White blood cell count and mortality in patients with acutepulmonary embolism

Carmen Venetz,1 Jos�e Labarere,2 David Jim�enez,3 and Drahomir Aujesky1*

Although associated with adverse outcomes in other cardiovascular diseases, the prognostic value of anelevated white blood cell (WBC) count, a marker of inflammation and hypercoagulability, is uncertain inpatients with pulmonary embolism (PE). We therefore sought to assess the prognostic impact of the WBC ina large, state-wide retrospective cohort of patients with PE. We evaluated 14,228 patient discharges with aprimary diagnosis of PE from 186 hospitals in Pennsylvania. We used random-intercept logistic regressionto assess the independent association between WBC count levels at the time of presentation and mortalityand hospital readmission within 30 days, adjusting for patient and hospital characteristics. Patients with anadmission WBC count <5.0, 5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6 3 109/L had a cumulative 30-day mortality of10.9%, 6.2%, 5.4%, 8.3%, and 16.3% (P < 0.001), and a readmission rate of 17.6%, 11.9%, 10.9%, 11.5%, and15.0%, respectively (P < 0.001). Compared with patients with a WBC count 7.9–9.8 3 109/L, adjusted odds of30-day mortality were significantly greater for patients with a WBC count <5.0 3 109/L (odds ratio [OR] 1.52,95% confidence interval [CI] 1.14–2.03), 9.9–12.6 3 109/L (OR 1.55, 95% CI 1.26–1.91), or >12.6 3 109/L (OR2.22, 95% CI 1.83–2.69), respectively. The adjusted odds of readmission were also significantly increased forpatients with a WBC count <5.0 3 109/L (OR 1.34, 95% CI 1.07–1.68) or >12.6 3 109/L (OR 1.29, 95% CI 1.10–1.51). In patients presenting with PE, WBC count is an independent predictor of short-term mortality andhospital readmission. Am. J. Hematol. 88:677–681, 2013. VC 2013 Wiley Periodicals, Inc.

IntroductionEvidence suggests that the white blood cell (WBC) count

is associated with adverse outcomes in patients with cardio-vascular diseases, such as acute coronary syndrome, heartfailure, and ischemic stroke [1–7]. In acute coronary syn-drome, WBCs cause myocardial injury via vascular plugging,direct injury to myocytes and the coronary endothelium, andpro-inflammatory cytokines [3]. Neutrophilic infiltration andthe production of reactive oxygen species has been postu-lated as a mechanism for left ventricular failure [8].

Acute pulmonary embolism (PE) is a major health prob-lem; in 2010, 190,000 patients were discharged with a pri-mary diagnosis of PE from US hospitals [9], with anestimated average 30-day mortality of 9% [10]. Early deathafter PE is strongly associated with right ventricular dys-function [11]. Animal and autopsy studies demonstratedthat neutrophils are not only involved in the development ofvenous thrombosis but that an influx of neutrophils andother WBCs may contribute to right ventricular dysfunctionfollowing PE [12–17]. Moreover, the WBC count also corre-lates with levels of fibrinogen, factor VII, and factor VIII andthus, may be a marker for hypercoagulability [18]. Basedon a previously observed association between an elevatedWBC count and short-term mortality in patients with PE[19], we hypothesized that an elevated WBC at baseline, amarker of inflammation and hypercoagulability, may indicatea worse prognosis in patients with this illness.

We therefore examined the association between theadmission WBC count and 30-day mortality and hospitalreadmission in a large, statewide sample of patients withacute PE. If found, such association may be useful to risk-stratifying patients with acute PE, given that WBC count isa low-cost, routinely available laboratory parameter.

Methods

Patient identification and eligibility

We identified patients with PE discharged from 186 non-governmental (i.e., non-Veterans Administration) acute care hospitalsin Pennsylvania, USA (01/01/2000 to 11/30/2002) using the

Pennsylvania Health Care Cost Containment Council (PHC4) database[10]. This database contains information on demographic characteris-tics, insurance status, International Classification of Diseases, NinthRevision, Clinical Modification (ICD-9-CM) diagnosis and procedurecodes, hospital region and number of beds for all patients.

