safermedicines report - acmedsci.ac.uk › file-download › 34854-51e517b083b77.pdf · as unique...

1

Risk-benefit assessment working group report

Safer Medicines Report


November 2005

2


Contents

Page

Summary and recommendations 3

Chapter one: Introduction 8

Chapter two: Drug safety: perceptions, awareness and expectations of risk 9

Chapter three: Introduction to risk assessment and risk management: general principles 11

Chapter four: Risk management 12

Chapter five: Risk assessment: the essential elements 14

Chapter six: Concordance of toxicity in animals and adverse events in humans 26

Chapter seven: Species-specific toxicity and inter-individual variation 28

Chapter eight Dose selection 30

Chapter nine: Idiosyncratic adverse events 31

Chapter ten: Assessment of carcinogenic potential 34

Conclusions 44

References 45

Abbreviations 52

3


� better mechanisms and methods of communicatingrisk to patients and doctors, and to receivefeedback, to include briefings by well informededucators familiar with the drug and the disease

� revisions to existing regulatory guidelines, asindicated by new knowledge, to ensure continuingutility of testing requirements and to encouragedevelopment and deployment of modern methods

� harnessing the potential of new technologies andestablishing priorities for work to validate newbiomarkers for drug and patient safety, and riskassessment

� consultation to establish new paradigms for dataanalysis including deployment of expert systems inknowledge management, data mining, and systemsbiology

� expanding the knowledge base to develop andrefine predictive physiologically-based,pharmacokinetic and pharmacodynamic modelingand simulation tools to increase precision in riskassessment

� collaborative efforts between industry, academiaand regulatory scientists to investigate themechanistic and constitutive basis of species-specific toxicity or atypical manifestations oftoxicity in animals; determine their potential utilityas unique tools to investigate certain forms ofhuman toxicity

� consistent standards for characterizing drug safetyand risk with transparency of methods and logicapplied to risk assessment; publication of outputs toshare best practice

� establishing the infrastructure for a National Centrefor Safety Assessment to provide focus andleadership to the foregoing initiatives

Among the greatest threats to successfully deliveringthese initiatives are the challenges relating to: a)managing a burgeoning volume of data generated bylarge scale screening technologies b) setting the rightpriorities for translating new information into reliableknowledge c) converting new knowledge into usablemethods d) maintaining public confidence thatinvestment in new methods will deliver safermedicines e) maintaining trust and confidence at the

Considering the number of safe and efficaciousmedicines approved during the last 50 years it isreassuring that the number withdrawn because ofunacceptable side effects is, by comparison, verysmall. The reported rate of marketed drug withdrawalshas been fairly stable (1) or has declined slightly overthe past two decades, down from 3.5% to 1.2% ofslightly more than 500 new drugs approved in theUSA between 1979 and 1998 (2,3). Given dramaticadvances in medical knowledge over the last decade,and the diversity of new drugs currently underdevelopment, it is timely to review current practices inthe light of these new developments.

This report deals only in broad outline with certainaspects of drug safety and the complex process that isrisk assessment. Aspects of risk assessment pertainingto reproductive function, embryo-fetal development,post-natal development and the special needs ofchildren as they relate to drug safety are beyond thescope of this high level review.

In considering the remit of the project “SaferMedicines – new approaches to ensuring the safety ofmedicines” sponsored by the Academy of MedicalSciences’ Forum, the Risk-Benefit AssessmentWorking Group has identified a number of initiativesthat should be undertaken, with governmentagencies, academia and the pharmaceutical industryworking in partnership, to further ensure the safety offuture medicines. These initiatives are summarizedunder each of the recommendations for action thatfollow, and are further detailed within the text of thisreport. In brief these initiatives include:

� adoption of “One Medicine” as an operationalprinciple, wherein all disciplines cooperate fully tosupport an unbroken line of scientific enquiry fromdiscovery through to delivery of a new medicine topatients

� actions to increase public awareness that nomedicine is without risk

� steps to improve process and communication at theinterface between industry, academia andregulatory scientists

� commitment to an hypothesis-driven, evidence-based approach to drug safety evaluation and riskassessment

Summary

interface between industry and regulatory agencieswhile managing these uncertainties, and f) evolving anew paradigm for drug safety evaluation and riskassessment wherein active pursuit of knowledge aboutrisk is the new benchmark for establishing the safety ofnew medicines.

These challenges clearly call for a sustainedinvestment in pharmaceutical R&D and academicresearch in all the relevant disciplines across the ‘OneMedicine’ spectrum. They also demand a sustainedlevel of inward investment to provide a stream ofqualified people with the requisite knowledge andskills needed to maintain the tradition of excellence inUK science and innovation in pharmaceutical R&D.

With intense public attention focused on drug safety atthe present time, moving these initiatives forward willprogressively enhance the scientific underpinning ofdrug safety evaluation and risk assessment. Thedevelopment of better methods and tools with whichto characterize, manage and avoid risk of drug adverseeffects is key to providing patients with a steady streamof treatments for hitherto difficult and intractablediseases.

Recommendations

Numerous organizations and institutions have thecapacity, knowledge and expertise, as well as theadministrative infrastructure, required to assess themerits of the recommendations listed below. Themembership of professional societies in the UK andEurope, representing the disciplines of Pharmacology,Toxicology, Pathology and various branches ofMedicine, along with the respective Colleges,Industrial Associations and the Academy of MedicalSciences can provide valuable impetus to the workrequired to facilitate implementation of thoseinitiatives deemed worthy of further investment.

Recommendation 1: UK Health and regulatoryauthorities, in partnership with the UK basedpharmaceutical industry, should undertake a variety ofinitiatives to increase public awareness that someelement of risk is endemic to taking a medicine.

� There is a need to review current systems andmechanisms for disseminating information aboutthe safety, risks and benefits of newly approveddrugs to patients, health care professionals and the

4


public at large. Mechanisms for receiving patientfeedback are needed to gauge the level of toleranceto risk that different patient groups are willing toaccept set against the benefits of their medicine.Regulators and industry need to find mechanismsto address net patient benefit-risk such that they arenot inappropriately penalized by a litigious system.

� The public must be informed that lifestyle anddietary habits are important in managing risk, andavoiding adverse effects when taking any medicine.

� Approval to market a new drug should becontingent on establishing a panel of educatorsto provide ongoing educational briefings andupdates to patients and health care professionals,these to be administered within the framework ofexisting Continuous Professional Developmentprogrammes.

Recommendation 2: UK and EU Health andregulatory authorities should publish a‘Memorandum of Understanding’ endorsingflexibility and innovation in the process of drug safetyand risk assessment.

� There is a need to foster a culture of trust, dialogueand transparency between scientists in industry andregulatory agencies. Early engagement indiscussion of key issues leading to creation of a riskmanagement plan for a new drug candidateentering development would provide a vehicle tofacilitate scientific exchange.

� A Risk Management Plan should be an integral partof all interim regulatory submissions serving as aflexible blueprint throughout the course ofdeveloping a new drug candidate with the objectiveof identifying and predicting potential safety issueson a case-by-case basis. Such investigative workshould be hypothesis-driven aiming to prioritisepotential risk factors in a manner relevant to thedisease target, the drug’s mechanism of action, thedisease indication, special sub-populations at risk,and overall consideration of risk in relation topotential benefits.

� Actively exploring the boundaries of risk is a vitalpart of confirming safety, and this requires latitudewithin the existing framework of regulatoryguidelines to employ novel study designs and non-routine testing paradigms in both pre-clinical andclinical phases of drug safety assessment. This alsocalls for inter-agency dialogue and unification of

5


administrative processes and communicationnetworks in order to avoid duplication of effort andinconsistency in regulatory decision-making.

Recommendation 3: Representatives of the UKand EU Health and regulatory authorities andrepresentatives of the pharmaceutical industry shouldjointly review the ICH guidelines with a view toensuring economies of scale and relevance of currentpractices to human risk assessment based onaccumulated knowledge and access to newtechnologies.

� While ICH guidelines have proven effective it istimely to consider their applicability within thecontext of recent advances in technology andinvestigative capabilities. The new paradigm indrug safety assessment should be hypothesis-drivenwith emphasis on predicting toxicity and exploringthe potential for risk in humans.

� For example, the need for 2 species in generaltoxicology and carcinogenicity studies, the need fortoxicology studies greater than 90 days in duration,and the requirements for pre-clinical studies priorto the first exposure of human volunteersparticipating in ‘Experimental Medicine’ studiesshould be re-assessed in the light of past humanexperience with all classes of medicine.

� There is a need to facilitate transference andvalidation of new biomarkers of importance tohuman risk assessment (so-called bridgingbiomarkers) from the research laboratory into theclinic. Consensus must be reached on appropriatetesting and validation strategies including means ofgaining access to archives of chemical structuresand relevant data held by pharmaceuticalcompanies without breaching intellectual propertyinterests.

� The possibility of developing new methods toevaluate candidate drugs for toxic effects in lowerorder organisms should be investigated asknowledge of cellular pathways related to commonpathogenic mechanisms are revealed by analysis oflarge scale gene and protein expression data.

Recommendation 4: Representatives of the UKand EU Health and regulatory authorities, thepharmaceutical industry and academia shouldconvene in scientific session to address the utility ofcomputational ‘knowledge management’ and advanced

‘data mining’ tools to detect signals of toxicity oradverse effect in pre-clinical and clinical data.

� There is a need for strategies to review outliersubjects in animal and human studies focusing inparticular on temporal, quantitative and qualitativeexposure-response relationships in each case for thepresence of individual risk factors.

