodium, chloride, carbon dioxide, D-dimer, magnesium, amylase, lipase, creatine kinase MB isoenzymes, and troponin T; tests of liver function; the activated partial-thromboplas-tin time; and the red-cell indexes were normal. Screening for troponin I and serum toxins was negative. The results of other laboratory tests are shown in Table 1.

An ECG showed atrial fibrillation with a ven-tricular rate of 102 beats per minute, with ST-segment elevations (2 mm, convex) in leads aVR and V1 and downsloping ST-segment depressions (3 to 4 mm) in leads I, II, III, aVF, and V3 through V6 (Fig. 1 in the Supplementary Appendix, avail-able with the full text of this article at NEJM.org). Normal saline was transfused rapidly through two peripheral intravenous catheters, and 325 mg of aspirin was administered. The patient then had increased flushing. Fifty minutes after his arrival, the blood pressure was 140/81 mm Hg. A chest radiograph showed low lung volumes with inter-stitial prominence and was otherwise normal.

During the next 7 hours, the temperature rose to 38.1°C and abdominal pain increased, with tenderness in the upper abdomen. Normal saline, potassium chloride, magnesium sulfate, ondan-setron, and metoclopramide were administered intravenously, and his nausea transiently im-proved. Urinalysis showed yellow clear urine, with trace ketones, trace urobilinogen, 1+ biliru-bin, and 2+ albumin, with 20 to 50 white cells (reference range, 0 to 2) and a few squamous cells (reference range, no cells) per high-power field. There were more than 100 hyaline casts (reference range, 0 to 5) and 10 to 20 granular casts (reference range, none) and mucin per low-power field. The urinalysis was otherwise nor-

mal. The oxygen saturation increased to 100% while the patient was breathing ambient air.

Eight hours after the patient’s arrival, an ECG showed sinus rhythm, at 100 beats per minute, and no ST-segment depressions (Fig. 2 in the Supplementary Appendix. Computed tomography (CT) of the abdomen and pelvis with intravenous and oral contrast material showed hypodense lesions in the liver and spleen that were consis-tent with hemangiomas; the study was otherwise normal. Ciprofloxacin was begun. He was ad-mitted to the cardiac telemetry unit. The tem-perature rose to 39.4°C, systolic blood pressure decreased to 80 to 90 mm Hg, oxygen saturation decreased to 89% while he was breathing ambi-ent air, and vomiting occurred. Oxygen supple-mentation (4 liters) was restarted, and he was transferred to the medical intensive care unit.

Two hours later (18 hours after the patient’s arrival at the hospital), the systolic blood pres-sure decreased to 75 mm Hg. Diphenhydramine, ranitidine, phenylephrine, hydrocortisone, epi-nephrine, acetaminophen, vancomycin, ceftriax-one, and oseltamivir were administered. Trans-thoracic echocardiography showed an estimated ejection fraction of 76%, a mildly dilated right ventricle, right ventricular hypertrophy with nor-mal right ventricular function, right atrial dilata-tion, trace mitral regurgitation, trace pulmonary insufficiency, and findings consistent with mild pulmonary-valve stenosis. Examination of a blood smear showed no malarial forms, and testing for antibodies to the human immunodeficiency virus and Borrelia burgdorferi, a nasal swab for influenza virus types A and B, and stool specimens for white cells, ova and parasites, rotavirus, and Clostridium difficile toxin were negative. The erythrocyte sedi-mentation rate and tests of thyroid function were normal; other results are shown in Table 1. Hypo-tension resolved within 2 hours, and phenyleph-rine was discontinued. The patient was trans-ferred to the medical floor on the third day.

During the next 3 days, tests for antinuclear antibodies, hepatitis viruses (A, B, and C), and Helicobacter pylori were negative, cultures of spec-imens of blood and urine remained sterile, and a stool culture grew normal enteric flora. Antibi-otics were discontinued. Abdominal ultrasonog-raphy showed two lesions consistent with hem-angiomas in the liver and nonspecific thickening of the gallbladder wall.

On the sixth day, a test result was received.



case records of the massachusetts gener al hospital


Differ en ti a l Di agnosis

Dr. Mandakolathur R. Murali: Dr. Dudzinski, would you describe the cardiac studies?

Dr. David M. Dudzinski: The initial ECG (Fig. 1 in the Supplementary Appendix) shows atrial fibril-lation with a rapid ventricular response and an Ashman beat (a wide QRS complex after a short

Table 1. Laboratory Data.*

Variable Reference Range, Adults†On

Admission12 Hr after Admission

18 Hr after Admission

38 Hr after Admission

Hematocrit (%) 41.0–53.0 (men) 40.8 34.4 32.6 32.1

Hemoglobin (g/dl) 13.5–17.5 (men) 13.3 11.4 10.8 10.6

White-cell count (per mm3) 4500–11,000 13,800 18,200 22,300 11,200

Differential count (%)

Neutrophils 40–70 59 87 92 85

Lymphocytes 22–44 30 1 4 10

Monocytes 4–11 2 0 4 3

Eosinophils 0–8 1 0 0 2

Band forms 0–10 7 12 0 0

Metamyelocytes 0 1 0 0 0

Platelet count (per mm3) 150,000–400,000 327,000, large forms 225,000 294,000 270,000

Prothrombin time

Seconds 10.8–13.4 16.7 17.8

International normalized ratio 1.5 1.6

Potassium (mmol/liter) 3.4–4.8 2.6 4.2 3.9 4.0

Urea nitrogen (mg/dl) 8–25 11 22 26 11

Creatinine (mg/dl) 0.60–1.50 1.35 2.30 2.20 1.14

Estimated glomerular filtration rate (ml/min/1.73 m2)

>60 >60 34 36 >60

Glucose (mg/dl) 70–110 122 170 168 143

Protein (g/dl)

Total 6.0–8.3 5.8 5.7 5.7 5.9

Albumin 3.3–5.0 3.4 3.5 3.4 3.2

Globulin 2.6–4.1 2.4 2.2 2.3 2.7

Phosphorus (mg/dl) 2.6–4.5 2.8 1.7 2.9 2.0

Calcium (mg/dl) 8.5–10.5 9.2 7.8 7.7 7.9

Lactic acid (mmol/liter) 0.5–2.2 1.8 3.6 2.7 1.4

NT-pro-BNP (pg/ml) 0–450 (<50 yr) 679

C-reactive protein (mg/liter) <8.0 57.3

Cortisol (µg/dl) <10.0 (between 8 p.m. and 8 a.m.) 29.6

Complement (mg/dl)

C3 86–184 47

C4 20–58 13

* To convert the values for urea nitrogen to millimoles per liter, multiply by 0.357. To convert the values for creatinine to micromoles per liter, multiply by 88.4. To convert the values for glucose to millimoles per liter, multiply by 0.05551. To convert the values for calcium to milli-moles per liter, multiply by 0.250. To convert the values for phosphorus to millimoles per liter, multiply by 0.3229. To convert the values for cortisol to nanomoles per liter, multiply by 27.59. NT-pro-BNP denotes N-terminal fragment of pro−brain (B-type) natriuretic peptide.

† Reference values are affected by many variables, including the patient population and the laboratory methods used. The ranges used at Massachusetts General Hospital are for adults who are not pregnant and do not have medical conditions that could affect the results. They may therefore not be appropriate for all patients.





RR interval following a long RR interval) and diffuse ST-segment depression with ST-segment elevation in leads V1 and aVR. Explanations for profound ST-segment changes include epicardial coronary disease, vasospasm, metabolic distur-bances, cardiomyopathy, and subendocardial ischemia — especially in a patient with a criti-cally low blood pressure.1 Echocardiographic studies were obtained for evaluation of persis-tent hypotension (Videos 1 and 2, available at NEJM.org, and Fig. 3 in the Supplementary Ap-pendix). There was mild right ventricular hyper-trophy and dilatation with diastolic septal flat-tening, right atrial dilatation, moderate tricuspid regurgitation, mild pulmonic stenosis with dom-ing of the valve,2 and an estimated right ventricu-lar systolic pressure of 43 mm Hg. These find-ings led to the consideration of carcinoid heart disease, but in carcinoid heart disease, pulmonic regurgitation is common (80%), and almost all patients have thickened, shortened tricuspid leaf-lets; neither of these findings was present in this case.3 Echocardiographic examination repeated after the patient was euvolemic was unchanged except for a reduction in tricuspid regurgitation to trace, a decrease in right ventricular systolic pressure to 29 mm Hg, and resolution of right ventricular dilatation. Although a pulmonary em-bolus was not formally ruled out, there was no echocardiographic explanation for this patient’s persistent hypotension.

Dr. Murali: I am aware of the diagnosis in this case. In arriving at a diagnosis, a systematic analysis of the events in a timeline (both histori-cal and clinical evolution) could form a useful matrix.

Episodes of flushing, light-headedness, vom-iting, diarrhea, and hypotension began 12 years before admission, precipitated by physical and mental stress. Hypotension lasted about 12 hours, but generalized weakness lasted 3 to 4 days. From a twice-yearly occurrence, these episodes escalat-ed to once every 2 months in the year preceding this admission. This hospitalization was pre-ceded by an upper respiratory infection, and the episode occurred 30 minutes after ingestion of a pill containing acetylsalicylic acid. The prominent features were facial f lushing, “beefy red” ears, vomiting, and chest tightness, with ECG changes, diaphoresis, and hypotension leading to near-syncope. The theme that emerges from this time-

line is a recurrent and escalating systemic dis-ease with flushing and hypotension as the salient clinical features.

Although the historical timeline provides the scaffold on which to build the differential diag-nosis, the evolution of the clinical features in the emergency department and telemetry unit leads us to a clinical diagnosis. The cardinal features on examination were scleral injection, diffuse blanching erythema over the upper body, warm and well-perfused skin, hypotension, atrial fibril-lation and rapid ventricular response, and ST-segment changes suggestive of ischemia — in the absence of hives, angioedema, and wheezing. Administration of acetylsalicylic acid during re-suscitation exacerbated the flushing, without wheezing or urticaria. Therapy with intravenous fluids, diphenhydramine, ranitidine, phenyleph-rine, hydrocortisone, acetaminophen, and anti-microbial agents resulted in resolution of the hypotension. Thus, disorders characterized by episodic flushing, emesis, hypotension, constitu-tional features of weakness, and depression, with escalation in severity and precipitation by physical and mental stress and acetylsalicylic acid, need to be considered in the differential diagnosis.

FlushingFlushing and hypotension are clues to the diag-nosis. Flushing is a sensation of warmth accom-panied by transient erythema, usually over the face, neck, ears, chest, and limbs and is due to vasodilatation with increased cutaneous blood flow. Vasodilatation is mediated either by neuro-genic (autonomic) regulation of cutaneous vascu-lar smooth muscle or by direct action of vasodila-tor stimuli such as histamine, substance P, and prostaglandins (e.g., PGD2) on endothelial cells.4,5 The resultant vasodilatation and increased vascu-lar permeability contribute to distributive shock, which occurred in this patient.6 The effects of histamine acting through H1 and H2 receptors include vasodilatation and vascular permeability, thus explaining the patient’s flushing and hypo-tension. Histamine acting through the H3 recep-tor affects neurogenic vasodilatation through lo-cal neuron–mast-cell feedback loops, and its effect on the central nervous system includes alterations in emotions and memory as well as cognitive ab-normalities, possibly explaining the depression and weakness that followed the acute episodes.7

Videos of echocardio-

graphic studies are available at

NEJM.org





A further clue to the diagnosis comes from the characteristics of the flushing — that is, wheth-er it is associated with sweating (wet flushing) or not (dry f lushing). Flushing due to neurogenic stimuli is accompanied by sweating, because of autonomic innervation of the eccrine sweat glands. This process occurs with fever, exercise, heat ex-posure, and menopause, as well as with neuro-logic disorders with autonomic dysfunction such as post-encephalitic syndromes and Parkinson’s disease. These entities are easily excluded in this case. Vasodilator substances, in contrast, cause dry f lushing, which is what this patient had.4,5 Vasodilator stimuli can be exogenous or endog-enous, and we need to decide which applies in this case (Table 2).