We included inpatients aged �18 years who were discharged with aprimary diagnosis of PE based on the following ICD-9-CM codes: 415.1,415.11, 415.19, and 673.20–24 [10]. To ensure that we identified themost severely ill patients with PE as the primary reason for hospitaliza-tion, we also included inpatients with a secondary diagnosis code forPE,and one of the following primary codes that may represent complica-tions or treatments of this condition: respiratory failure (518.81), cardio-genic shock (785.51), cardiac arrest (427.5), secondary pulmonaryhypertension (416.8), syncope (780.2), thrombolysis (99.10), and intuba-tion or mechanical ventilation (96.04, 96.05, 96.70–96.72).

We excluded all other patients who had a secondary ICD-9-CMcode for PE or those who were transferred from another health carefacility, because such patients are more likely to have PE as a compli-cation of hospitalization and we did not know whether PE was diag-nosed and treated before the patient was transferred. We excludedfollow-up records for patients who were subsequently transferred to

Additional Supporting Information may be found in the online version of thisarticle.1Division of General Internal Medicine, Bern University Hospital, Bern, Swit-zerland; 2Techniques de l’Ing�eni�erie M�edicale et de la Compl�exit�e, UMR5525 Centre National de la Recherche Scientifique, Universit�e JosephFourier-Grenoble 1, Grenoble, France; 3Respiratory Department, Ram�on yCajal Hospital, IRYCIS, Madrid, Spain

Conflict of interest: Nothing to report

*Correspondence to: Drahomir Aujesky, Division of General Internal Medi-cine, Bern University Hospital, Inselspital, 3010 Bern, Switzerland. E-mail:drahomir.aujesky@insel.ch

Contract grant sponsor: National Heart, Lung, and Blood Institute; Contractgrant number: 1-R21-HL075521-01A1; Contract grant sponsor: SwissNational Science Foundation; Contract grant number: 33CSCO-122659/139470.

Received for publication 24 March 2013; Revised 3 May 2013; Accepted 7May 2013

Am. J. Hematol. 88:677–681, 2013.

Published online 14 May 2013 in Wiley Online Library(wileyonlinelibrary.com).DOI: 10.1002/ajh.23484

VC 2013 Wiley Periodicals, Inc.

American Journal of Hematology http://wileyonlinelibrary.com/cgi-bin/jhome/35105677

Research Article

other hospitals. We also excluded patients without the identifiersrequired for linkage to the necessary clinical data and those for whommortality information was not available. For this analysis, we alsoexcluded patients without a documented WBC count on admission.

Patient and hospital characteristics

Patient demographic characteristics (age, gender, race, and insur-ance status) were abstracted from the PHC4 Database [10]. Baselineclinical variables and laboratory parameters (including WBC counts)were obtained by linking eligible patients to the Atlas Database (Med-iQual, Marlborough, MA). This Database includes clinical findings atpresentation for all inpatients treated at non-governmental acute carehospitals in Pennsylvania [10]. The PHC4 and Atlas databases werematched by PHC4 staff using unique patient identifiers (patient date ofbirth, gender, and social security number); we had no access to perso-nal patient identifiers [10].

We quantified severity of illness using the Pulmonary EmbolismSeverity Index (PESI), a prognostic model for patients with PE thatwas developed and validated using these clinical data from the PHC4and Atlas databases [10]. On the basis of the PESI, each patient isclassified into one of five severity classes (I–V), with 30-day mortalityranging from 1.1% to 24.5% [10]. To ascertain whether patientsreceived thrombolytic therapy, we used ICD-9-CM procedure codes(99.10) from the PHC4 and Atlas databases.

We abstracted the hospital region within Pennsylvania, number ofbeds per hospital site, and annual number of PE admissions for eachsite from the PHC4 database. We defined hospital teaching statusbased on data from the Council of Teaching Hospitals of the Associa-tion of American Medical Colleges. Because 76% of teaching hospitals,only 12% of nonteaching hospitals, had at least 350 hospital beds, wecreated a composite hospital-level variable for our statistical modelingbased on teaching status and size (i.e., small nonteaching hospitalswith fewer than 350 beds, large nonteaching hospitals with at least 350beds, and teaching hospitals).