� Greater emphasis needs to be placed on relationalanalysis of conventional laboratory data forevidence of mutually reinforcing patterns ofresponse that may point to incipient organ systemtoxicity.

� Resolution and integration of genomic andproteomic data with data derived from measures ofconventional endpoints will require a ‘systemsbiology’ approach to elucidate complex cellularpathways related to defined modes of toxicity.

Recommendation 5: Representatives of the UKand EU Health and regulatory authorities, thepharmaceutical industry and academia shouldconvene in scientific session to establish guidingprinciples for use in exposure assessment and dose-response analysis for human risk assessment. There isa need for greater consistency and quality of output.

� The product of this review should be the issuance ofguidance notes on the essential principles thatshould be followed with specific reference to thelogic supporting selection of dosimetry parameters,the methods and models employed for inter- andintra-species extrapolation of exposure, and theextent of uncertainty and the potential for error.

� A second output of the review should bepublication of non-competitive or anonymised datain the form of case studies illustrating the bestexamples of robust exposure assessments and dose-response analyses with the objective of sharing bestpractices.

Recommendation 6: The evolution of in silico andin vitro methods for prediction of drug kinetics anddrug-drug interactions in vivo and the incorporation ofsuch data into sophisticated physiologically-basedpharmacokinetic models should be encouraged bothin virtual mode and for simulation of conditionsrelevant to patient and disease demographics.

� Standardisation of in vitro models for absorption,models of blood-brain barrier transfer under

defined pathological states, standardised screens forenzyme induction, standardization of methods forenzyme inhibition, and integration of in vitroADME studies with cell-based systems for evaluationof toxicity are priorities for further development.

� Additional investment is needed to expanddatabases of disease prevalence, genetic variability,enzyme/transporter abundancies and interactionsin order to refine predictive algorithms for use inreal as well as virtual human populations. Thisfacility would enable linkage of physiologically-based PK models to mechanistic pharmaco/toxico-dynamic (PD/TD) models in order to put PKvariability into context for projections of efficaciousas well as potentially toxic exposures.

Recommendation 7: Representatives of the UKand EU Health and regulatory authorities, thepharmaceutical industry and academia shouldconvene in scientific session to establish guidingprinciples for risk characterization. There is a need forimprovements in transparency, consistency andquality of output.

� The product of this review should be the issuance ofguidance notes on the essential elements of a soundmechanism-based, weight-of-evidence approach torisk characterization, where risk is invariablyviewed in relation to benefit.

� There is a need to develop consensus on the utilityand validation of biomarkers of host response andnovel biomarkers of exposure that would permitrisk characterization in animals and humans withinthe confines of a given species, or sub-group ofpatients, or individual patients. This approachshould be considered in the context of other relatedstrategies to deliver personalized medicine.

� Risk-benefit summaries should be published after adrug is approved for marketing in order to sharebest practice and insights into the basis forregulatory decision-making.

Recommendation 8: Create an organizationalmatrix to function as a National Centre for SafetyAssessment to promote an evidence-based approachto drug safety and risk assessment in the pre-clinicaland clinical disciplines.

� A National Centre for Safety Assessment couldfulfill several important functions:

6


� oversee development and implementation ofeducational and training programmes

� provide focus and leadership on issuesrelating to policy making

� coordinate clinical and experimentalinvestigations of adverse drug reactions andprovide a vehicle for collaboration andsharing of samples with other centres in theUS and Europe

� review gaps in knowledge relating to thecause of idiosyncratic and other forms ofunanticipated reactions to drugs in regularuse, and institute appropriate actions

� engage with other institutions internationallyto design and populate databases and toestablish accessible tissue, cell and serumbanks

Recommendation 9: Establish an internationalprogramme of collaborative research betweenscientists in industry and academia, to be funded atleast in part by industry, to investigate phenotypic,genotypic, and other constitutive differences that mayaccount for manifestations of species-specific toxicity.

� The principal objectives would be to:

� elucidate the basis for species differences intoxic response to certain classes of drugsperceived to pose a high risk for humans

� characterize genetic polymorphisms andallelic variations in toxicology species thatmay be relevant to the pathogenesis of class-specific or target- specific toxicity

� accumulate a shared repository of phenotypicand genotypic data from animals intoxicology studies exhibiting atypicalresponses to various classes of drugs;determine how this information might beused to develop more informative models ofhuman risk

� increase knowledge of species differencesin organ pharmacology and physiologythat would contribute to more accuratephysiologically-based pharmacokineticmodeling

� provide evidence to underpin the selection ofanimal species and/or animal models for

7


toxicity testing having greater relevance andpredictive capacity for human risk assessment

Recommendation 10: Representatives of the majorHealth and regulatory authorities internationally, andthe pharmaceutical industry, should convene to reviewexisting ICH guidelines for assessing carcinogenichazard with a view to adopting a modified testingstrategy in line with contemporary knowledge. TheWorking Group proposes that in the short term:

� analysis of structure-activity relationships and abattery of genotoxicity tests would eliminate overtlygenotoxic chemicals from consideration as new drugcandidates, as currently practiced

� a bioassay in one rodent species, the rat, would beused to detect carcinogenic potential other than bydirect genotoxicity

� in the event of a tumorigenic effect in the rodentbioassay additional mode of action studies would beconducted to ascertain the significance andrelevance of such observations to human riskassessment

� further modifications of testing strategies could beintroduced in the longer term in the light ofexperience and accumulated knowledge and thefurther evolution of new and advanced methods forassessing carcinogenic hazard

Recommendation 11: Representatives of the UKand EU Granting Authorities, Academic Authorities,Health Authorities, and the pharmaceutical industryshould convene to review the training of basicscientists and physicians, both at the undergraduateand postgraduate level.

� there is a need to ensure that the re-supply ofscientists and physicians in the future is consistentwith meeting the challenges and objectives implicitin recommendations 1 through 10

� appropriate steps should be taken to ensure thatmedical curricula include courses with a specificfocus on drug safety, risk assessment and riskcommunication

8


suggests that most of the time, in the majority ofcases, existing processes and procedures havedelivered efficacious and safe medicines topatients. Whilst it is appropriate to ask whetherwe can do more to avoid even these rareinstances of adverse reactions, it is alsoimportant to consider whether the scrutiny ofdrug candidates has now become so stringentthat even potentially safe and beneficial drugsare failing to reach the marketplace, to theultimate detriment of patients.

1.5 In this report we review certain of the activitiesconnected with the discovery and developmentof a new drug in order to give perspective andcontext to the process of assessing risk inrelation to benefit. It is clear that successfuldelivery of safe medicines in the twenty firstcentury depends more than ever before on theexercise of original thinking, innovativeapplication of science and technology,interdisciplinary collaboration and seamlesscommunication between all participants in theenterprise. There is a need to build trustbetween industry and regulatory agencies such that the safety of each new drug may be addressed on its scientific merits, case-by-case. There is a need to build consensus across Regulatory Agencies on issues of highestpriority so that regulatory decisions affectingdrugs in development and those on the marketare based on objective evidence. Equally, when failure occurs, there must beencouragement and latitude to carry out a full scientific investigation, without punitive consequences, so that by sharingknowledge similar events may be averted infuture.

1.6 Finally, the report stresses the importance ofusing this evidence base as the platform fromwhich to communicate relevant informationabout a drug’s safety to the public and healthcare professionals.

1.1 The sum of activities required to deliver a newmedicine can be likened to those preceding thelaunch of a manned space mission. Both arecomplex enterprises imbued with someuncertainty. Progress is conditional on thepotential for gain outweighing the risk. Despiteall efforts to manage risk rare missions do end infailure, mostly for reasons that are difficult topredict. As with manned space missions so toowith a new medicine: complete safety cannot beassured at all times.

1.2 Eight hundred million dollars, eight to tenyears, and hundreds of thousands of pages ofdocumentation sum up the cost, time, and effortto demonstrate that a new drug is safe andefficacious for it to gain approval byinternational Regulatory Agencies (4,5).

1.3 Throughout this long and costly process theincidence, duration and severity of all adverseevents are assiduously recorded. After approvalis given to market a drug adverse events arerecorded by Pharmaceutical Companies andHealth Authorities and other agencies aroundthe world. Pharmaceutical companies areobligated by law to provide safety updates on aregular schedule.

1.4 Despite this intensive scrutiny, a smallproportion of new medicines, at varying timesafter approval as marketed products, have beenlinked with serious adverse reactions causingthem to be withdrawn (1,2,3). The overallincidence of serious adverse reactions is difficultto estimate with complete accuracy but onereview concludes that between one in thirty andone in sixty physician consultations result fromadverse drug reactions (6). The frequency ofadverse drug reactions may range from 1 in1,000 to 1 in 40,000 patients for a particulardrug (2,7,8,9). In rare cases adverse reactionsmay cause death while others are transient andeasily remedied. On balance, the record

Chapter One - Introduction

9


2.1 Patients elect to take medicines for theirailments on the advice of their physician; manyexpect their medicines always to be safe andeffective. Such patients are not attuned to thepossibility that a medicine prescribed by theirphysician will cause them harm.

2.2 Drug failures related to safety or efficacy resultin widespread and generally negative publicity.This erodes public confidence and trust in thesystem. Yet, given the millions of prescriptionsdispensed each day, adverse reactions to drugtreatment are relatively rare and the benefits tohuman health and society at large are generallywell appreciated (10,11).