AnaphylaxisCould this patient’s illness be a manifestation of recurrent anaphylaxis due to a food allergy, a med-ication, or even an idiopathic cause? Anaphylaxis is a systemic mast-cell–activation or basophil-activation syndrome in which mediators are re-leased from normal mast cells, usually because of exposure to an exogenous substance such as food or medication. In 2003, the World Allergy Orga-nization proposed that anaphylaxis be defined as a “severe, life-threatening generalized or system-ic hypersensitivity reaction” that may be immuno-logic (whether it is IgE-mediated or not) or non-immunologic (Table 2). Kiwifruit shares antigens with latex, and the patient’s reaction to kiwifruit is suggestive of an oral allergy syndrome; howev-er, his flushing and hypotensive episodes were not precipitated by the ingestion of kiwifruit. Cutane-ous hives, angioedema, or both are present in more than 90% of patients with anaphylaxis, and respiratory tract involvement (hoarseness, stridor, rhinitis, or wheezing) is present in 40 to 60%.7 In the absence of these manifestations and with no documented association of previous episodes with specific foods, medications, or latex, anaphylaxis due to any of these causes is unlikely in this case. Acetylsalicylic acid can potentiate mast-cell acti-vation resulting from other causes, exacerbating flushing in this patient.

This leaves us with idiopathic anaphylaxis — a syndrome of recurrent anaphylaxis that is not as-sociated with a known trigger.8 Before we accept this as a cause, we need to examine possible en-dogenous causes of this patient’s symptoms.

Sepsis

The rapid clinical deterioration (with fever, flush-ing, tachycardia, and hypotension with a poor re-sponse to fluid therapy) and the neutrophilic leu-kocytosis with “bandemia” (an elevated level of band forms of white cells) and an increased pro-thrombin time suggest a systemic response to in-flammatory cytokines, as is seen in sepsis.9 Right-upper-quadrant tenderness, oliguria, lesions in the liver, and the thickened gallbladder wall sug-gested an infectious process. This suspicion led to immediate therapy with broad-spectrum antibi-otics. The patient was admitted during the H1N1 influenza pandemic, and his associated upper

Table 2. Causes of Vasodilator-Mediated Flushing Accompanied by Hypotension.*

Causes Types and Associated Features

Exogenous

Medications Nicotinic acid, calcium-channel blockers, phosphodiesterase-5 inhibitors, vanco-mycin, angiotensin-converting−enzyme inhibitors

Foods

Pharmacologic Capsaicin, ethanol, sulfites, and mono-sodium glutamate

Toxic Scombroidosis from histamine and cis-urocanic acid formed in bacteria-contaminated spoiled fish such as tuna, mackerel, and mahimahi

Anaphylaxis

Immunologic IgE- and FcεR1-mediated stimuli such as allergy to penicillin, insect venom, latex, heterologous serum, and chimeric monoclonal antibodies; non−IgE-mediated reactions to stimuli such as blood products, acetylsalicylic acid, ra-diocontrast mediums, and some drugs

Nonimmunologic Physical exercise, cold stimuli, opiates, curare

Primary or idiopathic Undetected cause

Endogenous

Sepsis

Carcinoid syndrome

Mastocytosis

Medullary carcinoma of thyroid

VIP-secreting tumors

Pheochromocytoma

Idiopathic systemic capillary leak syndrome

* VIP denotes vasoactive intestinal peptide.





respiratory infection led his care team to admin-ister oseltamivir. Inflammatory cytokines lead to hepatic production of acute-phase proteins such as C-reactive protein, the level of which was ele-vated in this patient.9 However, the recurrent na-ture and temporal profile of these episodes im-plicate a noninfectious systemic disease. Negative serologic tests and cultures over the next 3 days ruled out infectious causes of this illness.

Flushing Syndromes Produced by Endocrine Tumors

Neuroendocrine tumors, including pheochromo-cytoma, vasoactive intestinal peptide–producing tumors, medullary thyroid carcinoma, and carci-noid tumors, may all secrete substances that cause flushing, hypotension or hypertension, diarrhea, and respiratory symptoms in various combina-tions. The characteristics of the flushing episodes are atypical for pheochromocytoma, the absence of a thyroid nodule makes medullary thyroid car-cinoma unlikely, and the remitting symptoms and protracted course make VIPoma unlikely. Analy-sis to detect mediators of these syndromes was nonetheless warranted and was negative.

carcinoid syndromeCould this patient have the carcinoid syndrome, which is characterized by cutaneous flushing, di-arrhea, wheezing, and cardiac valvular lesions? Episodes are often precipitated by the ingestion of alcohol and chocolate and do not occur after exercise or the use of acetylsalicylic acid, as in this patient. Facial telangiectasia and cyanosis and pellagra-like skin changes may be seen in chronic cases.10 None of these features were noted in this patient. Echocardiographic findings raised the question of carcinoid heart disease, but the find-ings were also consistent with cardiac changes in a triathlete or volume overload during resuscita-tion. The abdominal CT was negative for ileal or appendicular masses, and the liver lesions were consistent with hemangiomas, not metastases. These features make the carcinoid syndrome un-likely, but it should be ruled out with a 24-hour urine test for 5-hydroxyindoleacetic acid.

Idiopathic Systemic Capillary Leak SyndromeIdiopathic systemic capillary leak syndrome is characterized by hypotension, hypoalbuminemia, and hemoconcentration. This condition is often preceded by an upper respiratory infection and

features of distributive shock initially with warm, flushed skin. It is a diagnosis of exclusion, and the biochemical profile is absent in this patient.11

MastocytosisThis patient’s presentation is consistent with a mast-cell activation syndrome, but it is not typical of anaphylaxis. The clinical features of flushing and hypotension with involvement of the cardio-vascular, gastrointestinal, and nervous systems in the absence of urticaria, angioedema, and upper-airway involvement suggest systemic mastocyto-sis, a mast-cell neoplasm (Table 3). In the absence of hematologic abnormalities, hepatosplenomeg-aly, and tissue dysfunction, the clinical diagnosis is indolent systemic mastocytosis.

Determination of the serum tryptase level is essential to establish the diagnosis of systemic mastocytosis and differentiate it from anaphy-laxis. Some other laboratory features could be explained by the diagnosis of mastocytosis. Ac-tive monomers of β-tryptase can generate ana-phylatoxins C3a, C4a, and C5a in vitro.13,14 This patient had decreased levels of C3 and C4, which could be due to an elevated tryptase level. The elevated C-reactive protein level might be a re-flection of the systemic effect of interleukin-1 secreted from the activated mast cells. Bone mar-row examination will be important in confirming the diagnosis.

Dr. Eric S. Rosenberg (Pathology): May we have the medical students’ diagnosis?

A Harvard Medical Student: We considered allergic and immunologic, hematologic, vascular, neuro-genic, and endocrine disorders. Our differential diagnosis included a carcinoid syndrome, pheo-chromocytoma, the hypereosinophilic syndrome, allergic reaction, thyrotoxicosis, and mastocyto-sis. We thought mastocytosis the most likely diagnosis.

Dr. Rosenberg: Dr. Vassallo, would you tell us what your clinical impression was and what di-agnostic tests were performed?

Dr. Milo Vassallo (Allergy and Immunology): I interpreted the clinical history to be most con-sistent with recurrent episodes of anaphylaxis. In the differential diagnosis, I considered idiopathic anaphylaxis, but urticaria and respiratory symp-toms were never prominent in this patient. The flushing improved rapidly in the medical inten-sive care unit with treatment with antihistamines, implicating histamine and mast cells rather than





other causes of flushing such as carcinoid and gastrointestinal tract tumors.

The diagnostic test was analysis of the serum tryptase level. This result reflects both the total mast-cell burden and mast-cell activation. There are two assayable forms of tryptase: total tryp-tase, which is preformed in the mast-cell gran-ules, and β-tryptase, which is formed by prote-olysis on activation and degranulation of mast cells.15 The total tryptase level is useful in the diagnosis of anaphylaxis only if measured within hours after the onset of symptoms. The patient’s total tryptase level in a specimen of blood drawn approximately 3 hours after the onset of symp-toms was 2983 ng per milliliter (reference range, <11.5); the β-tryptase level was 1350 ng per mil-liliter (reference range, <1). A total serum tryp-tase level measured 6 days later, when the patient was asymptomatic, was 54.8 ng per milliliter, and the β-tryptase level was less than 1 ng per milliliter; this result indicates an increased mast-cell burden.

These findings provided support for a diag-nosis of indolent systemic mastocytosis. The test to confirm the diagnosis was bone marrow bi-opsy and aspiration.

Clinic a l Di agnosis

Indolent systemic mastocytosis complicated by anaphylaxis due to aspirin.

Dr . M a nda kol athur R . Mur a li’s Di agnosis

Acute manifestation of indolent systemic masto-cytosis triggered by upper respiratory infection and aspirin.

Pathol o gic a l Discussion

Dr. Robert P. Hasserjian: The bone marrow–biopsy specimen contained multiple large aggregates of pale cells with oval nuclei, accounting for about

Table 3. World Health Organization Categories of Mastocytosis.*

Category Definition

Cutaneous mastocytosis Mast-cell infiltrates with four clinical variants; criteria for systemic mastocytosis not met (see below)

Urticaria pigmentosa

Diffuse cutaneous mastocytosis

Cutaneous mastocytoma

Telangiectasia macularis eruptiva perstans

Systemic mastocytosis Major criterion: multifocal, dense aggregates of mast cells (≥15) in sections of bone marrow, another extracutaneous organ or organs, or both; minor criteria (one required; three required if major criteri-on absent): >25% of mast cells morphologically atypical, expres-sion of CD2 or CD25 by mast cells, presence of KIT codon 816 mutation, serum total tryptase >20 mg/ml

Indolent systemic mastocytosis No mast-cell−related organ dysfunction; no associated hematologic non−mast-cell lineage disorder (see below); skin lesions usually present

Systemic mastocytosis with an associated hematologic non−mast-cell lineage disorder

Evidence of one of the following: myelodysplastic myeloproliferative neoplasm, myelodysplastic syndrome, acute myeloid leukemia, or lymphoid neoplasm (lymphoma or plasma-cell myeloma)

Aggressive systemic mastocytosis Organ dysfunction due to mast-cell infiltration (in bone marrow, liver, spleen, gastrointestinal tract, or bones); skin lesions usually absent

Mast-cell leukemia >10% immature mast cells in blood or >20% in bone marrow

Extracutaneous mastocytoma Solitary mast-cell tumor without cytologic atypia

Mast-cell sarcoma Solitary mast-cell tumor with high-grade cytologic atypia

* Data are from Horny et al.12





20% of the overall marrow cellularity (Fig. 1A and 1B). The remaining marrow was markedly hyper-cellular and showed trilineage maturing hemato-poiesis with increased eosinophils. Immunohis-tochemical studies revealed that the abnormal cells were mast cells expressing mast-cell tryptase and CD117 (the product of the KIT gene) (Fig. 1C). Unlike normal mast cells, these cells expressed CD25 (Fig. 1D) and CD2. The bone marrow aspi-rate showed normal hematopoiesis with no dys-plasia or increased blasts. Rare abnormal spindle-shaped mast cells were present (Fig. 1B, inset). Flow cytometry showed no abnormal cells. Cytoge-netic analysis revealed a normal karyotype (46,XY). Fluorescence in situ hybridization (FISH) showed no rearrangement of the Fip1-like 1 gene (FIP1L1) or the gene that encodes platelet-derived growth

factor receptor α (PDGFRA). Subsequently, poly-merase-chain-reaction assay of a peripheral-blood sample revealed a point mutation at codon 816 (Asp-816→Val) of the KIT gene.

The diagnosis of systemic mastocytosis is based on both clinical and pathological features (Table 3).15 The presence of multiple large mast-cell aggregates in the bone marrow of this patient constitutes a major criterion for the diagnosis of systemic mastocytosis. In addition, there was prominent mast-cell atypia (spindle-shaped cells with hypogranulation), aberrant expression of CD2 and CD25, a KIT codon 816 mutation, and elevated serum tryptase levels, fulfilling all diag-nostic criteria for systemic mastocytosis.

Bone marrow biopsy is an important diagnos-tic test for systemic mastocytosis, since the large

A B

DC

Figure 1. Findings on Examination of Bone Marrow Aspirate.