Admission white blood cell count and outcomes

Because the distribution of mortality by WBC count was J-shaped(figure in the Supporting Information), we categorized WBC counts intofive categories by subdividing patients with WBC counts �5.0 3 109/Linto quartiles (5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6 3 109/L), whereaswe assigned patients with WBC counts <5.0 3 109/L to a separate cat-egory labeled subquartile [2]. This categorization ensured that the sec-ond quartile, which had the lowest mortality and readmission rate,would serve as the appropriate reference group for comparing theodds of death and readmission.

Our primary study outcome was all-cause mortality within 30 daysafter admission by linking patients to the National Death Index usingunique patient identifiers, including social security number, name, dateof birth, and sex. The National Death Index has a sensitivity and speci-ficity of 97% for identifying mortality [20–22]. To ascertain our second-ary outcome, hospital readmission for any reason to any acute carehospital in Pennsylvania within 30 days of presentation, we used thePHC4 database.

Statistical analyses

We compared baseline patient characteristics across the five catego-ries of WBC counts using chi-square tests for categorical variables andKruskal–Wallis rank tests for continuous variables. We used survivalanalyses and the log-rank test to compare the cumulative 30-day mortal-ity and hospital readmission rates by WBC level. Surviving patients werecensored at 30 days. To adjust for severity of illness, we also stratifiedour comparisons of mortality by WBC count categories and the five PESIseverity risk classes. Missing values were assumed to be normal, a strat-egy that was successfully used to derive the PESI [10].

We used multivariable logistic regression to examine the independ-ent association between admission WBC count and mortality, afteradjusting for patient demographics (age, gender, race, and insurancetype), comorbid diseases (history of cancer, chronic lung disease, andheart failure), and physical examining findings (pulse �110/minute, sys-tolic blood pressure <100 mmHg, respiratory rate �30/minute, alteredmental status, temperature <36�C, and arterial oxygen saturation<90%) comprising the PESI, laboratory values (hemoglobin, sodium,creatinine, glucose, and troponin), thrombolysis, and hospital character-istics (region within Pennsylvania, annual volume of PE, and size andteaching status). To account for patient clustering within hospital, weused random-intercept logistic regression with the two levels defined bypatient and hospital site.

We examined whether the addition of WBC count to PESI improved30-day mortality risk prediction using the increase in the area under

the receiver operating characteristic (ROC) curve and the net reclassifi-cation improvement (NRI) index. The NRI index quantifies the magni-tude of improvement in risk prediction offered by the addition of theWBC count to the PESI based on upward and downward movementsin the predicted probabilities of 30-day mortality [23]. We expected thatthe addition of WBC count to the PESI would increase the model-based predicted probabilities for patients who died within 30 days ofadmission and decrease the model-based predicted probabilities forpatients who were alive.

We used the same logistic regression model to examine the associ-ation between WBC count and readmission within 30 days in patientsdischarged alive. Patients who were still hospitalized 30 days afteradmission and those without a documented readmission status wereexcluded from this analysis. To account for heterogeneity in length ofhospital stay, we also calculated the number of hospital-free days fromdischarge to 30 days of follow-up or readmission, which ever occurredfirst, and compared the readmission incidence rates across categoriesof WBC count. All analyses were performed using Stata 11.0.

ResultsOf the 17,733 patient discharges who met our inclusion

criteria, we excluded 323 patients with only a secondarycode indicative of PE (1.8%), 767 patient transfers fromanother hospital (4.3%), 265 subsequent transfers to anotherhospital (1.5%), 777 without a match to key clinical findings(4.4%), and 70 patients without a linkage to the NationalDeath Index (0.4%), leaving a sample of 15,531 patient dis-charges with PE (Fig. 1). Of these, we excluded 1303 (8.4%)discharges with an undocumented WBC count on admission.Our final study sample comprised 14,228 patient dischargeswith a diagnosis of PE from 186 Pennsylvania hospitals(Fig. 1). Overall, 822 (5.8%) had a WBC count below5.0 3 109/L and 3278 (23.0%) had a WBC count above12.6 3 109/L. Compared with the 14,228 enrolled patients,the 1303 patients excluded because of an undocumentedWBC count were significantly younger (median age, 66 vs.67 years; P 5 0.02) and less likely to be men (37.1 vs.40.4%; P 5 0.02) and to have heart failure (13.0 vs.16.1%;P 5 0.003) but were more likely to have a history of cancer(22.6 vs. 19.3%; P 5 0.004). They were also less likely tohave a pulse �110/minute (12.8 vs. 18.1%; P<0.001), asystolic blood pressure <100 mmHg (7.8 vs. 10.7%;P 5 0.001), a respiratory rate �30 breaths/minute (8.7 vs.15.1%; P< 0.001), and an arterial oxygen saturation <90%(3.6 vs. 8.3%; P< 0.001) on admission.