Challenges and opportunities

2.3 It is unlikely that all risk associated with takinga medicine will ever be eliminated completely.Whilst every effort is taken to minimize harm,some degree of risk is endemic to anytherapeutic intervention. There will always bean element of uncertainty in determining theprecise level of risk. Balancing risk againstbenefit is individual to each patient.

2.4 It is important that patients recognise theindividual nature of their treatment and the needto report any untoward effects. The presentmeans of communicating this kind of informationto patients is inadequate. Patients should be madeaware that they all may not experience the samebenefit of treatment, and that the greatest benefitand the greatest risk may not be the same for eachperson; in other words the risk-benefit ratio maybe different for the individual compared to theaverage of all patients in general.

2.5 Patients should be encouraged by their physicianto seek up-to-date information about the safety oftheir medicines and to anticipate that thisinformation may change over the course of time.This would enable patients to participate fully indiscussions of possible risk and the hoped-forbenefits offered by their treatment.

2.6 Accurate feedback from patients would be ofimmense value in informing regulators andhealth care professionals whether they feel thebenefit of treatment justifies the risk. Thisinformation would help to ensure thatregulatory decisions were consistent with theneeds of patients.

2.7 Patients should be informed that most adversereactions to drugs arise from taking the wrongdose, or the right dose at the wrong time, ortaking mixtures of medicines at the same time.Patients are largely unaware that alcohol, dietand smoking all add to the risk of their sufferingside effects or failing to get the full benefits oftreatment. Compliance can be a majordeterminant of therapeutic outcome. However,the impact of poor compliance varies with thedrug. Patients need to be better informed aboutthe importance of compliance for their specifictreatment.

2.8 When a new medicine claiming improvedefficacy and safety is introduced on to themarket physicians and healthcare professionals,particularly those in general medical practice,have little time to assimilate the detailscontained in the large volume of scientific andmedical evidence supporting such claims.

2.9 Pharmaceutical companies, scientists inregulatory agencies, members of the medicalcommunity, and the public at large each have adifferent view of the nature of risk. What is anacceptable level of risk to one group is oftenunacceptable to another. Failure to achieveconsensus on the acceptable level of risk topatients can end development of a promisingnew drug, or lead to withdrawal of a useful drugfrom the market.

Recommendations

2.10 UK Health and regulatory authorities, inpartnership with the UK based pharmaceuticalindustry, should undertake a variety of

Chapter two - Drug safety: perceptions, awareness andexpectations of risk

10


initiatives to increase public awareness thatsome element of risk is endemic to taking amedicine.

2.11 Action plans should be developed to achievethe following objectives:

� to increase public awareness andunderstanding that drug safety is based on abalanced appraisal of risk and benefit, andthat patients play a key role in ensuring safeuse of medicines,

� to establish formal mechanisms with which tosolicit and record patient feedback in order togauge their level of tolerance to risk. Patientfeedback should be subjected to periodic andindependent review after approval to marketa new drug is granted,

� to conduct a comprehensive review ofexisting systems and methods ofcommunication used to disseminateinformation about the safety, risks andbenefits of newly approved drugs tophysicians, patients, and the public at large,and,

� to establish a mechanism by which a panel ofeducators is appointed to brief health careprofessionals and patients about the risks andthe benefits of newly approved drugs.Approval to market a new drug should becontingent on establishing such a panel.Educational briefings should be administeredwithin the existing framework of “ContinuousProfessional Development” programmes.

11


3.1 Risk assessment is a process of gatheringevidence to determine the potential of a newmedicine to cause undesirable effects in patientsunder certain conditions of exposure and use.

3.2 The output of a risk assessment defines thoseconditions that may pose a hazard, eitherrelated to the dose, the disease, or an attribute ofthe patient receiving treatment. In this caseregulatory policy seeks to manage risk throughlabeling a drug with appropriate precautions.

3.3 Safety assessment is principally concerned withsetting limits of exposure consistent with theavoidance of harm. In this case regulatorypolicy adopts the precautionary position thatseeks to eliminate risk. Such differences inregulatory policy are believed by some toaccount for differences in the rates of drugapproval by health authorities in variousregions of the world (7). There are concerns thatsuch inconsistencies will limit innovation andcurtail the development of much needed newmedicines without manifest improvement indrug safety.

3.4 The process of risk analysis involvescollaboration across several disciplines andagencies. For ease of description it may bebroken down into the following constituentactivities:

� hazard identification – the qualitative andquantitative assessment of adverse effectscaused by a new chemical entity using a rangeof test systems e.g. humans, animals or cells intissue culture,

� hazard characterization – the exploration ofmechanisms underlying, or related to, the

development of adverse effects in a testsystem; establishing the relationship betweenthe incidence and prevalence of adverseeffects caused by exposure to a range ofdoses/concentrations of the administereddrug,

� exposure assessment – the characterization of thepattern of exposure, preferably internalexposure, to a drug with respect to duration,frequency, and intensity of peak and averageconcentrations in plasma and, where possible,in target tissues using modeling techniques asappropriate, and,

� risk characterization – the integration ofinformation obtained in the course of ‘hazardidentification,’ ‘hazard characterization,’ and‘exposure assessment’ in order to estimate theprobability of adverse effects occurring inhumans under specific conditions of exposureand the identification of any potentiallysusceptible sub-groups,

� risk communication – the translation of allinformation relevant to risk assessment intoclear messages acceptable to regulators andeasily understood by patients and writers ofprescriptions,

� risk management – encompasses all decisionsbased on the outcome of risk assessment. Thismay cause termination of a drug indevelopment, limit the dose given to patientsin clinical trials or restrict testing to certainpatient categories. A marketed drug may becontraindicated in certain patients or may bewithdrawn altogether.

Chapter three - Risk assessment and risk management: general principles

12


4.1 Risk management embodies all activities anddecisions necessary to ensure drug and patientsafety. The proceedings of an internationalworkshop sponsored by the Centre forMedicines Research International in 2002 (12)considers the role of risk management strategiesin drug development commenting on thestrengths and weaknesses of current practices.The current risk management structure for thedetection, assessment, and communication ofrisk is optimized to ensure the safety of newdrugs at the time of approval and post licensingwhen the product is launched on the market(13). However, regulatory scientists recognizethe need to formulate new approaches to riskmanagement, particularly during the course ofdevelopment, where there is a need to exploreand predict the potential for risk in addition tocompiling evidence of safety (13,14).

Challenges and Opportunities

4.2 The drug discovery and development process iscarried out in accord with stringent ethical, legaland regulatory requirements. New drugcandidates are subjected to comprehensivetesting for efficacy and safety in animals andhumans. These investigations are carried outaccording to standardized testing protocols thatmeet the regulatory requirements of, amongstothers, the three major Health and regulatoryauthorities in the USA, Europe and Japan.

4.3 Consultation between industry and regulatoryscientists to produce a risk management planearlier in the discovery-development processwould promote trust and confidence while, atthe same time, supporting innovation in riskassessment. Such a collaborative approachwould allow drug development to proceed in aless risk-averse environment and wouldencourage greater latitude to investigate thenature of adverse effects in a pro-active ratherthan in a reactive mode. This would introduce anew paradigm into drug development whereobservations confirming safety as well asinvestigations relevant to exploring the

boundaries of risk would be encouraged. Aneffective risk management strategy shouldanticipate and even predict undesirable sideeffects while at the same time laying downcontingencies for managing untoward events.This approach would ameliorate risk to patientsin clinical trials but most importantly wouldallow full exploration of a drug’s potentialbenefits in the disease under investigation.

Recommendations

4.4 UK and EU Health and regulatoryauthorities should publish a “Memorandum ofUnderstanding” endorsing flexibility andinnovation in the process of drug safety and riskassessment within the framework of existingguidelines. A central tenet should be theencouragement of open dialogue with scientistsin pharmaceutical companies in the process ofcreating a Risk Management Plan for a newcandidate drug as follows:

� risk management plans should be developedon a “case-by-case” basis in consultation withregulatory scientists. Investigative workshould be “hypothesis-driven” consistent withthe disease, the patient’s needs, the drug’smode of action, its chemical structure, itsmetabolism and elimination, and theanticipated schedule of dosing,

� the plan should be regarded as a flexibleblueprint allowing hypotheses to be tested inorder of priority. The plan should describeactions to avoid risk and to manage adverseevents should they occur. Plans should bemodified in the light of emerging data fromanimal studies and clinical trials,

� risk management plans should be arequirement of all regulatory applications toprogress development of a new drug andshould be updated on a regular basis,

� risk management plans should be included innew drug applications such that they are thebasis for designing further studies to evaluate

Chapter four - Risk management

13


drug safety after approval to market isgranted, and,

� regulatory and industry scientists should beencouraged to consult jointly with academicexperts and others with relevant expertise to

refine risk management plans and to resolvedifferences of opinion or concerns relating tothe interpretation of new experimental dataor safety issues arising.

14


5.1 The objective of risk assessment is to define anyconditions of exposure in humans that mayresult in adverse effects following administrationof a drug. It is not the purpose to predict theprecise incidence of adverse effects in aparticular patient population or to estimate thelevel of exposure that will most frequently beassociated with adverse effects. Although thereare a number of largely uncontrolled variablesthat determine an individual’s response to adrug, the one constant in medical practice is thedose administered to the patient.