The bone marrow–biopsy specimen contains multiple aggregates of pale cells that are often located adjacent to bone trabeculae surrounding a central core of mature lymphocytes (Panel A, hematoxylin and eosin). The pale cells have oval nuclei and abundant cytoplasm and are admixed with numerous eosinophils (Panel B, hematoxylin and eosin); on the aspirate smear, the rare mast cells are abnormally spindled with uneven cytoplasmic granulation (in-set, Wright–Giemsa stain). Immunohistochemical analysis shows that the mast cells express the lineage-specific marker tryptase (Panel C) and aberrantly express CD25 (Panel D).





mast-cell aggregates are highly characteristic of the disease. As in this case, mast cells may be rare in the aspirate smears because of fibrosis associated with the mast-cell aggregates.15 When the diagnosis of systemic mastocytosis was sug-gested on the basis of the bone marrow exami-nation in this patient, additional considerations arose in the differential diagnosis.

Systemic mastocytosis is associated with a clonal non–mast-cell hematologic neoplasm (e.g., a myelodysplastic syndrome or chronic myelo-monocytic leukemia) or acute myeloid leukemia in 30 to 40% of cases (Table 3).16,17 In these cases, the prognosis is typically determined by that of the associated non–mast-cell neoplasm. Many of these myeloid neoplasms have been shown to share KIT mutation and cytogenetic abnormalities with the mast-cell proliferation, indicating an origin from a common precursor cell.18,19 In this patient, there was no morpho-logic evidence of an associated myelodysplastic syndrome, myeloproliferative neoplasm, or acute leukemia, and the karyotype was normal.

Another important consideration in the diag-nosis of systemic mastocytosis is the exclusion of a myeloid neoplasm with FIP1L1–PDGFRA rear-rangement. This neoplasm is characterized by diffusely increased bone marrow mast cells, in-creased serum tryptase levels, and eosinophilia, and it may mimic systemic mastocytosis. Many patients with this neoplasm have a response to specific targeted therapies, such as imatinib me-sylate, that are ineffective in systemic mastocy-tosis.16 The presence of aggregated (rather than diffusely scattered) mast cells, a KIT mutation, and a negative FISH study ruled out a myeloid neoplasm with FIP1L1–PDGFRA rearrangement.

Dr. Dudzinski: The prominent yet transient ST-segment changes observed on the admission ECG may be a manifestation of epicardial coronary vasospasm due to supraphysiologic concentra-tions of histamine from mast-cell degranulation. This mechanism, variously termed hypersensitiv-ity coronary syndrome, allergic angina, or the Kounis syndrome, has been described in response to a host of allergic stimuli and can be mani-fested as ECG changes and chest pain.20-22

Dr. Vassallo: Upper gastrointestinal endoscopy on the sixth day showed duodenal ulcers. The patient remained in stable condition and was discharged on the eighth day. He was prescribed ranitidine and cetirizine for prophylactic man-

agement of flushing and pruritus and epineph-rine autoinjectors for use during acute reactions. Two days later, CT of the chest was performed.

Dr. James Y. Song: CT of the chest performed after the administration of intravenous contrast material for evaluation of pulmonary embolism (Fig. 2A) shows loss of corticomedullary differ-entiation and coarse trabeculae in the thoracic

A

B

C

Figure 2. Imaging Studies of the Chest.

An axial CT scan of the chest (Panel A) obtained after the administration of contrast material shows loss of corticomedullary differentiation and heterogeneous osteosclerosis within the 10th thoracic vertebral body (arrow). The bilateral fifth ribs show patchy bone scle-rosis (Panel B, arrows) and there is a focal lytic lesion (≤7 mm in diameter) within the left posterior eighth rib (Panel C, arrow). An additional lytic lesion was de-scribed in the right third rib (not shown). The diffuse sclerotic and lytic appearance is characteristic of mas-tocytosis.





spine. The ribs show scattered regions of patchy sclerosis (Fig. 2B) alternating with normal mar-row attenuation, and there is a focal lytic lesion within the left eighth rib (Fig. 2C). This mixed lytic and sclerotic infiltrative appearance in the axial skeleton is consistent with mastocytosis. In retrospect, similar subtle changes were probably present in the spine on the abdominal and pelvic CT scan obtained on admission.

Dr. Rosenberg: Dr. Castells, can you discuss the care of this patient with systemic mastocytosis?

Discussion of M a nagemen t

Dr. Mariana C. Castells: With its variable clinical phenotypes,23 systemic mastocytosis is a great masquerader, and as in this patient, the onset of symptoms can precede the diagnosis by many years — a median of 9.5 years in one study.24 Exercise, alcohol, trauma, infections, contrast dyes, medications including nonsteroidal antiin-flammatory drugs (NSAIDs) and antibiotics, an-esthesia, and surgery can induce release of mast-cell mediators, triggering symptomatic episodes. When I questioned this patient, he said that his triggers were alcohol, stress, emotions, infections, foods with high fat content, and NSAIDs. He also had chronic symptoms, including daily flushing, chronic fatigue, depression, anxiety, bone pain, fractures, and chest pain that had led to multiple cardiac evaluations.

Urticaria pigmentosa is associated with indo-lent systemic mastocytosis in more than 80% of all cases; scratching of these reddish brown mac-ules, which are scattered over the body except for the palms, soles, and scalp, triggers urtica-tion and erythema, known as Darier’s sign. If skin findings are overlooked or subtle, the other-wise nonspecific symptoms may lead to a fruit-less search for other causes. When I examined this patient after discharge, I noted several brown lesions on the anterior chest, with a positive Darier’s sign, findings compatible with urticaria pigmentosa.

The management of mastocytosis depends on the clinicopathological subtype. Reduction of the mast-cell burden is indicated in cases of aggres-sive mastocytosis and mast-cell leukemia, and treatment of the non–mast-cell hematologic dis-order is indicated in cases of mastocytosis with such a disorder. This patient has indolent sys-temic mastocytosis, and treatment in these pa-tients is aimed at avoiding triggers, limiting re-lease of mediators, and blocking their actions. For this patient, I recommended H1- and H2-histamine–receptor blockade (cetirizine and ra-nitidine); oral disodium cromoglycate, which blocks the release of mediators from mast cells25; leukotriene-receptor blockade (montelukast); and a proton-pump inhibitor (omeprazole).26 He car-ries two epinephrine self-injectable devices to treat hypotensive events. Imatinib would not be useful, since its site of action is abrogated by the D816V mutation in this patient. Unfortunately, the patient has been noncompliant with his medi-cations, despite extensive counseling, and he has had two further episodes of flushing and hypo-tension during the 15 months since the diagnosis was established. Both episodes were treated in the emergency department with intravenous fluids and antihistamines.

Dr. Steven E. Goldfinger (Gastroenterology): I note that the patient was once given epinephrine for his flushing. The fact that profound hypotension did not develop tends to rule out carcinoid tu-mors, because epinephrine can precipitate a car-cinoid crisis, which could result in death.

A nat omic a l Di agnosis

Indolent systemic mastocytosis.

This case was presented at the Medicine Grand Rounds.Dr. Castells reports receiving grant support from Ovations for

the Cure and consulting and other fees from Schering-Plough; Dr. Hasserjian, consulting fees from Genzyme; and Dr. Dudzin-ski, honoraria from Lippincott Williams & Wilkins. No other potential conflict of interest relevant to this article was reported.


References

1. Habibzadeh MR, Ewy GA. ST-seg-ment elevation in lead aVR greater than that in lead V(1) secondary to diffuse myocardial ischemia from prolonged hy-potension. Clin Cardiol 2007;30:251-4.2. Weyman AE, Hurwitz RA, Girod DA, Dillon JC, Feigenbaum H, Green D. Cross-

sectional echocardiographic visualization of the stenotic pulmonary valve. Circula-tion 1977;56:769-74.3. Pellikka PA, Tajik AJ, Khandheria BK, et al. Carcinoid heart disease: clinical and echocardiographic spectrum in 74 pa-tients. Circulation 1993;87:1188-96.

4. Izikson L, English JC III, Zirwas MJ. The flushing patient: differential diagno-sis, workup, and treatment. J Am Acad Dermatol 2006;55:193-208.5. Wilkin JK. Flushing reactions: conse-quences and mechanisms. Ann Intern Med 1981;95:468-76.





6. Kaliner M, Sigler R, Summers R, Shel-hamer JH. Effects of infused histamine: analysis of the effects of H-1 and H-2 his-tamine receptor antagonists on cardio-vascular and pulmonary responses. J Al-lergy Clin Immunol 1981;68:365-71.7. Lieberman PL. Anaphylaxis. In: Ad-kinson NF Jr, Bochner BS, Busse WW, Holgate ST, Lemanske RF Jr, Simons FER, eds. Middleton’s allergy: principles & practice. 7th ed. Vol. 2. Philadelphia: Mosby, 2009:1027-49.8. Ditto AM, Harris KE, Krasnick J, Mill-er MA, Patterson R. Idiopathic anaphy-laxis: a series of 335 cases. Ann Allergy Asthma Immunol 1996;77:285-91.9. Bone RC. Immunologic dissonance: a continuing evolution in our understand-ing of the systemic inflammatory re-sponse syndrome (SIRS) and the multiple organ dysfunction syndrome (MODS). Ann Intern Med 1996;125:680-7.10. Modlin IM, Kidd M, Latich I, Zikusoka MN, Shapiro MD. Current status of gas-trointestinal carcinoids. Gastroenterology 2005;128:1717-51.11. Dhir V, Arya V, Malav IC, Suryanaray-anan BS, Gupta R, Dey AB. Idiopathic systemic capillary leak syndrome (SCLS): case report and systematic review of cases reported in the last 16 years. Intern Med 2007;46:899-904.12. Horny H-P, Metcalfe DD, Bennett JM, et al. Mastocytosis. In: Swerdlow S, Cam-po E, Harris NL, et al., eds. WHO classifi-

cation of tumours of haematopoietic and lymphoid tissues. Lyon, France: IARC Press, 2008:53-63.13. Fukuoka Y, Schwartz LB. Active monomers of human beta-tryptase have expanded substrate specificities. Int Im-munopharmacol 2007;7:1900-8.14. Schwartz LB, Kawahara MS, Hugli TE, Vik D, Fearon DT, Austen KF. Genera-tion of C3a anaphylatoxin from human C3 by human mast cell tryptase. J Immunol 1983;130:1891-5.15. Schwartz LB. Diagnostic value of tryptase in anaphylaxis and mastocytosis. Immunol Allergy Clin North Am 2006;26: 451-63.16. Pardanani A, Lim KH, Lasho TL, et al. Prognostically relevant breakdown of 123 patients with systemic mastocytosis as-sociated with other myeloid malignan-cies. Blood 2009;114:3769-72.17. Travis WD, Li CY, Yam LT, Bergstralh EJ, Swee RG. Significance of systemic mast cell disease with associated hemato-logic disorders. Cancer 1988;62:965-72.18. Miranda RM, Tang G, Jones DM, Huh YO, Medeiros LJ, Wang SA. Systemic mas-tocytosis with associated clonal hemato-logical non-mast cell lineage disorder (SM-AHNMD) shows distinct hematolog-ic features and reveals a common clonal origin of myeloid and mast cells. Mod Pathol 2010;23:Suppl:311A. abstract.19. Sotlar K, Colak S, Bache A, et al. Vari-able presence of KITD816V in clonal hae-

matological non-mast cell lineage diseases associated with systemic mastocytosis (SM-AHNMD). J Pathol 2010;220:586-95.20. Kounis NG. Kounis syndrome (aller-gic angina and allergic myocardial infarc-tion): a natural paradigm? Int J Cardiol 2006;110:7-14.21. Kounis NG, Kounis GN, Kouni SN, Soufras GD, Niarchos C, Mazarakis A. Al-lergic reactions following implantation of drug-eluting stents: a manifestation of Kounis syndrome? J Am Coll Cardiol 2006;48:592-3.22. Rohr SM, Rich MW, Silver KH. Short-ness of breath, syncope, and cardiac ar-rest caused by systemic mastocytosis. Ann Emerg Med 2005;45:592-4.23. Castells M, Austen KF. Mastocytosis: mediator-related signs and symptoms. Int Arch Allergy Immunol 2002;127:147-52.24. Horan RF, Austen KF. Systemic mas-tocytosis: retrospective review of a de-cade’s clinical experience at the Brigham and Women’s Hospital. J Invest Dermatol 1991;96:Suppl:5S-14S, 60S-65S.25. Soter NA, Austen KF, Wasserman SI. Oral disodium cromoglycate in the treat-ment of systemic mastocytosis. N Engl J Med 1979;301:465-9.26. Escribano L, Akin C, Castells M, Schwartz LB. Current options in the treat-ment of mast cell mediator-related symp-toms in mastocytosis. Inflamm Allergy Drug Targets 2006;5:61-77.Copyright © 2011 Massachusetts Medical Society.