Comparison of baseline patient characteristics bywhite blood cell count level

Patients with a WBC count >12.6 3 109/L were older,more likely to have chronic lung disease and heart failure,and more likely to have clinical and biological signs of dis-ease severity (tachycardia, hypotension, tachypnea, alteredmental status, hypothermia, hypoxemia, hyponatremia, andelevated creatinine, glucose, and troponin levels) thanpatients with a WBC count �12.6 3 109/L (table in the Sup-porting Information). There was a higher proportion ofpatients in PESI risk class V among patients with a WBCcount >12.6 3 109/L. Patients with a WBC count <5.0 3 109/L were more likely to have a history of cancer and anemiaand to be in PESI risk class IV than patients with a WBC�5.0 3 109/L.

Association of white blood cell count and 30-daymortality

Overall 1300 of the 14,228 patients (9.1%) died within 30days of admission. Patients with an admission WBC count<5.0, 5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6 3 109/L had acumulative probability of 30-day mortality of 10.9%, 6.2%,5.4%, 8.3%, and 16.3%, respectively (P< 0.001) (Fig. 2).When stratified by each of the five PESI risk classes atpresentation, mortality was highest among patients with a

research article

678 American Journal of Hematology

WBC count >12.6 3 109/L, followed by patients with aWBC count <5.0 3 109/L (Table I).

After adjustment, the odds of 30-day mortality remainedsignificantly increased for patients with a WBC count<5.0 3 109/L (odds ratio [OR] 1.52, 95% confidence inter-val [CI] 1.14–2.03), 9.9–12.6 3 109/L (OR 1.55, 95% CI1.26–1.91), and >12.6 3 109/L (OR 2.22, 95% CI 1.83–2.69) (Table II). No evidence of first-order interaction wasfound between PESI risk classes and WBC count catego-ries with regard to 30-day mortality (P 5 0.98).

The addition of WBC count to the PESI was associatedwith improved risk prediction for 659 (50.7%) of 1300patients who died within 30 days of admission and withworse risk prediction for 641 (49.3%) patients. Conversely,risk prediction improved for 8068 patients (62.4%) anddeteriorated for 4860 patients (37.6%) who were alive by30 days of follow-up. The corresponding NRI index was0.26 (95% CI 0.21–0.32), indicating improvement in riskprediction. This finding was consistent with a modest butstatistically significant increase in the area under the ROCcurve (0.78 [95% CI 0.77–0.79] vs. 0.76 [95% CI 0.75–0.77] for the PESI with and without WBC count, P< 0.001).

Association of white blood cell count and 30-dayreadmission

The 30-day readmission rate was estimated in 13,261patients, after the exclusion of 840 patients who died in thehospital, 93 who were still hospitalized 30 days after admis-sion, and 34 with unknown readmission status. Overall,1647 (12.4%) patients were readmitted within 30 days.Patients with an admission WBC count <5.0, 5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6 3 109/L had a cumulative proba-bility of 30-day readmission of 17.6%, 11.9%, 10.9%,11.5%, and 15.0%, respectively (P<0.001). After adjust-ment, patients with a WBC count <5.0 3 109/L (OR 1.34,

95% CI 1.07–1.68) and >12.6 3 109/L (OR 1.29, 95% CI1.10–1.51) had significantly increased odds of 30-day read-mission (Table II).