5.2 In future, with a better understanding of thecauses of variability in the response to drugs, itmight be possible to assess risk to the individualpatient. In an idealized situation this wouldenable the prescribing physician to adjust thedose or the schedule of dosing according to thepatient’s metabolic or genetic predispositions aswell as the severity or sub-type of their disease(15,16). Thus, in the future, risk assessment andrisk management will be an integral part of eachtreatment regimen with adjustments beingmade according to the needs of the individualpatient.

5.3 The following sections present in broad outlinethe various activities that underpin the processof risk assessment. Aspects of risk assessmentpertaining to reproductive function, embryo-fetal development, post-natal development andthe special needs of children as they relate todrug safety are beyond the scope of this highlevel review. Certain of the principles apply topost-marketing surveillance but this review doesnot address the specialist aspects of researchassociated with pharmaco-epidemiology orpharmacovigilance.

Hazard identification

5.4 The primary objective is to characterize thenature and severity of adverse effects that maybe elicited by a drug over a range of doses inanimals or humans. International regulatoryguidelines and articles of the Geneva Convention

and the Declaration of Helsinki prohibitexposure of humans to a new drug candidatewithout prior assessment of risk. To datelaboratory animals have been the surrogates ofrisk for humans who volunteer to participate inclinical trials of hitherto untested drugs. Whilethe search for alternative test systems continues,and progress in certain areas has been possible,it does not appear that cell-based systems willreplace all animal use for this purpose in theforeseeable future.


5.5 Cell culture cannot duplicate conditions in thewhole animal that: i) regulate the supply ofoxygen and nutrients, ii) facilitate the removalof waste products, iii) maintain blood pressureand flow to organs and tissues. All of thesefunctions are highly integrated in consciousmammalian organisms and they govern thebody’s response to drugs and other chemicals inthe environment.

5.6 However, cell-based systems may be adapted toinvestigate specific mechanisms of toxicitywhere these can be studied under the controlledconditions of cell culture. Such in vitro methodsare used frequently to supplement studies inwhole animals.

5.7 Drug development is regulated by guidelinesrecently agreed under the auspices of theInternational Conference on Harmonisation ofTechnical Requirements for the Registration ofPharmaceuticals for Human Use (ICH). Theseguidelines have been in effect for slightly morethan a decade and stipulate the studies requiredin animals and humans to secure approval tomarket a new drug. The time is right to re-examine the requirements detailed in theseguidelines in the light of experience of drugsafety testing and the availability of new testingmethods over the last decade.

5.8 Large-scale screening technologies (genomics,proteomics, metabonomics) and other tools of

Chapter five - Risk assessment – the essential elements

15


molecular biology are providing new insightsinto mechanisms of toxicity that areincreasingly important in drug safety and riskassessment. It is widely acknowledged thatharnessing the resolving power of these newtechnologies in a fashion that delivers validbiomarkers of drug efficacy, safety, and risk is amajor challenge for the next decade. Theuniverse of chemical structures, biologicaltargets, evoked pathologies and the range ofhost responses is seemingly limitless (17).

5.9 Large government funded research agenciessuch as some branches of the EPA and theNIEHS in the US are heavily invested in theseactivities and they are engaged with scientistsfrom pharmaceutical and other companies,academia and European agencies in multi-national collaborative projects (18). Thechallenge is to translate a burgeoning volume ofinformation into usable diagnostic andpredictive assays of utility in drug safetyassessment. While changes in certain genefamilies are claimed to correlate withobservations of pathology, few if any have yetbeen validated to a point where they can beused in risk assessment (19). It is alsorecognized that certain gene changes are proneto artifact arising from variations inexperimental procedures (20). There is clearlya need to focus these efforts on the particularneeds of drug safety assessment, and UK andEU based scientists and their colleagues in therespective regulatory agencies should be fullyengaged in setting priorities for further work.

5.10 With knowledge of human and animal genomesequences now available comparison of codingregions and gene regulatory elements, togetherwith appropriate functional studies, should helpin the selection of animal species best adapted toexpress the biological and/or toxicologicalactivity of a new drug candidate and, therefore,of greatest utility in determining risk to humans.This would extend from considerations ofspecies similarities in metabolic pathways anddetoxification mechanisms, as well as anyspecies-specific susceptibility to chemicaltoxicity or immune mediated adverse effects.

5.11 There is a growing body of evidence suggestingthat fewer genes and proteins than hitherto

evaluated may be critically involved inregulating cell signalling pathways whichcontrol cell replication, programmed cell deathand other key functions (21,22); similarly, it isconceivable that host responses to chemical ormetabolic stress, reactive intermediates orexternal environmental agents, may beassociated with pathognomonic changes inparticular subsets of genes and proteins.

5.12 As better resolution of these critical cellular andmolecular pathways is achieved it is possible toenvisage that cell-based assays equipped withsuitable response elements, target genes andreporter constructs as well as lower orderorganisms will provide facile and validatedtesting methods with which to predict specifichazards. For example, Caenorhabditis elegans(23), saccharomyces cerevisiae (22) or the zebra fish(24) may serve as useful test systems in thefuture in a manner similar to Salmonella andmouse lymphoma cells used currently forevaluation of genotoxicity (25).

5.13 The concept that key biological processes areregulated by the interaction of chemicals in adefined ‘chemical space’ with targets in acorresponding ‘biological space’ suggests thatthere may be a finite repertoire of chemicalstructures that elicit activity at specificmolecular targets of toxicity. As mentionedabove there are many ongoing efforts tovalidate genomic and proteomic biomarkersemploying prototypic toxicants with a knownrange of toxic effects. This strategy may sufferthe disadvantage of discovering markers only ofpredictable responses and outcomes. There is aneed to access more diverse chemical structuresand data related to their potency and specificityof action, as well as inactive congeners to serveas negative controls, to validate known andnovel biological targets of relevance to toxicmechanisms evoked by pharmaceuticals. Suchinformation will also be of value in developingcomputer-based quantitative structure-activityrelationships for defined endpoints.

5.14 The discovery and development of new drugcandidates utilizes all the tools of basic andapplied research across the spectrum ofdisciplines engaged in the process. There is stillmore opportunity for further integration of

expertise and knowledge between scientificdisciplines, still earlier in the discovery-development process, to generate hypothesesand pose questions in a systematic fashion thatwill increase our ability to forecast and managerisk of adverse effects in patients.

5.15 Regulatory guidelines developed under theauspices of the ICH are largely based on achecklist of tests that must be completed prior toregistration of a new drug. While theseguidelines provide a good regulatory frameworkit is important they be interpreted in a mannerthat will foster innovation and meet the needsof contemporary drug safety assessmentprogrammes.

Recommendations

5.16 A new round of consultation betweenrepresentatives of the pharmaceutical industryand members of the UK and EU HealthAuthorities and Regulatory Agencies should beinstituted to review the existing ICH guidelinesand bring them up to date. In the context ofupdating current practice the guidelines should:

� promote a “hypothesis-driven” approach toall aspects of drug safety assessment testingstrategies on a “case-by case” basis

� endorse investigation of mechanisms oftoxicity in animals and adverse events inhumans; risk assessment should be evidence-based

� sanction suitable alternative study designsand approaches to data analysis in pre-clinicaland clinical development

5.17 In the context of study requirements theconsultation should re-examine the following:

� the requirements for pre-clinical testing tosupport first dose administration of drug tohumans in Experimental Medicine studies

� the requirement for 2 species in generaltoxicology studies and carcinogenicity studies,

� the need for toxicology studies of greater than90 days duration,

16


� the criteria for dose selection in generaltoxicology and carcinogenicity studies

5.18 Consensus should be sought on priorities andmechanisms for funding collaborativevalidation studies of molecular biomarkers oftoxicity

5.19 Validation of molecular biomarkers related toknown and novel pathways relevant tomechanisms of toxicity will require access toarchives of diverse chemical structures. Activeas well as inactive isomers and congeners will berequired for adequately controlled experiments.Steps should be undertaken to establish thefeasibility of obtaining access to molecules fromthe archives of pharmaceutical companieswithout breaching intellectual propertyinterests. The Molecular Libraries initiative inthe US provides access to public databases ofchemical information with which to probe adiverse range of biological systems (21).

5.20 Opportunities should continue to be sought byUK and EU based scientists to initiate andsponsor basic and applied research with a viewto developing alternative in vitro screeningassays for mechanism-based toxicity related totarget and non-target effects. While such assaysmay only play an adjunctive role in riskassessment within the next 5 years, they mayprove to be sufficiently reliable as a means ofeliminating less desirable molecules with highertoxic liabilities early in the developmentprocess.

Hazard characterization

5.21 Hazard characterization seeks to determine themode or the mechanism of toxicity caused by adrug in a test system. Most investigations oftoxicity carried out currently in drug safetyassessment are in fact investigating the moderather than the mechanism of toxicity, assessedby means of qualitative and quantitativesurrogate markers which, by weight of evidenceare consistent with a plausible mechanism.

5.22 Investigations of mechanisms of toxicity, on theother hand, aim to link specific elements in apathway to the causative stimulus that evokes

17


cell damage and tissue pathology. Theavailability of new methods to correlate toxicchanges with changes at the level of genes andproteins will enable a more precisedetermination of mechanisms of toxicity to bemade in future.


5.23 The principal objective of hazardcharacterization is to determine if the toxicityin animals is a relevant risk to humans, or toascertain if the adverse event seen in a smallsub-set of patients presents a risk to allpatients.