Lantern Slides Updated: Complete PowerPoint Slide Sets from the Clinicopathological Conferences

Any reader of the Journal who uses the Case Records of the Massachusetts General Hospital as a teaching exercise or reference material is now eligible to receive a complete set of PowerPoint slides, including digital images, with identifying legends, shown at the live Clinicopathological Conference (CPC) that is the basis of the Case Record. This slide set contains all of the images from the CPC, not only those published in the Journal. Radiographic, neurologic, and cardiac studies, gross specimens, and photomicrographs, as well as unpublished text slides, tables, and diagrams, are included. Every year 40 sets are produced, averaging 50-60 slides per set. Each set is supplied on a compact disc and is mailed to coincide with the publication of the Case Record.

The cost of an annual subscription is $600, or individual sets may be purchased for $50 each. Application forms for the current subscription year, which began in January, may be obtained from the Lantern Slides Service, Department of Pathology, Massachusetts General Hospital, Boston, MA 02114 (telephone 617-726-2974) or e-mail [email protected].




e d i t o r i a l


Choice of Bronchodilator Therapy for Patients with COPDJadwiga A. Wedzicha, M.D.

Chronic obstructive pulmonary disease (COPD) is a progressive airway inflammatory condition that is associated with accelerated decline of lung function and is characterized by worsening dys-pnea with episodes of increased number and se-verity of symptoms, termed exacerbations.1 The main objectives of managing COPD are a reduc-tion in the severity of symptoms and the preven-tion of exacerbations.2 Bronchodilator therapy is central to the management of COPD; this treat-ment produces modest increases in lung function as measured by spirometry and a reduction in the dynamic hyperinflation that leads to the dys-pnea of COPD.3 Patients with COPD often pres-ent to their physician with dyspnea, and broncho-dilators are often the first therapy prescribed. Initially, short-acting bronchodilators, such as al-buterol, may be used, but in patients with persis-tent symptoms, long-acting bronchodilators pro-vide more uniform relief. Currently there are two classes of long-acting bronchodilators available — long-acting β2-agonists and long-acting anti-cholinergic agents, and both classes of agents, as compared with placebo, have been shown to pro-vide relief from symptoms in patients with COPD.

Although exacerbations are more common with increasing severity of the disease, a substan-tial number of patients, even those with moder-ate COPD (defined as COPD in which the patient’s forced expiratory volume in 1 second [FEV1] is 50% or more of the predicted value) are suscep-tible to frequent exacerbations.4 Patients who have frequent exacerbations are also likely to have more symptoms, worse health status,5 faster dis-ease progression,6 and an increased risk of death7; therefore, it is important to make the diagnosis of COPD in these patients accurately and to treat them effectively. Subgroup analyses of recent large COPD trials have shown that treatment with long-

acting β2-agonists or long-acting anticholiner-gic agents, in addition to decreasing the fre-quency of exacerbations in patients with severe disease, also reduce exacerbations in patients with moderate COPD.8,9

Since long-acting bronchodilators benefit these patients, it is important to ask which long-acting bronchodilator should be the initial choice for patients with COPD, especially those with mod-erate disease. The National Institute for Health and Clinical Excellence of England and Wales, in its 2010 update of COPD treatment guidelines, reviewed all studies that compared long-acting β2-agonists and long-acting anticholinergic agents and came to the conclusion that there was no evidence to favor one treatment over another.10

In this issue of the Journal, Vogelmeier and colleagues report the results of a study that makes some progress in addressing this impor-tant question.11 The Prevention of Exacerbations with Tiotropium in COPD trial (POET-COPD; ClinicalTrials.gov number, NCT00563381), a large international study, compared the effect of a long-acting anticholinergic agent (tiotropium, at a dose of 18 µg once daily) with a long-acting β2-agonist (salmeterol, at a dose of 50 µg twice daily) over the course of 1 year in 7376 patients with COPD who had an FEV1 of 70% of the predicted value or less. The focus of the study was the reduction of exacerbations, and the enrollment criteria in-cluded a documented history of at least one ex-acerbation in the previous year requiring therapy or hospitalization. The primary end point was the time to the first exacerbation, and secondary and safety end points included other exacerbation outcomes and serious adverse events, including death. The results showed that, as compared with salmeterol, tiotropium prolonged the time to the first exacerbation, with a 17% reduction in the



editorial


risk of an exacerbation. There was also a greater reduction with tiotropium than with salmeterol in the other exacerbation outcomes, such as the annual number of moderate exacerbations (re-quiring treatment with systemic glucocorticoids, antibiotics, or both) and the annual number of severe exacerbations (requiring hospitalization). The incidence of serious adverse effects, includ-ing cardiovascular complications and deaths, was similar between the two groups.

An important point to note about this trial was that it was not a direct comparison of a long-acting β2-agonist with a long-acting anticholin-ergic agent, since concomitant medications were allowed; more than 50% of the patients were re-ceiving, on a regular basis, inhaled glucocorti-coids that also reduce exacerbations.8 The fact that the patients had had at least one treated exacerbation in the previous year may explain the high use of concomitant medications among the patients in this trial. However, post hoc analyses showed that the effect of tiotropium with respect to the reduction in exacerbations was similar re-gardless of the severity of the COPD and regard-less of whether the patients were receiving in-haled glucocorticoid therapy on a regular basis.

Large multicenter studies involving patients with COPD are often difficult to perform, and previous large trials have attempted to study a wide range of outcomes. In contrast, the focus on COPD exacerbations in the study by Vogelmeier et al. enabled the data on exacerbations to be carefully collected and validated. This trial thus provides a good model for future COPD trials, which should be focused on a specific and rele-vant disease outcome. A strength of the study was that the reporting of exacerbation outcomes was detailed and was supplemented with daily diary cards to confirm exacerbations and the intervals between exacerbations, thus providing more ac-curate event rates.

The main implications of this trial are for the initial care of symptomatic patients with mod-erate disease and a history of recent exacerba-tions. The trial evidence suggests that with respect to exacerbation outcomes, tiotropium, adminis-tered once daily, is superior to salmeterol, admin-istered twice daily.11 However, novel, once-daily, long-acting β2-agonists such as indacaterol are now becoming available in Europe, and there is some evidence that outcomes with indacaterol are similar to those with tiotropium.12 There is

no evidence for the superiority of tiotropium in patients with mild COPD (those in whom the FEV1 is >70% of the predicted value) or sympto-matic patients with moderate COPD but without a history of exacerbations. However, in patients with progressive COPD, combinations of inhaled long-acting β2-agonists, long-acting anticholin-ergic agents, glucocorticoids, and new antiin-flammatory agents such as oral phosphodiester-ase-4 inhibitors may be indicated. Future trials involving patients with COPD will need to study which therapies and which specific combinations are optimal for which COPD phenotypes and disease severities, so that we can reduce the ad-verse effects of this disabling disease.

Disclosure forms provided by the author are available with the full text of this article at NEJM.org.

From the Academic Unit of Respiratory Medicine, University College London Medical School, University College London, London.

1. Wedzicha JA, Seemungal TAR. COPD exacerbations: defin-ing their cause and prevention. Lancet 2007;370:786-96.2. Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management and prevention of chronic obstructive pulmonary disease: 2009 update. (http://www.goldcopd.com.)3. O’Donnell DE, Revill SM, Webb KA. Dynamic hyperinflation and exercise intolerance in chronic obstructive pulmonary dis-ease. Am J Respir Crit Care Med 2001;164:770-7.4. Hurst JR, Vestbo J, Anzueto A, et al. Susceptibility to exacer-bation in chronic obstructive pulmonary disease. N Engl J Med 2010;363:1128-38.5. Seemungal TAR, Donaldson GC, Paul EA, Bestall JC, Jeffries DJ, Wedzicha JA. Effect of exacerbation on quality of life in pa-tients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 1998;157:1418-22.6. Donaldson GC, Seemungal TAR, Bhowmik A, Wedzicha JA. The relationship between exacerbation frequency and lung func-tion decline in chronic obstructive pulmonary disease. Thorax 2002;57:847-52. [Erratum, Thorax 2008;63:753.]7. Soler-Cataluña JJ, Martínez-García MA, Román Sánchez P, Salcedo E, Navarro M, Ochando R. Severe acute exacerbations and mortality in patients with chronic obstructive pulmonary disease. Thorax 2005;60:925-31.8. Jenkins CR, Jones PW, Calverley PM, et al. Efficacy of salmet-erol/f luticasone propionate by GOLD stage of chronic obstruc-tive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study. Respir Res 2009;10:59.9. Decramer M, Celli B, Kesten S, Lystig T, Mehra S, Tashkin DP. Effect of tiotropium on outcomes in patients with moderate chronic obstructive pulmonary disease (UPLIFT): a prespecified subgroup analysis of a randomised controlled trial. Lancet 2009;374:1171-8.10. CG101 chronic obstructive pulmonary disease (update): full guideline. London: National Institute for Health and Clinical Excellence, 2010. (http://guidance.nice.org.uk/CG101/Guidance.)11. Vogelmeier C, Hederer B, Glaab T, et al. Tiotropium versus salmeterol for the prevention of exacerbations of COPD. N Engl J Med 2011;364:1093-103.12. Donohue JF, Fogarty C, Lötvall J, et al. Once-daily broncho-dilators for chronic obstructive pulmonary disease: indacaterol versus tiotropium. Am J Respir Crit Care Med 2010;182:155-62.Copyright © 2011 Massachusetts Medical Society.




s o u n d i n g b o a r d


Specialization, Subspecialization, and Subsubspecialization in Internal Medicine

Christine K. Cassel, M.D., and David B. Reuben, M.D.

At a time when most authorities believe that the country desperately needs more generalists, the American Board of Internal Medicine (ABIM) is adding new subspecialties. Specifically, in the past 2 years the ABIM has launched certification in the fields of hospice and palliative care and advanced heart failure and has begun a process for internal-medicine certification with a focused practice in hospital medicine. The ABIM has also approved the subspecialty of adult congeni-tal heart disease to move forward to the Ameri-can Board of Medical Specialties (ABMS) for fi-nal approval. In addition, the ABIM has received requests from specialty societies to approve sev-eral new subspecialties, including medical infor-matics, clinical pharmacology, vascular medicine, addiction medicine, and obesity medicine. Each of these applications raises issues of a societal nature (i.e., the benefits to the public of having clear standards for emerging areas of medical specialization) versus issues of practicality (e.g., the cost of creating and maintaining certifica-tion examinations and the ongoing worry about fragmentation of care). These issues have been coupled with the concerns of different special-ties that favored or opposed particular subspe-cialty designations. Most important, we receive clear but contradictory messages from physi-cians: on the one hand, “recognize what I do” (i.e., create a subspecialty for my niche practice); on the other hand, “stop fragmenting an already overfragmented system.”

Meanwhile, accountability standards for phy-sicians are proliferating, and maintaining board certification is requiring more of physicians.1 Other certifying boards, such as those of surgery and pediatrics, face similar pressures. The per-spective of the ABIM may be generalized to other disciplines. To understand the ABIM’s decision-making process for adding new specialties, it is

important to examine the historical and current forces behind the drive for additional examina-tions and performance assessments.

History of Specializ ation

Specialization has characterized scientific prog-ress in medicine for nearly two centuries. With the centennial celebration of the hugely influen-tial Flexner Report,2 great attention has been paid to Flexner’s call for educational reform. Un-derlying his call for more research-based educa-tion was the growth of a scientific consensus of allopathic medicine and the physiology-based and biochemistry-based understanding of hu-man illness leading to ever more specialization in the pursuit of greater knowledge and exper-tise. Articles that appeared in the Journal 3,4 in 1936 and 1950 identified the crea tion of board certification of specialty status as an “inge-nious” way for the profession, independent of government, to control the “dangers of special-ism.” The authors were most concerned about unqualified practitioners claiming to be special-ists and were particularly alarmed by the prac-tice of specialist advertising, considered to be unethical at the time.