The median (25–75th percentiles) length of hospital stayfor patients who were discharged alive by 30 days offollow-up were 5 days (4–8), 6 days (4–8), 6 days (4–8), 6days (4–8), and 6 days (5–9) for admission WBC counts<5.0, 5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6 3 109/L,respectively (P< 0.001). The corresponding readmissionincidence rates were 8.2, 5.3, 4.9, 5.2, and 7.2 per 1000hospital-free days (P<0.001), indicating that differences inreadmission rates by WBC count were not related to differ-ing length of hospital stay across categories of WBC count.

Figure 1. Patient flow chart.

Figure 2. Kaplan–Meier estimates of 30-day mortality were 10.9%, 6.2%, 5.4%,

8.3%, and 16.3% for patients with an admission white blood cell count of <5.0,5.0–7.8, 7.9–9.8, 9.9–12.6, and >12.6x109/L, respectively (P<0.001).

research article

American Journal of Hematology 679

DiscussionOur results demonstrate that after adjustment for patient-

and hospital-related confounders, and thrombolytic therapy,patients with an elevated WBC count (>9.8 3 109/L) had asignificantly higher 30-day mortality. This was also true,albeit to a lesser extent, for patients with a low WBC count(<5.0 3 109/L). Importantly, the higher mortality amongpatients with elevated/low WBC count was observed acrossall PESI risk classes. Our findings are consistent with a ret-rospective study reporting an independent associationbetween an elevated WBC count (>11.0 3 109/L) and 30-day mortality in 150 patients with PE [19]. Similarly, datafrom the RIETE registry showed that an elevated WBC issignificantly associated with adverse outcomes in patientswith acute venous thromboembolism who have cancer [24].

There are several biologically plausible explanations forthe association between an elevated WBC count andincreased mortality in PE. The relationship between elevatedWBC count and left ventricular dysfunction in patients withacute coronary syndrome and the various potential mecha-nisms of myocardial injury mediated by leukocytes havebeen described in detail [3]. There is growing evidence fromanimal and autopsy studies that acute PE with at leastmoderately severe pulmonary hypertension results in rightventricular myocyte lysis and infiltration by neutrophils, mac-rophages, and lymphocytes in humans and rats [12–15,17]and that this inflammation independently amplifies injury[25]. Therefore, an elevated WBC count may indicatePE-related right heart dysfunction, a known factor foradverse prognosis in patients with PE [11]. Evidence alsosuggests that the WBC count correlates with levels of fibrino-gen, factor VII, and factor VIII [18]. Thus, an elevated WBCmay be a marker for hypercoagulability, which may carry aworse prognosis [26]. Future prospective studies shouldexamine whether an elevated WBC count reflects right ven-tricular dysfunction and hypercoagulability in patients withacute PE.

Another explanation for the observed associationbetween elevated WBC count and adverse outcomes maybe the presence of unmeasured prognostic factors that areunrelated to right ventricular dysfunction. Besides infectionor inflammation, elevated WBC count occurs in a variety ofnon-cardiovascular clinical situations, such as trauma,intensive physical exercise, therapy with drugs such as ste-roids or lithium, malignancy, poisoning, psychosis, diabeticacidosis, and general in-hospital admissions and has beenshown to be a predictor of mortality in respiratory disease,malignancy, and head trauma [27]. Large, prospective stud-ies conducted in healthy populations have also shown anassociation between a raised WBC count and mortality[28,29]. The relationship between elevated WBC count andadverse outcomes remains incompletely understoodbecause of its association with a multitude of underlyingdisease states, suggesting that elevated WBC may be anunspecific reaction to a variety of conditions, perhaps aspart of a generalized stress response [30]. Thus, it is possi-ble that WBCs have no causal role in death following PE

but rather reflect an inflammatory response either to throm-bosis or to its risk factors.

Consistent with a study examining the relationshipbetween WBC count and mortality in patients with acutemyocardial infarction, we found a J-shaped distribution ofmortality by WBC count [2], with patients with WBC levels<5.0 3 109/L having an increased mortality rate. Given thatmore than 36% of these patients had cancer and 65% wereanemic, the low WBC count and the higher mortality likelyreflect greater severity of illness and higher prevalence ofcomorbid conditions, in particular cancer and/or itstreatments.