5.24 The decision to embark on an investigativeprogramme to determine the importance of apotential hazard to long-term patient safetyshould be governed by consideration of theperceived level of risk in relation to benefit, aswell as the level of unmet medical need, andthe availability of alternative treatment options.Not all hazards identified in animals or a sub-setof patients in the course of drug developmentmerit the same level of investment andcommitment of resources.

5.25 Investigating the mode of action related to aparticular toxicity may adopt a systems-basedapproach starting with the toxic phenotype inanimals or patients with further investigationsdirected at specific questions and hypotheses.The tools for this purpose are rooted in thebasic research laboratory, providing an almostunlimited capacity to deploy conventional andnovel methods in vitro and in vivo.

5.26 A well-structured investigative programmeaimed at gaining a better understanding of themechanistic basis of a particular hazard shouldalso aim to define biomarkers to signalimpending risk, thus adding to the quality of riskassessment and the effectiveness of riskmanagement strategies.

5.28 The concept of biomarkers is not new in animalor human medicine. Conventional laboratoryvalues signaling changes in haematology orclinical chemistry parameters outside the

normal range are biomarkers of toxicity andare key determinants of safety. There is a needto extract more measurements from theselaboratory values by closer scrutiny of minortrends or deviations in individual patient dataand relating these changes to dose, or exposure,or disease status. Sporadic, non-dose related,short-lasting deviations outside the normalrange, or dose-related trends within the normalrange of laboratory values are easilyoverlooked by statistical analysis of groupmeans. These subtle trends or deviations mayhold the key to predicting toxicity or adverseevents that, in the past, have arisenunexpectedly either in later clinical trials orafter marketing.

5.29 Although technically challenging, it is wellrecognised that collation and integration ofdisparate data sets at different levels (primaryor derivative) may reveal mutually reinforcingpatterns that point to a biological response thatotherwise would not be evident by othermethods of analysis (26). Integration ofdisparate data sets of relatively simplelaboratory and clinical data such as bodyweight, red blood cell count, serum glucose,blood pressure, heart rate, etc may revealevidence of a common pathogenesis related toincipient sub-clinical toxicity. The applicationto such data of methods developed in the fieldsof bioinformatics and data-mining may be ofvalue here.

5.30 The repertoire of non-conventional biomarkersis growing. They are capable of signalling organ specific toxicity, for example segmentalrenal tubular damage (28), isoforms ofglutathione-S-transferase (27), and KIM-1,generalized reaction to tissue injury such asfibrosis e.g. collagen type IV (29) or damage tovascular endothelium e.g. von Willebrandfactor (30) or vascular smooth muscle e.g.caveolin (31,32). It is widely recognized thatvalidation and qualification of biomarkers foruse in risk assessment will be challenging fromseveral vantage points: for example with respectto diagnostic sensitivity and specificity as well asrelevance to the mode or mechanism of actionunderlying the pathogenesis and expression of aparticular form of toxic injury (33).

Recommendations

5.31 More effective integration of disparatelaboratory data sets from pre-clinical andclinical studies is needed in order to detectsubtle signals that would otherwise be obscuredby routine statistical analysis.

5.32 Complex analytical tools ranging from‘knowledge management’ through to advanced‘data mining’ techniques should be deployed toextract maximum value embedded in dataderived from measurement of conventional andnon-conventional biomarkers of toxicity.

5.33 Individual subjects (animal and human) withsubtle trends or minor short lasting deviations inlaboratory values should be investigated withreference to temporal, quantitative andqualitative assessment of exposure responserelationships, as considered appropriate.

5.34 UK and EU based scientists should developconsensus on the requirements for validation ofbiomarkers. While the FDA has issued helpfulcomments (34) on the definitions of biomarkersfrom a technical and medical use perspective,the level of validation required to qualify abiomarker as a diagnostic aid in post marketingsurveillance of efficacy and risk has not beenconfirmed in formal regulatory guidance notes.

5.35 To utilize and interpret all data effectively in thequest to unravel complex inter-connectedcellular pathways operating at different levelswill require interdisciplinary applications ofcomputational models and a ‘systems biology’approach to confirming the mode and locus oftoxicity. Such models should ultimately acquirepredictive capacity.

Exposure assessment and dose-responseanalysis

5.36 Exposure assessment and dose-responseanalysis relate incremental changes in relevantdoses/concentrations of a drug (and/or itsmetabolites) to incremental changes in function,

18


organization, or structural integrity at the site(s)of action. The interpretation of this relationshipis key to risk assessment. The objective inhuman studies is to achieve a suitably widemargin between the maximum tolerated dose inPhase I volunteers and the projected range ofdoses required for efficacy in patients in PhaseII and Phase III clinical trials.

5.37 For drugs that induce toxicity in animal studiesthe interval expressed as a ratio of “effect-dose”to “no-effect-dose” is termed the therapeuticindex. There are variations in how this indexmay be calculated but the principle is the samein each case. With a relatively non-toxic drug itmay be possible to establish a level of exposurethat, without adverse effect, provides anadequate safety margin over exposure at thehuman therapeutic dose.


5.38 Dose-response analysis is a central plank of riskassessment for it defines the conditions ofexposure under which an adverse effect is likelyto occur. A long-standing issue of concern inrisk assessment is the practice of extrapolatingfrom high doses causing overt toxicity inanimals to lower doses in the pharmacodynamicrange, and the validity of conclusions derivedfrom this extrapolation.

5.39 If extrapolation of risk is based exclusively oneffects seen at excessively high doses there is adanger that the effect will not relate to thedrug’s primary mode of action or to relevantmechanisms of toxicity because of non-linearkinetics and saturation of metabolism. Thisshifts the mode of action to a region of the dose-response curve that bears no relevance toestimating human risk. In effect, the riskassessment is shifted from a “low-dosecategory” to a “high-dose category”. In generalthe low end of the dose response curve shouldbe regarded as more relevant to human riskassessment. Risk assessment must be based onplausible interpretation of the data otherwiseopportunities to fully characterize thetherapeutic potential of a new drug candidatemay be denied because of unfounded safetyconcerns (35).

19


5.40 Prior to the establishment of pharmacokineticsas an integral part of toxicology studiesestimation of a “safe” starting dose in humanswas based on safety factors whereby the lowestdose not causing toxicity in animal species wasdivided by 100 to allow for uncertainty inextrapolating from animals to man and fordifferences in individual sensitivity (36). Thesecalculations were usually based on the externaldose in mg/kg.

5.41 Currently, exposure is expressed as theconcentration of drug in circulating plasma overtime. Peak concentration (Cmax) and the areaunder the concentration-time curve (AUC) aremore frequently used in dose-response analysis.It is a reasonable assumption that Cmax andAUC values for unbound drug reflectconcentrations of drug at targets within, orreadily accessible to, the vascular compartment.However, they are less accurate measures ofdrug concentration at targets in deeper lessaccessible compartments, more especially whentoxicity is caused by reactive metabolites.Making the wrong assumptions whencomparing Cmax and AUC across species canlead to a miscalculation of the therapeuticindex.

5.42 It is important to select the most appropriateexposure metric for use in risk assessment on acase-by-case basis (37). Information about thetime course and mechanism of toxicity, theorgan/tissue compartment affected, the toxicmoiety (parent drug or metabolite, or reactivemetabolite), localized exposure concentrationsgoverned by membrane transport (influx orefflux), fractional extraction (liver) or excretion(kidney) of the administered dose are amongfactors that should guide selection of the mostappropriate dose metric.

5.43 Extrapolation across species or between sub-groups of individuals may be prone to error inthe absence of data to confirm a commonmechanism of toxicity and similarities inpotency of response at the target. Speciesdifferences in pharmacokinetics and/ormetabolism and differences in drug dispositionall affect the concentration of drug at the target.These underlying differences in host responseand biodisposition are among the main reason

why allometric scaling, based on a function ofbody mass and surface area, is subject tosignificant error, especially for extensivelymetabolized drugs.

5.44 Physiologically-based pharmaco/toxico-kinetic(PB-PK/TK) modeling and simulation methodsmay overcome uncertainties and inaccuraciescaused by extrapolating Cmax or AUC valuesacross or within species, or between patient sub-groups (vide infra). Iterative comparisons of PB-TK-TD relationships obtained by modelinganimal data with corresponding analyses of PB-PK relationships in patients would add weightand accuracy to inter-species and inter-groupexposure predictions.

5.45 The utility of any exposure metric in defining a“no-effect” dose for risk assessment isconditional on the mechanism underlying theeffect(s) observed being consistent with theexistence of a threshold. This presumes amechanism where there is no effect on the targetbelow a critical concentration. This highlightsthe importance of understanding the mode ormechanism of toxicity as a basis to support arational dose-response analysis.

5.46 With the exception of drugs for cancer it isstandard practice for drugs that are positive inassays for genotoxicity not to be progressed indevelopment, for it is assumed that they wouldhave the potential to be direct-acting genotoxiccarcinogens. They are believed to interact withDNA directly and effectively do not possess athreshold.

5.47 Non-genotoxic carcinogens on the other handproduce tumours through several establishedmodes of action, some related to hormonalimbalance (38), hypo- or hypermethylation ofDNA (39), or perturbation of processesregulating cell replication or apoptosis(40,41,42). These non-DNA interactive, indirectmodes of action portend a threshold. Theexistence of a threshold and a clear no-effectdose in the rodent bioassay is used as a basis forrisk assessment. In most cases the marginbetween this notional threshold in the rodentbioassay and the effective therapeutic doserange in humans is sufficient to support theconclusion that the new drug candidate will not

pose a carcinogenic risk to humans. Moreover,there are confirmed examples where the modeof action of tumour induction in rodents is notrelevant to humans (see section 11.4), or wheretherapeutic exposure levels in humans do notevoke the stimulus or response underpinningthe induction of tumours in rodents.