As medical scientists specialized and devoted their intellectual energies to understanding more and more about narrower topic areas, general prac-titioners differentiated into physicians with spe-cific areas of expertise, devoting some or all of their work to that specific area. The first medical specialty to create its own assessment board was ophthalmology in 1917. Prompted by the growth of optometry as a separate discipline, the American Medical Association and the American Ophthalmological Society created an independent board of specialists to create standards that would recognize physicians whose knowledge and skills





demonstrated expertise in identifying and treat-ing disorders of the eye.

Four specialties created a federation called the ABMS in 1933, which encompassed 10 specialties by 1935; the ABIM was added in 1936. By the 1970s, there were 20 specialties, including primary surgical boards in orthopedics, urology, neurosur-gery, plastic surgery, and colorectal surgery. In-ternal medicine kept the unifying requirement of training in general internal medicine but un-derwent just as much subspecialization during that time. The ABIM approved requests for 4 subspecialties in the 1940s, another 6 in the 1970s, and 10 more since then (Fig. 1). Some of these subsubspecialties are built on further spe-cialization of large subspecialties such as cardiol-ogy and pulmonary medicine, whereas others, including sleep medicine, sports medicine, and geriatric medicine, span multiple disciplines and allow different pathways to certification from other boards.

Throughout these decades, some leaders voiced concern that the growing fragmentation of medical care5 would result in the loss or under-valuing of the personal or generalist physician, who was perceived as being essential to good patient care. With more and more specializa-tion, they worried that the generalists’ practice would become too limited in scope and an un-attractive choice for residents. This concern spawned the creation of a new kind of specialty in 1969 — a generalist discipline in family med-

icine (called family practice at the time). It also led to calls for strengthening and repopulating general internal medicine.5 At the same time, other leaders in medicine saw growing special-ization as strengthening internal medicine.6

Current Driving Forces

Now we are witnessing a resurgence of interest in new specialty designations and a simultaneous eruption of concern about diminishing strength and numbers in primary care specialties. What are the driving forces in the current environ-ment? Do these forces make a stronger case for the value of new specialties? Do they outweigh the concerns about cost, fragmentation, and phy-sician burden? What factors are determinative in these decisions?

As in earlier times, the progress of biomedi-cal science continues to be a major factor in the emergence of new subspecialties. Advanced heart failure and transplant hepatology are two exam-ples growing from the need for extended medi-cal care before and after transplantation, includ-ing device management in the case of cardiology. Creation of these subspecialties reflects the rec-ognition that there are some specific popula-tions of patients who would benefit from highly focused knowledge and skills obtained by ad-ditional training and certification beyond that of a general cardiologist or gastroenterologist. These tertiary subspecialists also require high

1

2/24/11



AuthorFig #

Title

ME

DEArtist

Issue date

COLOR FIGURE

Draft 3Cassel

Knoper

3/24/11

FortiJMD

Decade1930 1940 1950 1960 1970 1980 1990 2010 20202000

Allergy(moved to ABAI in 1971)

Cardiovascular disease

Tuberculosis (pulmonary disease after 1946–1948)

Gastroenterology

Medical oncology

Critical caremedicine

Geriatric medicine

Clinical cardiac electrophysiology

Sports medicine

Adolescentmedicine

Interventionalcardiology

Transplanthepatology

Sleep medicine

Hospice and palliative medicine

Advanced heart failure and transplantcardiology

Rheumatology

Nephrology

Infectious disease

Hematology

Endocrinology,diabetes, and

metabolism1941

1972

1973

1987 1988

1992

1993

1994

19992008

2007

2006

2010

Figure 1. Timeline of Subspecialties Approved by the American Board of Internal Medicine.

The American Board of Allergy and Immunology (ABAI) was founded in 1971.




sounding board

patient volume to maintain proficiency in their skills and are therefore usually based at referral centers. Certification for these subspecialties is limited to specialists with training in the under-lying organ-system specialty, which must be maintained. Conversely, disciplines such as ger-iatric medicine, palliative medicine, and hospi-tal medicine are based on clinical needs and the organization and delivery of care rather than on scientific and technical expertise in a specific organ system. Many new and emerging subspe-cialties are cross-disciplinary; sleep medicine crosses six different specialties, and palliative medicine is a subspecialty option for 10 differ-ent primary specialties. We anticipate that med-ical informatics will probably extend just as broadly across specialties.

The demand for board certification comes in part from the profession itself, with the major-ity of survey respondents reporting that “pro-fessional image” was the primary reason they sought or renewed certification.7 All the pro-posals for new subspecialties have come to the board initially from specialty societies, often with support from patient groups. In the cur-rent environment of growing demands for higher standards and greater transparency, health plans and hospitals are also seeking ways to ensure that physicians have the knowl-edge and skills required for patients to have reasonable confidence in their capabilities, and board certification is one of the criteria fre-quently used for this purpose. Some health care organizations and medical groups that employ physicians require that they be certified to join the medical staff.

Although the public values board certifica-tion,8 most people do not understand what cri-teria it represents or fully appreciate that differ-ent kinds of organizations offer certificates that represent varying degrees of rigor and clinical relevance. As evidence of interest in this type of information, most Internet-based “report cards” include statements about a physician’s board certification. It is not surprising that many new organizations have emerged that offer their own versions of a certificate. We believe that the cri-teria and requirements for certification in a med-ical specialty should be public and transparent, and both the ABMS and the ABIM post such in-formation on their respective Web sites.

Criteria Used to Establish New Specialties of Internal Medicine

The policies of the ABIM for establishing new areas of specialization have been “repeatedly and exhaustively re-examined”6 over the years, result-ing in two successive documents (in 1993 and 2006) to guide the board in deliberations about new subspecialties. The criteria currently used in considering a request for new subspecialty status in internal medicine are articulated in the 2006 report entitled New and Emerging Disciplines in Inter-nal Medicine — 2 (NEDIM–2).9 These criteria focus on evidence that the new discipline has a defin-able body of knowledge and a substantial number of clinical training programs, with the reason-able expectation that clinical services in the sub-specialty will play a beneficial role in patient care (Table 1). Such designations have usually required at least 1 year of accredited training. Subspecialty applications from clinical pharmacology, vascu-lar medicine, addiction medicine, and obesity medicine have not been approved to date because they failed to meet one or more of these criteria or they were deemed insufficiently mature, as reflected by the number of training programs or practitioners in the field.

Changes in Pr ac tice over the Course of a C areer

In the introduction to the 1998 updated version of her classic book American Medicine and the Public Interest: A History of Specialization,10 Rosemary Stevens opined, “Arguably, the structure of the medical profession is moving toward a system of special-ties defined by the job market rather than by the professional system of specialist qualifications.” If you add “defined by the practice area in which the physician focuses and in which the patient ex-pects expertise,” Stevens’s prediction of a specialty-medicine structure may be correct. Clear evidence of specific competencies is a core component of the consumer demand for transparency, and phy-sician specialists themselves are asking to be cer-tified in narrower areas of expertise than they were originally trained in. In response to these developments, maintenance of certification (MOC) — the process through which time-limited cer-tificates are renewed — might be becoming more fluid, reflecting what NEDIM–2 calls “rec-





ognition of focused practice.” The first example of this is the ABIM’s offering of an identified hospitalist pathway for maintaining certification in internal medicine with a focused practice in hospital medicine. It is a rigorous pathway, with clear requirements for a large enough number of patients, quality and safety metrics, and an ex-amination on hospital-based medicine. Although this pathway is approved by ABMS as a pilot of-fering, the ABIM and other boards, such as those in ophthalmology and radiology, have re-ceived additional requests for this kind of fo-cused MOC option.

Greater specificity in certification and recog-nition of focused practice should be appealing to consumers who want to know the areas in which their physicians are skilled and experi-enced; taken to its logical conclusion, however, it could unravel the traditional specialty-medi-cine structure that Stevens described. For exam-ple, should an endocrinologist who has focused only on diabetes and, to maintain certification, has focused only on demonstrating current knowledge and performance in management of diabetes still be considered an endocrinologist? The ABMS’s role has been to define these spe-cialty areas and, in the process, to respect the breadth of competence the public could expect from that specialist. Taking the focused-practice trend too far would mean additional, smaller areas of competence, although such a focus may

be of greater interest to the patient seeking care for a specific condition. If this approach were carried out thoughtfully, the original certificate would identify the areas of formal training with the understanding that the focus of practice could change over the course of a career. Both the original certification and the focused-practice areas would be available to the public on the Web site.

In the current consumer-centered environ-ment, it is difficult to argue against making more specific and more evidence-based informa-tion available to the public about the specialists who provide their care. Indeed, the Center for Medicare and Medicaid Services Physician Com-pare database, and perhaps other publicly avail-able databases, argues for the most meaningful information (i.e., beyond administrative and claims data) to be part of the information in-cluded.11 Our challenge is to weigh the public interest in light of the traditional meaning of board certification and the benefits of making more specific physician information available against the costs of developing accurate and rig-orous assessment tools. With physician practice evolving over the course of a career in medicine, the focused-practice approach may be better for consumers and may also be a way to respond to physicians’ requests to be evaluated on their ac-tual clinical practice in addition to their desig-nated area of training.12

Table 1. Criteria for Subspecialty Certification and for Recognition of Focused Practice.*

Criteria for Subspecialty Certification

Represents a unique body of knowledge that cannot be fully incorporated into the parent discipline

Has clinical applicability (i.e., the clinical practice is distinctly different from the parent discipline)

Is based on and contributes to the research base of medicine

Offers evidence that the discipline improves patient care

Requires supervision and direct observation in formal training settings, generally lasting at least 12 months

Involves complex technology or specific site-of-care skills

Has positive value for certification in the new discipline that outweighs any negative impact on the practice of general internal medicine or an existing subspecialty

Criteria for Recognition of Focused Practice through Maintenance of Certification (MOC)

Includes large numbers of internists who focus their practice in the discipline, while others may not practice in the focused area at all

Meets an important social need for the discipline and offers evidence that focusing practice in the discipline improves patient care

* Criteria are adapted from New and Emerging Disciplines in Internal Medicine — 2 (NEDIM–2).9




sounding board

Effec ts on Gener alist Pr ac tice

Despite concerns expressed by generalists in re-sponse to new subspecialty designations, it seems unlikely that primary care or generalist disci-plines would be strengthened by a moratorium on creating new subspecialties. Rather, the pay-ment system and organization of medical prac-tice are much more important contextual drivers for advancing primary care.13,14 Generalist disci-plines themselves are developing new knowledge standards as they evolve, including the new hos-pitalist designation geriatric medicine and the probable need for stronger emphasis on systems and information science for generalist physi-cians in a medical home. Medical students get the mistaken message that generalist disciplines are less intellectually exciting, when in fact, it is increasingly difficult to keep up with the breadth of knowledge needed in these fields, including clinical, technical, and managerial skills. Medi-cal group leaders, payers, and hospitals need to be able to identify physicians with these skills as new practice models are formed. Some have called for a specific recognition, perhaps through focused practice in MOC, to recognize the spe-cific systems, health information technology, and team-based competencies required for effective practice in a medical home or accountable care organization.15 Ultimately, primary care will be fostered through payment reform, systems re-form, and stronger standards, not through dimin-ished competition from the subspecialties.

Conclusions

New specialties can benefit both patients and physicians. However, a proliferation of special-ties without adequate justification may simply confuse the public without creating a social good. Use of specified criteria, such as those articu-lated in the ABIM’s NEDIM reports, can lead to rational decision-making that balances the po-tential benefit of recognizing more specific ex-pertise with the detriment of fragmentation of the profession. This approach extends beyond traditional specialization, which requires formal training, to the recognition of new areas of ex-pertise that physicians gain while in practice — that is, focused practice. Although the demands for new subspecialties come from physician

groups themselves, some of their members also complain about the burden of more require-ments. To be meaningful, the criteria for estab-lishing new specialties must be rigorous, but to be workable, they need to be aligned with other measurement and reporting requirements, such as those in pay-for-performance programs, state licensing processes, and hospital privilege issu-ing and credentialing. Certifying boards should continue their work with other accrediting and standard-setting organizations to make these designations as meaningful as possible to patients and physicians alike.


From the American Board of Internal Medicine, Philadelphia (C.K.C., D.B.R.); and the Division of Geriatrics, David Geffen School of Medicine, University of California, Los Angeles (D.B.R.).