Our study also shows that, after adjustment for patientand hospital characteristics and thrombolytic therapy, anelevated/low WBC count in the acute phase of PE is anindependent predictor for future hospital admissions, rein-forcing the prognostic importance of this laboratoryparameter.

Our findings may have both clinical and research implica-tions. Clinically, patients with PE who have a high/low WBCcount at the time of presentation carry a higher risk ofshort-term mortality and hospital readmission, and may,therefore, potentially benefit from more intensive surveil-lance in the hospital and after discharge. However, the use-fulness of the WBC count for risk stratification in PE mustbe further validated before its clinical use can be

TABLE I. Association of Mortality and Quartile of White Blood Cell Count Stratified by Severity of Illness

Admission white blood cell count (3109/L)

Pulmonary Embolism Severity Subquartile Quartile 1 Quartile 2 Quartile 3 Quartile 4Index risk class <5.0 5.0–7.8 7.9–9.8 9.9–12.6 >12.6 P-value

30-Day mortality, n/N (%)I 2/124 (1.6) 5/745 (0.7) 5/700 (0.7) 9/715 (1.2) 10/447 (2.2) 0.10II 4/153 (2.6) 22/816 (2.7) 15/793 (1.9) 24/704 (3.4) 27/529 (5.1) 0.02III 13/188 (6.9) 36/764 (4.7) 28/742 (3.8) 51/762 (6.7) 77/664 (11.6) <0.001IV 24/177 (13.5) 41/522 (7.8) 35/547 (6.4) 49/484 (10.1) 93/590 (15.7) <0.001V 47/180 (26.1) 110/587 (18.7) 99/584 (16.9) 145/663 (21.9) 329/1048 (31.4) <0.001

TABLE II. Association Between Level of White Blood Cell Count andOutcome

OutcomesAdjusted

odds ratioa

95%Confidence

interval P-value

30-day all-cause mortality <0.001

Admission WBC count (3109/L)<5.0 1.52 1.14–2.035.0–7.8 1.20 0.97–1.497.9–9.8 1.00 –9.9–12.6 1.55 1.26–1.91>12.6 2.22 1.83–2.69

30-day readmission rateb 0.002Admission WBC count (3109/L)<5.0 1.34 1.07–1.68

5.0–7.8 1.04 0.89–1.217.9–9.8 1.00 -9.9–12.6 1.06 0.90–1.24>12.6 1.29 1.10–1.51

WBC, white blood cell.a The odds ratios were adjusted for patient demographics (age, gender, race,

and insurance type), comorbid diseases (history of cancer, chronic lung dis-ease, and heart failure) and physical examining findings (systolic arterial bloodpressure <100 mmHg, pulse �110 beats/minute, respiratory rate �30 breaths/

minute, altered mental status, body temperature <36�C, and arterial oxygensaturation <90%) comprising the Pulmonary Embolism Severity Index, labora-tory values (hemoglobin, sodium, creatinine, glucose, and troponin), thrombo-lytic therapy, and hospital characteristics (region within Pennsylvania, annualvolume of pulmonary embolism, and size and teaching status).

b Adjusted odds ratios of 30-day readmission were estimated after the exclu-sion of 840 patients who died in the hospital, 93 who were still hospitalized 30days after admission, and 34 with unknown readmission status, leaving a sam-ple of 13,261 patients.

research article

680 American Journal of Hematology

recommended. To date, the PESI remains the most vali-dated clinical prognostic tool for PE. The prognostic benefitof adding the WBC count to the PESI appears to be rathermodest.

Further research is warranted to determine whetherWBC count correlates with other inflammatory markers,such as C-reactive protein and interleukins, right ventriculardysfunction, and hypercoagulability, and whether anti-inflammatory treatment is associated with improved out-comes for patients with PE. Evidence from animal studiessuggests that treatment with anti-PMN antibodies, anti-CINC antibodies, and the nonsteroidal anti-inflammatorydrug Ketorolac reduced the inflammation in right ventriculartissue and significantly improved right ventricular contractilefunction [25].