5.48 Since there is evidence that tumorigenesis is amulti-step process the availability of acorresponding series of biomarkers that could beused to monitor patients on a continuing basisover time would be a highly significant andpositive development, and especially valuable inconfirming the safety of new therapeutic classeswith novel mechanisms of action.

5.49 Biomarkers of host response are importantsurrogate indicators of hazard as well asexposure and there are almost limitlessopportunities for developing and deployingthem in animal studies and clinical trials.Biomarkers of host response might includeindicators of oxidative stress, tissue regenerationand repair, inflammation and fibrosis (43,44,45).Biomarkers of exposure might include urinarycatecholamines, circulating levels of steroidhormones and ACTH, and urinary metabolitesof the administered drug.

5.50 If these biomarkers can be quantifiedconsistently and demarcated in serial rank orderat intervals correlating with dose across thepharmacodynamic range then the therapeuticindex can be calculated very simply withoutrecourse to plasma drug concentrations ormodeling projections, or cross-species or sub-group extrapolations. If these relationships canbe shown to be consistent then a therapeuticindex can be derived for an individual or a groupwithin a given species. This therapeutic indexcould well be very similar numerically from onespecies to the other whereas the ratio of dose orAUC or Cmax may differ widely up or down.

5.51 Industry and regulatory scientists would readilyagree that a rational and plausible basis for theanalysis of dose-response relationships is key to asuccessful risk assessment. However, it is evidentthat the interpretation of these data may differwidely between regulatory agencies in differentparts of the world. This may mean that the ability

20


to proceed with development in one situation, orto market a drug in another, may be denied onone continent but approved on another.

Recommendations

5.52 Representatives of the UK and EU Health andregulatory authorities and the pharmaceuticalindustry should convene in scientific session toreview the strengths and weaknesses of currentapproaches to exposure assessment and dose-response analysis. Conclusions should appear inthe form of published guidance notes andprinciples. The guiding principles shouldensure that:

� the analysis is transparent, clearly articulated,and based on a composite analysis of allavailable data,

� appropriate justification for selection ofdosimetry parameters consistent with modeof action, pharmacokinetics and metabolism,

� models and modeling systems are describedand properly justified,

� biomarkers of toxicity, host response andexposure are used appropriately tocharacterise dose-response relationships fullyacross the dose range, with particular attentionto relationships within the pharmacodynamicrange, and,

� estimates of potential error and uncertainty inthe analysis are presented.

5.53 A second output of the review should be thepublication of non-competitive data in the formof case studies illustrating the best examples ofdose-response analyses and risk assessment withthe objective of sharing knowledge andpromoting best practice.

5.54 A concerted research effort is required todiscover the basis for constitutive differencesbetween animal species, between animal speciesand humans, and between subpopulations ofhumans, with particular reference to genotype,phenotype, organ physiology andpharmacology, and processes affecting bio-distribution of drug in organs and tissues, as theyrelate to mechanisms of drug toxicity.

21


Predicting pharmacokinetics,metabolism, and drug-drug interactions

5.56 The expression of toxicity both in animals andhumans depends on the concentration of thetoxic moiety at the target, the affinity andkinetics of the interaction between the toxicmoiety and the target, and the ability of the hostto adapt and repair the damage caused. Thesecomplex concentration-time relationshipsbetween host and drug govern ultimatelywhether the dose administered is safe orharmful. Clearly, information that predictswhen these complex interactions might co-existmaterially and temporally to cause tissuedamage would be of immense value inpredicting risk, managing risk and reinforcingpatient safety (46).


5.57 Early in the development of a new candidatedrug, information is sought to determine if it willbe adequately absorbed in animals and humans,whether it will exhibit desirable pharmacokineticand metabolic properties, and whether it islikely to interact with other medicationsadministered concurrently in a manner thatcould compromise patient safety. It is nowpossible to obtain much of this qualitative andquantitative information directly using human-derived in vitro systems (whole cells, sub-cellularfractions, or expressed enzymes) to help predictthe in vivo metabolism and kinetics of a drugcandidate, as well as providing information onvariability between individuals that may lead toadverse drug-drug interactions (46, 47).

5.58 The absorption, distribution, metabolism andelimination (ADME) characteristics of a newdrug candidate are similarly studied in vitro andin vivo in animal species, including those used intoxicology studies. This information is key tooptimizing the design of animal toxicity studieswith respect to species selection, dose andformulation, dosing regimen and samplingintervals, as well as highlighting potential targetorgans of toxicity. Analysis of ADME data is anintegral part of interpreting the findings inanimal toxicity studies and is an essential part ofcharacterizing risk for humans.

5.59 As the liver is the major site of xenobioticmetabolism in mammals, particular attentionhas been paid to the development of in vitrohepatic models. For example, advances inmolecular biology techniques have led to thedevelopment of expression systems to permit invitro metabolism and interaction studies withspecific hepatic xenobiotic metabolisingenzymes, including P450 (CYP), flavin-containing monooxygenase (FMO),UDPglucuronosyl-transferase (UGT) andsulphotransferase (SULT) forms (47)

5.60 Other heterologous expression systems havebeen developed to permit studies with variousreceptors and transport proteins (48). Extra-hepatic models include intestinal cell lines (e.g.Caco-2) to investigate absorption from the gut.

5.61 Another important and continuingdevelopment is the application of in silico (i.e.computer based) techniques to all areas of drugdiscovery and development (49). For example,in silico methods have been developed to helpdesign libraries of compounds with appropriatechemical, physical, pharmacokinetic andbiological properties.

5.62 Predicting kinetics. In silico methods,artificial membrane systems and Caco-2 humanintestinal cell lines have been developed toevaluate intestinal permeability and metabolism.Heterologous expression systems containingtransporter proteins (e.g. P-glycoprotein andorganic anion transporter proteins) and highthroughput procedures for measuring proteinbinding have also been developed to evaluateabsorption and tissue distribution of orallyadministered drug candidates.

5.63 Assessment of bioavailability (the fraction of anoral dose entering the systemic circulation fromthe gut), intrinsic clearance and metabolicstability of a drug candidate is carried out inassays conducted in liver microsomes, othersub-cellular fractions, intact cell systems(hepatocytes and liver slices) and expressedenzymes in conjunction with sensitive analyticaltechniques (47).

5.64 Predicting metabolic fate In silico models ofmetabolism, cultured hepatocytes or liver slices,and expressed enzymes are used to predict the

extent of Phase I (oxidation) or Phase II(conjugation) metabolism; to identify pathwaysof metabolism; and, to characterize the productsof liver metabolism. Using appropriate scalingfactors a suitable liver model is applied toestimate intrinsic clearance in vivo.

5.65 Information regarding specific enzymesresponsible for biotransformation of the drugcandidate (generally CYP forms) is obtained byreaction phenotyping studies employingcorrelation analysis, studies with expressedCYP forms, chemical inhibition and/orinhibitory antibodies (46, 47). This informationis important for predicting drug-druginteraction and for identifying drug candidatesthat are metabolized by enzymes known toexhibit genetic polymorphisms or other inter-individual variability in humans. Other non-CYP dependent enzyme activities may bestudied in similar fashion using various liverpreparations, or preparations from extra-hepatic tissues.

5.66 Predicting variability/polymorphisms.Genetic polymorphisms, or multiple genecopies, or differences in allelic frequencies incertain drug metabolizing genotypes mayaccount for some individuals being classified asultra-rapid, extensive or poor metabolisers ofdrugs or chemicals. These variations may leadto adverse effects or lack of efficacy. Thesedifferences may also account for differencesbetween ethnic groups. Differences in the levelsof expression of drug metabolizing enzymesmay also account for variability between youngand old, and between adult and fetus.Variability between populations of humans canbe assessed by using well characterised liverpreparations from a wide range of donors andfrom knowledge of the enzymes involved in themetabolism of the drug candidate (47).

5.67 Predicting drug-drug interactions.Potentially harmful drug-drug interactions canbe caused by induction but more frequently arecaused by inhibition of metabolizing enzymes(47, 50). A combination of in silico models, highthroughput screens and liver microsomes canbe used to identify specific CYP forms that maybe inhibited by a new drug candidate. Thenature of the inhibition, the Ki value and the

22


concentration causing 50% inhibition (IC50)can be determined. These data, coupled with invivo data, enable the risk of harmful drug-druginteraction to be assessed.

5.68 Hepatocyte cultures, or liver slices can be usedto confirm induction of CYP forms while cell-based reporter gene constructs can be used toscreen chemicals for such activity. Forexample, high-throughput pregnane-X-receptor ligand binding and activation assayshave been developed; this receptor is known toregulate CYP3A4. Cultured hepatocytes canalso be used to assess the effects of a drugcandidate on other metabolizing enzymes suchas UGT forms.

5.69 Besides effects on CYP forms in the liver, drug-drug interactions have also been describedinvolving both inhibition and induction oftransporter proteins (48).

5.70 Accurate and useful prediction of the kinetics ofdrugs and drug-drug interactions in vivo prior tothe efficient selection and design of confirmatorystudies in humans will rely increasingly on invitro data generated with human-based material(tissues, cells, sub-cellular fractions, expressedenzymes/transporters).