1. Levinson K, King TE Jr, Goldman L, Goroll AH, Kessler B. American Board of Internal Medicine maintenance of certifica-tion program. N Engl J Med 2010;362:948-52.2. Flexner A. Medical education in the United States and Cana-da: a report to the Carnegie Foundation for the Advancement of Teaching. Bulletin no. 4. New York: Carnegie Foundation for the Advancement of Teaching, 1910.3. The certification of specialists. N Engl J Med 1936;215:468-9.4. Fitz R. The rise of the practice of internal medicine as a specialty. N Engl J Med 1950;242:569-74.5. Petersdorf RG. The doctors’ dilemma. N Engl J Med 1978; 299:628-34.6. Reitemeier RJ, Benson JA Jr. Two views of “the doctors’ di-lemma”: as the Board sees it. N Engl J Med 1978;299:1308-9.7. Lipner RS, Bylsma WH, Arnold GK, Fortna GS, Tooker J, Cassel CK. Who is maintaining certification in internal medi-cine — and why? A national survey 10 years after initial certifi-cation. Ann Intern Med 2006;144:29-36.8. Brennan TA, Horwitz RI, Duffy FD, Cassel CK, Goode LD, Lipner RS. The role of physician specialty board certification status in the quality movement. JAMA 2004;292:1038-43.9. Final report of the Committee on Recognizing New and Emerging Disciplines in Internal Medicine (NEDIM) — 2. Phila-delphia: American Board of Internal Medicine, 2006. (http://www.abim.org/pdf/nedim-2-report.pdf.)10. Stevens R. American medicine and the public interest: a his-tory of specialization. Berkeley: University of California Press, 1998:xxvi.11. Obama B. Memorandum for the heads of executive depart-ments and agencies: transparency and open government. Wash-ington, DC: The White House, 2010. (http://www.whitehouse.gov/the_press_office/Transparency_and_Open_Government.)12. Drazen JM, Weinstein DF. Considering recertification. N Engl J Med 2010;362:946-7.13. Bodenheimer T, Lo B, Casalino L. Primary care physicians should be coordinators, not gatekeepers. JAMA 1999;281: 2045-9.14. Baron RJ, Cassel CK. 21st-century primary care: new physi-cian roles need new payment models. JAMA 2008;299:1595-7.15. What should outpatient internists do? Weblog: db’s medical rants. May 1, 2010. (http://www.medrants.com/archives/5472.)Copyright © 2011 Massachusetts Medical Society.



clinical implications of basic research



In Utero Hematopoietic Stem-Cell Transplantation — A Match for Mom

Ornella Parolini, Ph.D.

Once researchers recognized that adult stem cells can generate multiple cell types and contribute to tissue homeostasis, it became conceivable to exploit this potential to treat genetic or acquired disorders characterized by tissue degeneration or organ dysfunction. The concept of regenera-tive medicine was thus born, with the general aim of transplanting donor stem cells to replace or repair defective cells of the host.

Unfortunately, in the case of an HLA mis-match between donor and recipient, transplan-tation is hampered by the risks of immunologic recognition and rejection of the graft. However, a recent article by Nijagal and colleagues1 re-vives the discussion of the potential advantages of transplanting stem cells into the fetus early in gestation. Because in utero stem-cell trans-plantation can be carried out when the immune system is immature, it provides the theoretical opportunity to induce fetal tolerance of the for-eign cells and thereby avoid rejection and the need for immunosuppressive therapy. For these reasons, this potential clinical approach is attrac-tive for any disorder that is amenable to stem-cell transplantation and that can be prenatally diagnosed. However, despite successful results of in utero transplantation in animal models, achieved for the most part with hematopoietic stem cells (HSCs), positive outcomes of this procedure in humans have been limited to cases of inherited immunodeficiency diseases.2-4 One of the main hurdles to widespread application and success of in utero stem-cell transplanta-tion is the difficulty in achieving adequate levels of engraftment.

Nijagal and colleagues tested their hypothe-sis that maternal cells trafficking into the fetus impede effective in utero stem-cell transplanta-tion by mounting a sort of “immune protection” of the fetus, in which maternal cells effect the

rejection of cells allogeneic to both mother and fetus. Using mouse models, these researchers test-ed their hypothesis in a series of experiments to evaluate the role of the maternal immune re-sponse in limiting engraftment. First they found that in utero transplantation of fetal HSCs elic-ited an increase in trafficking of maternal T cells to the fetal blood. To test the hypothesis that maternal cells play a pivotal role in the fe-tal engraftment of allogeneic cells, they trans-ferred allogeneic fetal HSCs into fetuses of mothers with experimentally induced B-cell or T-cell deficiency (Fig. 1). Levels of engraftment were significantly higher in fetuses of mothers with T-cell deficiency than in fetuses of wild-type mothers or of mothers with B-cell deficiency. Finally, when in utero stem-cell transplantation was performed with HSCs matched to the moth-er, similar levels of engraftment were observed in fetal recipients of syngeneic and allogeneic fetal grafts.

The evidence put forward is clear and strik-ing, although further research is warranted to confirm these findings, as well as to determine whether maternal T cells are critical for the suc-cess of in utero stem-cell transplantation in other animal models. Differences in placentation be-tween animal species might predict differences in maternal-cell trafficking and thus the extent to which the observed phenomenon in mice is relevant to other animals. However, considering that the same types of hemochorial placentation and fetal–maternal chimerism have been described in humans, the conclusion of Nijagal et al. may be relevant to humans.

The clinical relevance of this study lies in the potential for improving engraftment with the use of cells that are either harvested from or matched to the mother’s cells (Fig. 1). Although collect-ing stem cells from pregnant women presents




clinical implications of basic research

some obstacles, they are not insurmountable, and such an approach may confer advantages that would extend the clinical applicability of in utero stem-cell transplantation beyond trans-plantation of HSCs to include other cell types — such as mesenchymal stromal cells — for the treatment of congenital diseases outside the hematopoietic system.

Finally, the notion that maternal T cells are responsible for the rejection of allogeneic trans-planted HSCs and are therefore critical to the success of engraftment underscores the complex and paradoxical relationship between mother and fetus during pregnancy.

Disclosure forms provided by the author are available with the full text of this article at NEJM.org

From Centro di Ricerca E. Menni, Fondazione Poliambulanza–Istituto Ospedaliero, Brescia, Italy.

1. Nijagal A, Wegorzewska M, Jarvis E, Le T, Tang Q, Macken-zie TC. Maternal T cells limit engraftment after in utero hema-topoietic cell transplantation in mice. J Clin Invest 2011;121:582-92.2. Wengler GS, Lanfranchi A, Frusca T, et al. In-utero trans-plantation of parental CD34 haematopoietic progenitor cells in a patient with X-linked severe combined immunodeficiency (SCIDXI). Lancet 1996;348:1484-7.3. Flake AW, Roncarolo MG, Puck JM, et al. Treatment of X-linked severe combined immunodeficiency by in utero trans-plantation of paternal bone marrow. N Engl J Med 1996;335: 1806-10.4. Peranteau WH, Endo M, Adibe OO, Flake AW. Evidence for an immune barrier after in utero hematopoietic-cell transplan-tation. Blood 2007;109:1331-3.Copyright © 2011 Massachusetts Medical Society.

Wild-typemother

Wild-typemother

Allogeneic fetalhematopoietic stem cells


Mother lacking B cells


Significantimprovement inengraftment levels

Mother lacking T cells

Matchedmaternal cells

Matchedmaternal cells

Maternalcells

E

A

B

C

D

Significantimprovement inengraftment levels

02/24/11



AuthorFig #TitleME

DEArtist

Issue date

COLOR FIGURE

Rev4 Dr. Parolini

03/24/2011

1

PhimisterDaniel Muller

Figure 1. Intrauterine Transplantation and Maternal T Cells.

A recent study by Nijagal and colleagues1 highlights a key role of maternal T cells in limiting engraftment af-ter in utero transplantation, in which allogeneic fetal hematopoietic stem cells (HSCs) were transplanted into the fetuses of wild-type mice (Panel A) and of mice lacking B cells (Panel B) or T cells (Panel C). Only T-cell deficiency resulted in significant improvement in en-graftment levels, as compared with the levels in wild-type mothers. A notable level of engraftment was also observed after transplantation of fetal HSCs that were genetically matched to the mother (Panel D). Taken to-gether, these results suggest that maternal T cells are responsible for failure of engraftment. It is therefore possible that the clinical success of in utero stem-cell transplantation could be improved by transplanting cells harvested from or matched to the mother’s cells (Panel E).





c o r r e s p o n d e n c e

Retraction: A Genomic Strategy to Refine Prognosis in Early-Stage Non–Small-Cell Lung Cancer. N Engl J Med 2006;355:570-80.

to the editor: We would like to retract our ar-ticle, “A Genomic Strategy to Refine Prognosis in Early-Stage Non–Small-Cell Lung Cancer,”1 which was published in the Journal on August 10, 2006. Using a sample set from a study by the American College of Surgeons Oncology Group (ACOSOG) and a collection of samples from a study by the Cancer and Leukemia Group B (CALGB), we have tried and failed to reproduce results supporting the validation of the lung metagene model de-scribed in the article. We deeply regret the effect of this action on the work of other investigators.Anil Potti, M.D.Chapel Hill, NC

Sayan Mukherjee, Ph.D.Duke University Medical Center Durham, NC

Rebecca Petersen, M.D.University of Washington Seattle, WA

Holly K. Dressman, Ph.D.Duke University Medical Center Durham, NC

Andrea Bild, Ph.D.University of Utah Salt Lake City, UT

Jason Koontz, M.D.Duke University Medical Center Durham, NC

Robert Kratzke, M.D.University of Minnesota Minneapolis, MN

Mark A. Watson, M.D., Ph.D.Washington University School of Medicine St. Louis, MO

Michael Kelley, M.D. Geoffrey S. Ginsburg, M.D., Ph.D. Mike West, Ph.D. David H. Harpole, Jr., M.D. Joseph R. Nevins, Ph.D.Duke University Medical Center Durham, NC [email protected]

1. Potti A, Mukherjee S, Petersen R, et al. A genomic strategy to refine prognosis in early-stage non–small-cell lung cancer. N Engl J Med 2006;355:570-80.

This letter (10.1056/NEJMc1101915) was published on March 2, 2011, at NEJM.org.

this week’s letters

1176 Retraction: A Genomic Strategy to Refine Prognosis in Early-Stage Non–Small-Cell Lung Cancer

1177 Apixaban vs. Enoxaparin after Hip Replacement

1178 Oral Rivaroxaban for Symptomatic VenousThromboembolism

1178 Drug-Eluting or Bare-Metal Stents in Large Coronary Arteries

1181 Safety and Effectiveness of a 2009 H1N1 Vaccinein Beijing



correspondence


Apixaban vs. Enoxaparin after Hip ReplacementTo the Editor: Lassen and colleagues (Dec. 23 is-sue)1 should be congratulated for striving to reduce thromboembolic events after hip replacement, but their choice of primary efficacy outcome, which combines clinically important variables (death and symptomatic venous thromboembolism) with venographically proven deep-vein thrombosis, pro-vokes concern. Neither mortality nor the rate of symptomatic thromboembolism differed signifi-cantly between the apixaban and enoxaparin groups; only subclinical deep-vein thrombosis was significantly more common after treatment with enoxaparin as compared with apixaban. This composite outcome measure is therefore mislead-ing. Furthermore, nearly 30% of the participants could not be evaluated for the primary end point. Although this factor was prespecified, it high-lights the limitations of including subclinical deep-vein thrombosis in the primary outcome. In addition, since the drugs were not started simul-taneously, assessment of their relative safety pro-file is compromised; commencing thrombopro-phylaxis before making the surgical incision will induce more bleeding. Finally, data on surgical-site infections and wound hematoma should be re-ported. Complications as a result of thrombo-prophylaxis occur more frequently than major thromboembolic events. Clinically relevant out-come measures and provision of comprehensive safety data are required if perioperative thrombo-prophylaxis is to change.Simon S. Jameson, M.R.C.S.Northumbria Healthcare NHS Trust Ashington, United Kingdom [email protected]

Robert D. Sanders, B.Sc., F.R.C.A.Imperial College London London, United Kingdom

Mike R. Reed, M.D.Northumbria Healthcare NHS Trust Ashington, United Kingdom

No potential conflict of interest relevant to this letter was re-ported.

1. Lassen MR, Gallus A, Raskob GE, Pineo G, Chen D, Ramirez LM. Apixaban versus enoxaparin for thromboprophylaxis after hip replacement. N Engl J Med 2010;363:2487-98.