Our study has potential limitations. First, patients in oursample were identified by use of ICD-9-CM codes for PErather than standardized radiographic criteria, and patienteligibility may therefore be subject to selection biases owingto hospital coding procedures. Although we could not vali-date the accuracy of ICD-9-CM codes for PE in our dataset,96% of patients with specific codes for PE had objectivelydocumented disease on the basis of chart review criteria ina prior study [31]. Second, our sample excluded 8.4% ofyounger, healthier, and less severely ill patients in whomthe WBC count was not measured at the time of admission.However, the exclusion of these lower risk patients, ofwhom probably a small proportion had an elevated/lowWBC count, is unlikely to change our study results. Third,because measures of right ventricular function, other inflam-matory markers (e.g., C-reactive protein, interleukins), andmarkers of hypercoagulability were not available in our data-base, we could not examine whether these measures arecorrelated with admission WBC levels. Similarly, we couldnot examine whether WBC counts are associated with PE-related cause of death. Fourth, we were not able to adjustour results for other potential confounders that may influ-ence WBC levels and prognosis, such as concomitantinflammatory or infectious diseases, the extent of cancer,and treatments (chemotherapy, steroids). Fifth, we had noinformation on WBC count after hospital admission and dis-charge; thus, the prognostic implication of transient versuspersistent elevation of WBC levels could not be analyzed.Sixth, we could not study the impact of leukocyte subtypeson patient prognosis. Finally, our study was observational innature and we could detect only statistical associations, notcausality, from our data. Thus, we cannot determinewhether WBC count has a direct effect on patient prognosisor is a mere marker of severity of illness and stress.

In conclusion, in this large sample of patients hospitalizedwith acute PE, an elevated/low admission WBC count wasassociated with a higher risk of 30-day mortality and read-mission. The WBC count may serve as an easy-to-usemarker to identify patients with PE who are at higher risk foradverse outcomes. Future studies should examine whetherWBC levels are correlated with other inflammatory markers,hypercoagulability, and right ventricular dysfunction inpatients with acute PE, and whether anti-inflammatory treat-ment is associated with improved outcomes.

References1. Barron HV, Harr SD, Radford MJ, et al. The association between white blood

cell count and acute myocardial infarction mortality in patients>or 565 yearsof age: Findings from the cooperative cardiovascular project. J Am Coll Cardiol2001;38:1654–1661.

2. Grzybowski M. The association between white blood cell count and acutemyocardial infarction in-hospital mortality: Findings from the National Registryof Myocardial Infarction. Acad Emerg Med 2004;11:1049–1060.

3. Turner SJ, Ketch TR, Gandhi SK, et al. Routine hematologic clinical tests asprognostic markers in patients with acute coronary syndromes. Am Heart J2008;155:806–816.

4. Cooper HA, Exner DV, Waclawiw MA, et al. White blood cell count and mortal-ity in patients with ischemic and nonischemic left ventricular systolic dysfunc-tion (an analysis of the Studies Of Left Ventricular Dysfunction [SOLVD]). Am

J Cardiol 1999;84:252–257.5. Fiutowski M, Waszyrowski T, Krzeminska-Pakula M, et al. Pulmonary edema

prognostic score predicts in-hospital mortality risk in patients with acute cardio-genic pulmonary edema. Heart Lung 2008;37:46–53.

6. Whiteley W, Jackson C, Lewis S, et al. Inflammatory markers and poor out-come after stroke: A prospective cohort study and systematic review of inter-leukin-6. PLoS Med 2009;6:e1000145.

7. Whiteley W, Jackson C, Lewis S, et al. Association of circulating inflammatorymarkers with recurrent vascular events after stroke: A prospective cohort

study. Stroke 2011;42:10–16.8. Bogaard HJ, Abe K, Vonk Noordegraaf A, et al. The right ventricle under pres-

sure: Cellular and molecular mechanisms of right-heart failure in pulmonaryhypertension. Chest 2009;135:794–804.

9. Centers for Disease Control and Prevention. National hospital discharge sur-vey 2010. Accessed December 23, 2012, http://www.cdc.gov/nchs/data/nhds/10Detaileddiagnosesprocedures/2010det10_numberfirstdiagnoses.pdf.