5.71 Provided that in vitro information is of highquality and is obtained at the appropriate timein the drug development process, itsincorporation into sophisticatedphysiologically-based population PK models(that incorporate patient/diseasedemographics, physiological, pathological,genetic and developmental variation) shouldfacilitate effective communication between latepre-clinical and early clinical development (51).

5.72 In addition, at earlier stages of developmentand in drug discovery, confidence in the useof purely in silico methods to predict ADMEproperties (active site modelling, QSARbased on physicochemical and structuralproperties) of both real and virtualcompounds should increase as moreexperience is gained. In turn, this should helpthe early differentiation of potentialdrugs from non-starter compounds likely tofail later on.

23


Recommendations

5.73 A number of priorities can be identified tosupport and extend the approach to predictingthe kinetics of drugs and drug-drug interactionsin humans:

� Maintenance of the supply of human tissuesamples while encouraging the developmentof techniques to use them more efficiently(cryopreservation, immortalisation of celllines, stem cell research).

� Standardisation of in vitro absorption modelsthat address both passive and activeprocesses.

� Standardisation of in vitro Blood-Brain-Barrier models that also mimic pathologicaleffects.

� More investment in in vivo studies of the roleof transporters in drug disposition to put invitro information into perspective and to helpto construct rational in vitro transporterscreens.

� Development of standardised in vitro screensfor enzyme induction and inhibition.

� Integration of tissue/cellular systems forstudying ADME with the evaluation ofcellular toxicology.

5.74 Some outstanding issues and needs withregard to in vitro/in silico – in vivoextrapolation include the following:

� Expanded data bases on demographics,disease prevalence, genetic variants, enzyme/transporter abundancies and their variances.

� Improved understanding of the interplaybetween enzymes and transporters and ofnon-CYP drug metabolism, to extendexisting algorithms and models.

� Better communication between largedatabases and predictive programmes.

� Improved collaboration between and withinacademia, industry and regulatory authoritiesin sharing databases and experience on invitro-in vivo extrapolation, PK-PD modellingand clinical trials simulation.

� Rationalisation of IP issues related to access toalgorithms, data and databases, withoutcompromise of confidentiality.

� Commitment to dedicate personnel(realignment) to carry out retrospective andprospective evaluation of algorithms andmodels.

� Commitment to training of skilled modellers.Industry needs to be more pro-active insupporting academia; universities need toupgrade their appreciation of the economicimportance of drug development and drugand environmental toxicology (safetysciences).

� Redistribution of expenditure in drugdevelopment such that frontloading with highquality in vitro data becomes more common,the appropriate information is obtained at theright time and costly clinical studies becomemore ‘confirm’ rather than ‘learn’.

� Linkage of predictive physiologically-basedpopulation PK models to mechanistic PDmodels to put PK variability into clinicalperspective.

� Linkage of predictive physiologically-basedpopulation PK models to data on cellular PKand toxicology and complex mechanisticbiological models – as part of a systems biologyapproach to the assessment of drug toxicity.

Risk characterization

5.75 Risk characterization is a key step in the overallanalysis of risk. The process involves theintegration of data drawn from studiesconducted by departments and disciplines thatcomprise a modern R&D organization. Theprocess utilizes the skills and knowledge of thespecialist as well as the generalist.

5.76 There are, on occasions, organizational andphilosophical barriers that may impact the qualityof the evidence gathered and the decisionsemerging from the process of riskcharacterization. Successful organizations arethose that promote interdisciplinarycommunication and collaboration with equalregard for the technical, intellectual andprofessional value of each contributing discipline.


5.77 Fundamental to risk characterization is theintegration of data and information arising fromthree key activities, namely, hazard identificationand the determination of the concentration of allchemical entities in test systems and theirrelevance to the clinical situation, dose-responseanalysis, and hazard characterization. Underideal circumstances the assessment of risk inrelation to benefit should adopt a weight ofevidence approach. In other words,extrapolation of data across systems or species, orwithin sub-groups of the same species, shouldalways relate the mechanism or mode of toxicityto relevant indices of exposure in order to ensurerelevant comparisons are made to support adecision on risk versus benefit.

5.78 Such a balanced weight of evidence approach isnot always adopted uniformly and consistentlyby regulatory agencies. Difficulties arise whenthe evidence is judged to be insufficient or themethodology flawed, when regulatory policydefaults to attempting to eliminate all possiblerisk, or when there is failure to agree that thebalance between risk and benefit for a givenindication is not consistent with patient safety.Given the diversity of drugs and indicationsunder development it is evident that regulatorydecisions should be undertaken on a case-by-case basis. There is a need for greatertransparency and consistency of practice, bothon the part of the pharmaceutical industry andthe regulatory agencies, to ensure regulatorydecisions are based on sound evidence andrational interpretation.

5.79 It is important to recognize that individualknowledge, institutional memory, clinicalexperience, and knowledge of therapeutics in aparticular indication or disease are paramount inassigning priority and importance to risk factors,and that these must be considered separately foreach indication, patient population, and dosage.It is generally agreed that there are no validatedmodels with which to derive objective andquantitative assessments of risk versus benefitthat apply universally to all drugs (52). It isvirtually impossible for a model to accommodateall the necessary criteria for risk or benefit.Therefore, a case-by-case assessment is clearlythe preferred approach.

24


5.80 Mussen (52) surveyed the European PublicAssessment Reports (EPARs) for 33 new drugs,in categories ranging from cancer therapy toViagra, approved between 1998 and 2000;EPARs summarise drug efficacy and safety andreflect the basis and the grounds for grantingregulatory approval to market a new drug.From this limited study Mussen concluded thatthere was more overall emphasis on efficacythan on safety in the concluding sections ofEPARs. He also noted that certain risk factorssuch as safety in sub-groups, drug-druginteractions, and the generalisability of thesafety profile were not discussed extensively inthe majority of cases. Possible reasons citedwere a lack of guidance in the preparation ofrisk-benefit conclusions, and the difficulty ofwriting a summary to meet the needs of adiverse audience of patients and specialists.

5.81 While it is important to preserve operationalflexibility in the process of scientific enquiryleading to risk characterization for a new drugcandidate, formal guidance on the essentialelements of the process is lacking at present.Similarly there is inadequate education andtraining in the basic scientific elements andprinciples that underpin the process

Recommendations

5.82 Representatives of the UK and EU Health andregulatory authorities and the pharmaceuticalindustry should convene to review the strengthsand weaknesses of current approaches to riskcharacterization. The results of this consultationshould be published guidance notes andprinciples to promote consistency of processand quality of output. The guiding principlesshould ensure that:

� the chosen methodology and the logicsupporting conclusions underpinning the riskcharacterization for the drug under reviewshould be transparent, clearly articulated andinclusive of all relevant data,

� primary as well as secondary risk factorsshould be identified and prioritisedaccording to disease severity, patientattributes, dose regimen, and conditions ofuse that may impact the safety of the drug ormarketed product,

25


� a weight-of-evidence, mechanism-basedapproach to risk characterization should beendorsed thus avoiding regulatory decisionsdefaulting to a ‘de minimis’ position; the useof a ‘framework’ approach to determiningweight of evidence should be explored,

� estimates of potential error and uncertainty inthe analysis and conclusions should bepresented, and,

� summaries of risk-benefit positions and thebasis of decisions rendered by regulatoryauthorities should be published in the openliterature when a new drug is approved formarketing; this information should providethe basis for creating communiqués topatients and health care professionals.

5.83 There is a need to consolidate expertise and tofocus activities within the UK in order toadvance the science and processes connectedwith risk characterization and risk assessment.This could be achieved by creating aninfrastructure to function as a National Centre ofSafety Assessment:

� the establishment of a National Centre ofSafety Assessment would provide focus andleadership to all efforts aimed at improving thescientific and regulatory aspects of processesthat underpin risk characterization, riskassessment and risk management in the UK.

� a National Centre of Safety Assessmentwould be responsible for development anddissemination of ‘best practice’ in approachesto risk characterization and risk assessment of drugs.

� a National Centre of Safety Assessment wouldfacilitate access to consultation on scientificand regulatory issues related to riskcharacterization and risk assessment, both ingeneral terms and on a case-by-case basis.

� the Centre would be well placed to overseethe development and implementation ofeducational and training programmesdesigned to strengthen interdisciplinary skillsin risk characterization and drug safetyassessment.

� a National Centre of Safety Assessment wouldfoster a culture of “One Medicine” wherein

the focus of all disciplines engaged in drugsafety assessment is on promoting anunbroken line of continuous scientific enquiryfrom discovery to successful approval andmarketing of a new medicine.

False positives and false negatives:implications for risk assessment

5.84 Studies to assess drug safety are conducted incompliance with rigorously controlledprotocols in accordance with regulatoryguidelines and the various codes of GoodLaboratory Practice, Good Clinical Practice andGood Manufacturing Practice. However, evenwell controlled experimental systems are liableto produce false positive or false negative resultsthat lead to misleading conclusions, so calledtype I or type II errors in statistics. In thecontext of predicting risk, false positive results“over-predict” the likelihood of adverse eventswhereas false negative results “under-predict”such outcomes. Cell-based assays, as well asstudies in animals and humans are liable tothese shortcomings, especially when data areextrapolated from one test system or species toanother.

5.85 Despite these well-recognized shortcomings,animal models are an indispensable resourcewith which to study the pathogenesis of humandisease and to investigate the efficacy and safetyof new drug candidates. Given our knowledgeof species diversity it is not surprising that theyare not perfect replicas able to predict everypotential outcome in humans. However,without insights and clues provided by animalmodels few of the major treatmentbreakthroughs of the past 50 years would havebeen possible.