The Authors Reply: We respectfully disagree with Jameson and colleagues. Subclinical deep-vein

thrombosis, especially proximal deep-vein throm-bosis, is a clinically relevant outcome because it is the source of fatal pulmonary embolism. Most deaths from pulmonary embolism occur as sud-den death and are not preceded by symptoms in the leg or cardiorespiratory system. Reducing the source of these events is clinically relevant. The apixaban regimen reduced major venous throm-boembolism (proximal deep-vein thrombosis or pulmonary embolism) as compared with a stan-dard-care, regulatory-approved regimen of enoxa-parin. Our comparison of these regimens is valid. Systematic literature reviews indicate that this preoperative dose of enoxaparin, administered 12 hours before surgery, does not add to bleed-ing.1,2 The apixaban regimen did not increase bleeding as compared with enoxaparin. This con-clusion is further strengthened when only those bleeding events that occur after the first dose of apixaban — which may be attributable to apixa-ban — are considered. Both surgical-site infec-tions and wound hematomas were uncommon (<2% overall), and occurred with similar fre-quency in both treatment groups (see the Supple-mentary Appendix, available with the full text of this letter at NEJM.org). Thus, our study provides clinically relevant data about the relative risk–benefit profile of the apixaban and enoxaparin regimens.Michael Rud Lassen, M.D.Copenhagen University Hospital Glostrup, Denmark [email protected]

Gary E. Raskob, Ph.D.University of Oklahoma Health Sciences Center Oklahoma City, OK

Alexander Gallus, M.D.Flinders University Adelaide, SA, Australia

Since publication of their article, the authors report no fur-ther potential conflict of interest.

1. Hull RD, Brant RF, Pineo GF, Stein PD, Raskob GE, Valentine KA. Preoperative versus postoperative initiation of low-molecu-lar-weight heparin prophylaxis against venous thromboembo-lism in patients undergoing elective hip replacement. Arch In-tern Med 1999;159:137-41.2. Strebel N, Prins M, Agnelli G, Büller HR. Preoperative or postoperative start of prophylaxis for venous thromboembolism with low-molecular-weight heparin in elective surgery? Arch In-tern Med 2002;162:1451-6.





Oral Rivaroxaban for Symptomatic Venous ThromboembolismTo the Editor: In their article on the use of oral rivaroxaban for the treatment of acute deep-vein thrombosis (DVT), the EINSTEIN investigators (Dec. 23 issue)1 report that rivaroxaban had non-inferior efficacy, as compared with subcutaneous enoxaparin followed by a vitamin K antagonist. In a separate double-blind, randomized superior-ity study comparing extended rivaroxaban thera-py with placebo, rivaroxaban was claimed to be superior. However, the net clinical benefit was not predefined as a separate secondary end point in the protocol and at ClinicalTrials.gov. The com-bination of major and clinically relevant nonmajor bleeding was defined as the principal safety out-come. Although the post hoc definition of net clinical benefit is questionable, it is nevertheless clinically relevant.

The greatest threats of DVT and its treatment are pulmonary embolism and major bleeding, whereas recurrent DVT and clinically relevant bleeding are primarily important from the per-spective of quality of life and cost.2,3 Therefore, we would like to know the clinical benefit of rivaroxaban when it is defined as the rate of re-current venous thromboembolism combined with the principal safety outcome, as well as when it is defined as the rate of recurrent pulmonary embolism and major bleeding.Gijs W. Landman, M.D.Isala Clinics Zwolle, the Netherlands [email protected]

Reinold O.B. Gans, M.D., Ph.D.University Medical Center Groningen, the Netherlands


1. The EINSTEIN Investigators. Oral rivaroxaban for sympto-matic venous thromboembolism. N Engl J Med 2010;363:2499-510.

2. Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin therapy on quality of life: evidence from a randomized trial. Arch Intern Med 1991;151:1944-9. [Erratum, Arch Intern Med 1992;152:825.]3. Gould MK, Dembitzer AD, Sanders GD, Garber AM. Low-molecular-weight heparins compared with unfractionated hepa-rin for treatment of acute deep venous thrombosis: a cost-effec-tiveness analysis. Ann Intern Med 1999;130:789-99.

The authors reply: Although the net clinical benefit was not reported as an a priori outcome in the study summary at ClinicalTrials.gov, it was included in the study’s statistical analysis plan, which was approved before enrollment of the first patient and was made available to health author-ities. This composite outcome gives equal weight to the primary efficacy outcomes and major bleed-ing, because of their similar overall clinical im-portance in terms of need for rehospitalization, long-term consequences (including the post-thrombotic syndrome, thromboembolic pulmo-nary hypertension, and damage to various organ systems), and case fatality rates. This overall risk–benefit outcome was in favor of rivaroxaban, with a hazard ratio of 0.67 (95% confidence in-terval, 0.47 to 0.95), and was consistent for the various subgroups.

Harry R. Buller, M.D.Academic Medical Center Amsterdam, the Netherlands [email protected]

Anthonie W.A. Lensing, M.D.Bayer HealthCare Pharmaceuticals Wuppertal, Germany

Martin H. Prins, M.D.Maastricht University Maastricht, the Netherlands


Drug-Eluting or Bare-Metal Stents in Large Coronary Arteries

To the Editor: Kaiser et al. (Dec. 9 issue)1 shed important light on the issue of the safety and ef-ficacy of drug-eluting stents in large coronary arteries. The characteristics of the patients and their lesions were very well matched between the groups, as disclosed in detail in the article. How-

ever, it is not clear whether there was any sig-nificant between-group difference in the mean reference-vessel diameter or post-procedural lu-minal diameter, since intravascular ultrasonog-raphy (IVUS) and quantitative angiography were not used in the study. These measurements are



correspondence


important predictors of restenosis, both with the use of bare-metal stents and drug-eluting stents.2 On the assumption that the stents were expand-ed as much as possible, as described in the study protocol,3 the diameters of the stents could give some idea about these measurements. It would be interesting to know whether there was any significant difference in the diameters of stents between the groups.Cemil Izgi, M.D.Gaziosmanpasa Hospital Istanbul, Turkey [email protected]


1. Kaiser C, Galatius S, Erne P, et al. Drug-eluting versus bare-metal stents in large coronary arteries. N Engl J Med 2010;363: 2310-9.2. Dangas GD, Claessen BE, Caixeta A, Sanidas EA, Mintz GS, Mehran R. In-stent restenosis in the drug-eluting stent era. J Am Coll Cardiol 2010;56:1897-907.3. Pfisterer M, Bertel O, Bonetti PO, et al. Drug-eluting or bare-metal stents for large coronary vessel stenting? The BASKET-PROVE (PROspective Validation Examination) trial: study protocol and de-sign. Am Heart J 2008;155:609-14.

To the Editor: The main conclusion of Kaiser et al. was that there was no significant difference in the occurrence of the primary end point (death from cardiac causes or nonfatal myocardial in-farction) with drug-eluting stents, as compared with bare-metal stents. However, the comparison of sirolimus-eluting stents with bare-metal stents showed a hazard ratio of 0.54 for the primary end point in favor of the sirolimus-eluting stent. The 95% confidence interval (0.31 to 0.93) sug-gests that the result is significantly different, but the P value is 0.13. I presume that this was the result of adjustment of data for multiple com-parisons. I wonder whether this statistical adjust-ment should have been performed, since there was only one primary end point and the pre-defined sample size was calculated for two sepa-rate comparisons.1 Could the difference be statis-tically significant in favor of the sirolimus stent? Also, given the seemingly beneficial effect of siro-limus, I would also be interested in a post hoc analysis combining both drug-eluting stents and bare-metal stents.Marnix W. van Bemmel, M.D.Deventerziekenhuis Zwolle, the Netherlands


1. Pfisterer M, Brunner-La Rocca H-P, Rickenbacher P, et al. Long-term benefit-risk balance of drug-eluting stents in daily practice: does stent diameter matter? Three-year follow-up of BASKET. Eur Heart J 2009;30:16-24.

To the Editor: The early studies of drug-eluting stents versus bare-metal stents were conducted in a blinded fashion.1 In these studies, no differences in rates of subacute stent thrombosis, myocardial infarction, or death were shown, findings that were confirmed in a meta-analysis2 that was per-formed to address the safety concerns raised by the Basel Stent Cost-Effectiveness Trial (BASKET).3 The prospective validation study by Kaiser et al. (BASKET–PROVE) further evaluated the risk. Un-expectedly, the early risk was higher with the use of bare-metal stents than with drug-eluting stents. In this study, the operators were not unaware of treatment assignment. It was found that the rate of early (<14 days) death and myocardial infarc-tion was increased with the use of bare-metal stents, suggesting that factors associated with implantation could be involved. It is possible that operators during the trial, concerned about safety with the use of drug-eluting stents, were perform-ing better post-dilatation procedures, used higher-pressure dilatations, or used higher final balloon diameters in the groups receiving drug-eluting stents. Can the authors give figures for the pro-portion of patients having high-pressure (>20 atm) post-dilatation procedures, the proportion of pa-tients who underwent IVUS, the highest implan-tation pressure used, and the nominal diameter of the biggest balloon used at implantation?Evald H. Christiansen, M.D., Ph.D.Aarhus University Hospital, Skejby Aarhus, Denmark [email protected]


1. Moses JW, Leon MB, Popma JJ, et al. Sirolimus-eluting stents versus standard stents in patients with stenosis in a native coro-nary artery. N Engl J Med 2003;349:1315-23.2. Stone GW, Moses JW, Ellis SG, et al. Safety and efficacy of sirolimus- and paclitaxel-eluting coronary stents. N Engl J Med 2007;356:998-1008.3. Brunner-La Rocca HP, Kaiser C, Pfisterer M. Targeted stent use in clinical practice based on evidence from the Basel Stent Cost Effectiveness Trial (BASKET). Eur Heart J 2007;28:719-25.

To the Editor: Kaiser et al. define a large coro-nary artery as one requiring a stent of 3.0 mm or more in diameter. However, patients requiring





stents larger than 4.0 mm were excluded because the Cypher stent was not available in that size. Thus, their results represent mainly the outcome of a population with “non-small” coronary arteries who were treated with either drug-eluting or bare-metal stents. Moreover, the mean stent diameter in each group is not reported. It would be inter-esting to know the outcome of patients treated with a bare-metal stent of 3.5 mm diameter. The results of this study should not be extrapolated to patients with larger coronary arteries (i.e., with a diameter of >3.5 mm) in which a bare-metal stent might have equivalent efficacy to a drug-eluting stent.1,2

Adel Aminian, M.D. Jacques Lalmand, M.D.Centre Hospitalier Universitaire de Charleroi Charleroi, Belgium [email protected]


1. Steinberg DH, Mishra S, Javaid A, et al. Comparison of effec-tiveness of bare metal stents versus drug-eluting stents in large (> or = 3.5 mm) coronary arteries. Am J Cardiol 2007;99:599-602.2. Quizhpe AR, Feres F, de Ribamar Costa J Jr, et al. Drug-eluting stents vs bare metal stents for the treatment of large coronary vessels. Am Heart J 2007;154:373-8.

The authors reply: The correspondents’ main questions relate to final stent diameters achieved in the different groups and the maximal dilata-tion pressures applied. Stents with a nominal diam-eter of 3.0 mm were used in 72% of patients, with stents of 3.5 mm used in 23% of patients (2% had smaller stents, and 3% had larger stents for bailout indications). Stents with a diameter of 4.0 mm or more were not used because of the unavailability of such sirolimus-eluting stents. However, in the initial BASKET population,1 only 23 of 988 consecutive patients (2.3%) needed stents with a diameter of 4.0 mm or more for native-vessel stenting, indicating that this patient group is clinically of little relevance. The mean (±SD) maximal stent diameter per patient was 3.5±0.3 mm in patients receiving sirolimus-elut-ing stents, 3.5±0.4 mm in those receiving evero-limus-eluting stents, and 3.5±0.4 mm in those receiving bare-metal stents; the mean maximal balloon pressures were 15.5±3.5 bar, 14.8±3.4 bar, and 14.5±3.2 bar, respectively.