10. Aujesky D, Obrosky DS, Stone RA, et al. Derivation and validation of a prog-nostic model for pulmonary embolism. Am J Respir Crit Care Med 2005;172:

1041–1046.11. Sanchez O, Trinquart L, Colombet I, et al. Prognostic value of right ventricular

dysfunction in patients with haemodynamically stable pulmonary embolism: Asystematic review. Eur Heart J 2008;29:1569–1577.

12. Watts JA, Gellar MA, Obraztsova M, et al. Role of inflammation in right ven-tricular damage and repair following experimental pulmonary embolism in rats.Int J Exp Pathol 2008;89:389–399.

13. Watts JA, Zagorski J, Gellar MA, et al. Cardiac inflammation contributes toright ventricular dysfunction following experimental pulmonary embolism inrats. J Mol Cell Cardiol 2006;41:296–307.

14. Iwadate K, Doi M, Tanno K, et al. Right ventricular damage due to pulmonaryembolism: Examination of the number of infiltrating macrophages. ForensicSci Int 2003;134:147–153.

15. Begieneman MP, van de Goot FR, van der Bilt IA, et al. Pulmonary embolismcauses endomyocarditis in the human heart. Heart 2008;94:450–456.

16. von Bruhl ML, Stark K, Steinhart A, et al. Monocytes, neutrophils, and plate-lets cooperate to initiate and propagate venous thrombosis in mice in vivo.J Exp Med 2012;209:819–835.

17. Iwadate K, Tanno K, Doi M, et al. Two cases of right ventricular ischemic injury

due to massive pulmonary embolism. Forensic Sci Int 2001;116:189–195.18. Bovill EG, Bild DE, Heiss G, et al. White blood cell counts in persons aged 65

years or more from the Cardiovascular Health Study. Correlations with base-line clinical and demographic characteristics. Am J Epidemiol 1996;143:1107–1115.

19. Huang CM, Lin YC, Lin YJ, et al. Risk stratification and clinical outcomes inpatients with acute pulmonary embolism. Clin Biochem 2011;44:1110–1115.

20. MacMahon B. The National Death Index. Am J Public Health 1983;73:1247–1248.

21. Williams BC, Demitrack LB, Fries BE. The accuracy of the National DeathIndex when personal identifiers other than Social Security number are used.Am J Public Health 1992;82:1145–1147.

22. Cowper DC, Kubal JD, Maynard C, et al. A primer and comparative review ofmajor US mortality databases. Ann Epidemiol 2002;12:462–468.

23. Pencina MJ, D’Agostino RB Sr., Steyerberg EW. Extensions of net reclassifica-tion improvement calculations to measure usefulness of new biomarkers. StatMed 2011;30:11–21.

24. Trujillo-Santos J, Di Micco P, Iannuzzo M, et al. Elevated white blood cell count

and outcome in cancer patients with venous thromboembolism. Findings fromthe RIETE Registry. Thromb Haemost 2008;100:905–911.

25. Watts JA, Marchick MR, Kline JA. Right ventricular heart failure from pulmo-nary embolism: Key distinctions from chronic pulmonary hypertension. J CardFail 2010;16:250–259.

26. Agterof MJ, Schutgens RE, Moumli N, et al. A prognostic model for short termadverse events in normotensive patients with pulmonary embolism. Am JHematol 2011;86:646–649.

27. Asadollahi K, Beeching NJ, Gill GV. Leukocytosis as a predictor for non-

infective mortality and morbidity. QJM 2010;103:285–292.28. de Labry LO, Campion EW, Glynn RJ, et al. White blood cell count as a pre-

dictor of mortality: Results over 18 years from the Normative Aging Study.J Clin Epidemiol 1990;43:153–157.

29. Jee SH, Park JY, Kim HS, et al. White blood cell count and risk for all-cause,cardiovascular, and cancer mortality in a cohort of Koreans. Am J Epidemiol2005;162:1062–1069.

30. Benschop RJ, Rodriguez-Feuerhahn M, Schedlowski M. Catecholamine-induced leukocytosis: Early observations, current research, and future direc-

tions. Brain Behav Immun 1996;10:77–91.31. White RH, Garcia M, Sadeghi B, et al. Evaluation of the predictive value of

ICD-9-CM coded administrative data for venous thromboembolism in theUnited States. Thromb Res 2010;126:61–67.

research article

American Journal of Hematology 681