There were no significant differences in maxi-mal stent sizes or maximal pressures applied be-

tween everolimus-eluting and bare-metal stents, but sirolimus-eluting stents were somewhat small-er than everolimus-eluting and bare-metal stents (−0.05 mm and −0.06 mm, respectively; P<0.05) despite having higher maximal dilatation pres-sures (P<0.001). This difference was most likely due to the stainless-steel platform of this stent, as compared with a cobalt–chromium platform in the other two stents. These small differences were clinically negligible, as shown by the virtu-ally identical 2-year outcomes of the two drug-eluting stents. Thus, the early risk of death or myocardial infarction in patients receiving bare-metal stents that was observed in our study and in several previous studies1-3 cannot be attributed to differences in stent size or maximal pressures applied.

We have not yet reported an analysis stratified by final stent size. Although this analysis was planned a priori, it was a secondary aim only and could not be included in the report of the pri-mary end-point results. For the primary analy-ses, adjustments for multiple comparisons were necessary because the control group (i.e., patients treated with bare-metal stents) was used as com-parator for both drug-eluting stents. Finally, it is possible that the newest generations of bare-metal stents may perform even better than the standard stents used, particularly in patients with large-vessel stenting, but the same is true for newer generations of drug-eluting stents. However, not all newer generations of stents may show im-proved outcomes, as we found in our study.

Christoph Kaiser, M.D. Matthias Pfisterer, M.D.University Hospital Basel Basel, Switzerland [email protected]


1. Kaiser C, Brunner-La Rocca HP, Buser PT, et al. Incremental cost-effectiveness of drug-eluting stents compared with a third-generation bare-metal stent in a real-world setting: randomised Basel Stent Kosten Effektivitäts Trial (BASKET). Lancet 2005; 366:921-9.2. Park DW, Yun SC, Lee SW, et al. Stent thrombosis, clinical events, and influence of prolonged clopidogrel use after place-ment of drug-eluting stent data from an observational cohort study of drug-eluting versus bare-metal stents. JACC Cardiovasc Interv 2008;1:494-503.3. Lagerqvist B, James SK, Stenestrand U, Lindbäck Nilsson T, Wallentin L. Long-term outcomes with drug-eluting stents versus bare-metal stents in Sweden. N Engl J Med 2007;356:1009-19.



correspondence


Safety and Effectiveness of a 2009 H1N1 Vaccine in BeijingTo the Editor: The study by Wu et al. (Dec. 16 issue)1 has serious methodologic issues that lim-it the interpretation of their estimate of the effec-tiveness of a 2009 pandemic influenza A (H1N1) vaccine. First, the cumulative incidence of influ-enza was 2.5 times as high in the unvaccinated cohort as in the vaccinated cohort even before a protective effect from the vaccine could have been achieved (Fig. 2 of the article). Because of bias or confounding, these two cohorts did not have the same risk of influenza. In addition, many stu-dents may have been infected with the pandemic virus before the study began. Vaccine trials in China that preceded this study reported that 18 to 25% of children who were 12 to 18 years of age were positive for the pandemic H1N1 antibody.2 At one Chinese school, 32% of the students were infected.3 A subsequent serologic survey conduct-ed in Beijing showed that 19.4% of children and young people between 6 and 17 years of age were seropositive for the pandemic virus.4 Yet in the study by Wu et al., less than 0.30% of the unvac-cinated cohort and 0.01% of the vaccinated cohort were infected. Numerous cases must have been missed in both cohorts before the study began and while it was being conducted.Nicholas S. Kelley, M.S.P.H. Michael T. Osterholm, Ph.D., M.P.H.University of Minnesota Minneapolis, MN [email protected]

Edward A. Belongia, M.D.Marshfield Clinic Research Foundation Marshfield, WI


1. Wu J, Xu F, Lu L, et al. Safety and effectiveness of a 2009 H1N1 vaccine in Beijing. N Engl J Med 2010;363:2416-23.2. Liang XF, Wang HQ, Wang JZ, et al. Safety and immunoge-nicity of 2009 pandemic inf luenza A H1N1 vaccines in China: a multicentre, double-blind, randomised, placebo-controlled trial. Lancet 2010;375:56-66. [Erratum, Lancet 2010;375:1694.]3. Li T, Liu Y, Di B, et al. Epidemiological investigation of an outbreak of pandemic influenza A (H1N1) 2009 in a boarding school: serological analysis of 1570 cases. J Clin Virol 2011;50: 235-9.4. Yang P, Huang F, Shi W, et al. A survey on serological epide-miology of influenza A (H1N1) 3009 in Beijing. Zhonghua Liu Xing Bing Xue Za Zhi 2010;31:485-8. (In Chinese.)

To the Editor: Wu et al. evaluate the effective-ness of a 2009 H1N1 vaccine. In this study, the

vaccinated students had all been participants in China’s 60th National Day Celebration, whereas most of the unvaccinated students had not been participants. Generally, the students in the cele-bration were healthier and better educated about the prevention of infection, and most were living in urban districts — all of which might lead to an overestimation of the effectiveness of the vaccine. Although aware of these differences between the two groups, the authors took no measures to elim-inate them. In fact, a total of about 80,000 primary-school or middle-school students participated in the celebration, including 25,037 vaccinated stu-dents and 55,000 unvaccinated students. I wonder why these 55,000 unvaccinated students were not selected as the control group, which might have eliminated the interference of the above-men-tioned factors.Xiang-Dong Mu, M.D.Peking University First Hospital Beijing, China [email protected]


The Authors Reply: In our study, both the un-vaccinated cohort and the vaccinated cohort came from 245 schools. The vaccinated cohort consist-ed of students who participated in the 60th Na-tional Day Celebration and were vaccinated during a 5-day period in September 2009. The unvaccinat-ed cohort included students who did and those who did not participate in the celebration. The stu-dents in the vaccinated cohort attended the same schools at the same times and had backgrounds similar to those in the unvaccinated cohort. Hence, both cohorts should have had similar exposure to influenza viruses.

The incidence data for our study were collect-ed from October 9 through November 15, 2009. Yang et al. performed their study from Novem-ber 27 through December 23, 2009, and their subjects were patients from outpatient sections of six hospitals (two children’s hospitals and four general hospitals) in Beijing. Since Yang et al. used convenience sampling and the study sub-jects did not represent the entire student popula-tion, there is no conflict in this regard between our report and theirs. In addition, there is some




corrections

debate about the cross-reactive antibody response between H1N1 and pandemic H1N1. The data from Chi et al. indicate that natural infection with the 2006 and 2008 seasonal H1N1 viruses in Taiwan may have induced a cross-reactive anti-body response to the 2009 pandemic H1N1 virus.1 We detected a similar phenomenon. We tested 301 serum samples from volunteers after H5N1 vaccination in Beijing in 2007 and observed a cross-reaction in 2.9% of the group between 3 and 11 years of age and in 11.3% of the group be-tween 12 and 17 years of age.2 There was also a report from the United States of a possible cross-reactive antibody response to the H1N1 and pan-demic H1N1 viruses.3 Our experience with sea-sonal flu vaccine and pandemic H1N1 flu vaccine also indicated the existence of this phenomenon. Thus, measurement of serum antibodies probably did not fully reflect the actual infection status of the population.4

We agree that some cases may have been missed in our study, such as silent infection and mild cases. However, this phenomenon should have af-fected both the vaccinated and unvaccinated sub-jects.

Jiang Wu, M.D. Min Lu, Ph.D. Ying Deng, M.D.Beijing Centers for Disease Control and Prevention Beijing, China [email protected]


1. Chi CY, Liu CC, Lin CC, et al. Preexisting antibody response against 2009 pandemic influenza H1N1 viruses in the Taiwan-ese population. Clin Vaccine Immunol 2010;17:1958-62.2. Clinical trial interim report of pandemic influenza A/H1N1 vaccine: serial number CR-PanFlu-4001, version 2009;0:32.3. Hancock K, Veguilla V, Lu X, et al. Cross-reactive antibody responses to the 2009 pandemic H1N1 influenza virus. N Engl J Med 2009;361:1945-52.4. Wu J, Zhong X, Li CK, et al. Optimal vaccination strategies for 2009 pandemic H1N1 and seasonal influenza vaccines in humans. Vaccine 2011;29:1009-16.Correspondence Copyright © 2011 Massachusetts Medical Society.

INSTRUCTIONS FOR LETTERS TO THE EDITOR

Letters to the Editor are considered for publication, subject to editing and abridgment, provided they do not contain material that has been submitted or published elsewhere. Please note the following:• Letters in reference to a Journal article must not exceed 175

words (excluding references) and must be received within 3 weeks after publication of the article.

• Letters not related to a Journal article must not exceed 400 words.

• A letter can have no more than five references and one figure or table.

• A letter can be signed by no more than three authors.• Financial associations or other possible conflicts of interest

must be disclosed. Disclosures will be published with the letters. (For authors of Journal articles who are responding to letters, we will only publish new relevant relationships that have developed since publication of the article.)

• Include your full mailing address, telephone number, fax number, and e-mail address with your letter.

• All letters must be submitted at authors.NEJM.org.Letters that do not adhere to these instructions will not be

considered. We will notify you when we have made a decision about possible publication. Letters regarding a recent Journal article may be shared with the authors of that article. We are unable to provide prepublication proofs. Submission of a letter constitutes permission for the Massachusetts Medical Society, its licensees, and its assignees to use it in the Journal’s various print and electronic publications and in collections, revisions, and any other form or medium.

corrections

Glycemic Control in the ICU (December 23, 2010;363:2540-6). In the Studies of Intensive Insulin Therapy subsection of Man-agement (page 2541), the first sentence in the third paragraph should have cited reference 8, rather than 9, and the second sentence in the final paragraph (page 2543) should have cited references 6 and 8, rather than 7 and 9. In the Glucose Variabil-ity subsection of Areas of Uncertainty (page 2543), the second sentence in the first paragraph should have ended, “. . . as is hypoglycemia that develops spontaneously,” rather than “. . . regardless of the glucose level.” Finally, in the Hypoglyce-mia subsection of Areas of Uncertainty (page 2544), the second sentence of the first paragraph should have cited reference 27, rather than 30. The article is correct at NEJM.org.

Emergence of a Multidrug-Resistant Pandemic Inf luenza A (H1N1) Virus (September 30, 2010;363:1381-2). In the third paragraph (page 1382), the date mentioned should have been January 4 rather than February 9, and in Figure 1 (page 1381), the dates on the x axis should have included Jan. 4, 2010, rath-er than Feb. 9, 2010. The article is correct at NEJM.org.

Case 4-2009: A 39-Year-Old Pregnant Woman with Fever after a Trip to Africa (January 29, 2009;360:508-16). In the third para-graph (page 508), the fifth sentence should have begun, “She took mefloquine weekly,” rather than “She took mefloquine daily.” The article is correct at NEJM.org.

THE JOURNAL’S WEB AND E-MAIL ADDRESSES:

For letters to the Editor: authors.NEJM.org

For information about the status of a submitted manuscript: authors.NEJM.org

To submit a meeting notice: [email protected]

The Journal’s Web pages: NEJM.org



images in clinical medicine


n engl j med 364;12 nejm.org march 24, 2011 e23

A 36-year-old man presented to an Ethiopian clinic with a 20-yearhistory of skin nodules, pain, and edema involving his legs and feet. He was otherwise healthy and worked as a farmer. Circulating filarial antigen tests

for the presence of Wuchereria bancrofti were negative; he was not tested for other types of filaria. This clinical presentation prompted a diagnosis of probable podo-coniosis (also known as nonfilarial elephantiasis or mossy foot). This locally en-demic, noninfectious condition is caused by the long-term exposure of susceptible persons to irritant volcanic soil. Colloid particles are thought to be absorbed through the skin and taken up by macrophages, leading to lymphatic fibrosis and elephantiasis. Affected persons are typically barefoot agricultural workers in the highland tropics. Social stigma associated with this condition is widespread; pa-tients are banned from schools, churches, and marriage. Economic productivity is often impaired. Podoconiosis is preventable with fastidious shoe wearing and foot hygiene. Treatment is limited to compression bandaging and elevation. The patient was instructed to wear shoes, but additional nodules continued to develop on un-covered areas of his sandaled feet.Copyright © 2011 Massachusetts Medical Society.

Podoconiosis

Whitney Lapolla, M.D.Stephen K. Tyring, M.D., Ph.D.

Center for Clinical StudiesWebster, [email protected]



the new england journal of medicine - march 24, 2011 - vol. 364 no. 12 364 12

Documents