Moreover, there has not been an epidemic of adverseevents accompanying the introduction of newprescription drugs during this period. As knowledge ofthe underlying biology of these systems increases,there is greater understanding of the basis for the falsepositives and the false negatives, thereby providinggreater insight into the true magnitude of any risk.

26


6.1 Although useful perspectives have beengleaned from retrospective reviews of data incompany files and regulatory submissions, thelatter available through freedom of information,there has not yet been a definitive prospectivesurvey to critically assess concordance oftoxicity seen in animal studies with observationsof adverse events in humans.


6.2 Data obtained from animal studies have thegreatest impact on drug safety prior toadministering the first dose to man. Clinicaltrials in Phase I, II, and III, are the main sourcesof data that govern decisions regarding drugsafety. Newly emergent pathology in chronicanimal studies (6 months or longer), not seen in30-day studies, is difficult to interpret but may incertain circumstances signal a risk to humans.This pathology may reflect tissue regenerationand repair (fibrosis and/or hyperplasia) inresponse to ongoing, drug-related, low-gradetissue damage, or may be related to spontaneousage-related changes unique to the species butunrelated to a direct effect of the drug.

6.3 In some cases drug related toxicity mayaccelerate progression of spontaneous age-related disease in laboratory animals. It is oftennecessary to investigate the basis of thesechanges if there is suspicion they maynegatively influence the balance of risk versusbenefit for humans. It is important therefore todistinguish laboratory animal specific changes(false positives) from changes that may signal arisk to humans (true positive).

6.4 It has been estimated that from 10 to 14% of newdrug candidates fail during the pre-clinicalphase of testing usually because of toxicity inanimals (53). Reasons for withdrawal fromdevelopment are usually related to severalfactors, namely:

� severe toxicity affecting the CNS, specialsense organs or the haemopoietic system,

� toxicity of major organs without adequatesafety margins,

� toxicity related to the drug candidate’sprimary mode of action without adequateseparation of the efficacious and toxicexposures and,

� toxicity whose mechanism is not known.

6.5 Because it is not considered safe to progressthese drug candidates into Phase I studies it isnot possible to determine if these findings wouldhave been reproduced in humans or whetherthey were indeed false positive observations.

6.6 It is well recognized that animal studies havebeen a major factor in ensuring that, of manythousands of drug candidates tested in Phase Istudies, only very few have resulted in seriousadverse events in human volunteers, and onlyrare cases of death have occurred (54). With arelatively low withdrawal rate of about 10% ofdrug candidates from later clinical trials beingattributed to adverse events, many related toeffects already seen in animals, it appears thatanimal studies make an important contributionto the safety of patients in clinical trials (55). Inother words the incidence of false negativeoutcomes in animal studies is comfortingly low.

6.7 The results of several published surveys havedemonstrated good concordance between thefinding of toxicity in animals and theoccurrence of similar or complimentary adversefindings in humans (55,56). ILSI/HESIsponsored a retrospective study (56), based ondata provided by 12 pharmaceutical companies,to assess the rate of concordance betweentoxicity in animal studies and adverse events insubsequent clinical trials. The overallconcordance rate was 71% when comparingdata from rodent and non-rodent species in theaggregate for any given drug. Importantly, 94%of the toxicities in animals occurred in studies of30 days or less. Drugs for cancer, viral infectionsand cardiovascular disease showed 80%concordance whereas drugs for endocrineconditions achieved only a 50% concordance

Chapter six - Concordance of toxicity in animals and adverseevents in humans

27


between animal and human data. The humanadverse events recorded for the purpose of thissurvey were generally severe. It appears themajority of non-concordant outcomes were notrelated to species differences in drugmetabolism or pharmacology between animalsand humans

6.8 Concordance was greatest for toxicities of thehaemopoietic system (80–90%) and least forskin conditions (38%) and abnormalities of liverfunction (41%). It is estimated that 50% ofhuman adverse reports are related to subjectiveclinical symptoms (headache, dizziness,anxiety, abdominal pain, nausea, myalgia,lethargy etc). There are no satisfactory meansby which to evaluate these effects objectively inanimals (57). The need for improved modelswith which to evaluate functional deficits in thecentral and peripheral nervous systems, and toassess cognitive function in animals, isaddressed in the report of the SafetyPharmacology Working Group.

6.9 A more critical analysis of concordancebetween observations in animals and adverseevents in humans should be possible wherecriteria for concordance or lack of concordanceare agreed in advance. At a minimum it isessential to consider the level of parity betweenthe mode of action of a drug and its associated

effects in animals relative to humans so thatcomparisons are not skewed because ofdifferences in exposures, access to target organsor receptors, genetic predispositions orenvironmental factors, frequency and timing ofdosing, or other constitutive differences thatcould jeopardize the validity of the comparativeanalysis. A protocol incorporating these criteriafor assessment of concordance is being pursuedin a prospectively designed study sponsored bythe ILSI/HESI organization.

Recommendations

6.10 Ensure that pre-defined criteria forinclusion/exclusion of data are established inadvance for all surveys of concordance betweenobservations of toxicity in animals and humans:the ongoing ILSI/HESI study of concordanceappears to have agreed such guidelines inadvance.

6.11 Extend the prospective ILSI/HESIconcordance study to include compilation ofdata from 6-months and 1-year toxicologystudies in the files of UK and EU Health andregulatory authorities to determine the extent towhich data from these studies have impacteddecisions affecting human risk and safety in therecent past.

28


7.1 Toxicity seen only in one or other animalspecies and seemingly without counterpart inhumans is categorized as a “species-specific”phenomenon, but the strength of evidence toconfirm or refute this conclusion variesconsiderably from one situation to another. Thisuncertainty frequently results in testing a newdrug candidate in additional species or strains ofanimal in an attempt to determine thesignificance to human risk. However, over thelast decade mode of action studies haveconfirmed that certain tumours in rodents areboth strain- and species-specific (see section11.4) and do not signal a risk to humans.

7.2 Individual animals in a study, or individualpatients in a clinical trial, or a post-marketingstudy of safety or efficacy, may show a transienteffect or a trend over time suggesting a possibledrug association. Such effects generally occur atvery low incidence, are frequently not dose-related, may not deviate outside the normalrange and, therefore, are not statisticallysignificant. This pattern of response is ofteninsufficient to conclude definitively that there isa drug-related effect.


7.3 Despite intensive investigation there are relativelyfew examples of species-specific toxicity that havebeen fully elucidated with reference to mode ormechanism of action. Most examples are non-genotoxic rodent carcinogens (58,59). Severalcardio-active and vaso-active drugs with differentmodes of action cause arterial toxicity in differentvascular beds in different animal species (60,61).There is often a divergence of opinion regardingthe relevance of these findings both for volunteerand patient safety, both from first exposure as wellas the concern about exacerbating coincidentvascular disease.

7.4 There is a paucity of knowledge regardingconstitutive differences between animal speciesthat may explain why some are highlyresponsive to certain classes of drug while others

are more or less refractory. Some of thesedifferences may be related to a lower level oftarget expression in one species versus another,pharmacokinetic or metabolic differences, orphysiologic or genetic differences inherent to thespecies. There is a need to characterizepolymorphic and allelic variations in the genomeof animal species showing marked differences insusceptibility to toxicity. Knowledge of the basisfor such differences would clarify theirsignificance to human risk assessment.

7.5 Human specific toxicity, in particular drug-induced hypersensitivity reactions have beenknown to occur at very low incidence withoutcounterpart in animal studies. These so calledidiosyncratic adverse reactions have beeninvestigated in patients but often without asuitable animal model with which to elucidatethe pathogenesis. There are already exampleswhere low incidence adverse events in clinicaltrials have been subjected to pharmacogeneticanalysis to map the genetic susceptibility ofaffected patients, and specific polymorphismshave been identified that relate to the functionaldeficit or adverse event. In the case of Tranilast(treatment for restenosis) a polymorphism in theUGT1A1 genotype (UDPglucuronosyltransferase 1, polypeptide A1) wasassociated with hyperbilirubinaemia in 40% ofaffected patients (62). In the case of Abacavir(anti HIV treatment) polymorphisms in twogenes (HLA –B57 and TNFalpha –238) bothlocated on chromosome 6 were found to behighly associated with a low incidencehypersensitivity reaction to the drug (63, 64).Further experience will demonstrate if it will bepossible to repeat this success as polymorphicdeterminants of variable response for a largernumber and range of novel drug targets aremapped in greater detail.

Recommendations

7.6 UK and EU based scientists in academia andindustry should establish a programme ofcollaborative research funded, at least in part,

Chapter seven - Species-specific toxicity and inter-individualvariation

29


by the pharmaceutical industry to explorephenotypic, genotypic and other constitutivedifferences that may account for manifestationsof species-specific toxicity. The objectives of thecollaboration would be:

� to discover a rational basis for speciesdifferences in toxic responses to certainclasses of drugs or modes of action thatcurrently are either denied regulatoryapproval to proceed into human trials or areconsidered a high risk for human safety,

� to characterize genetic polymorphisms andallelic variations in toxicology species thatmay be relevant to assessing the toxicpotential of specific drug classes andtherapeutic disease targets in general e.g.

nuclear or cytoplasmic receptors; such arepository of data could support selection ofone rather than two species for toxicologystudies,

� to accumulate a repository of phenotypic andgenotypic data from animals exhibitingatypical, non-dose related responses tovarious classes of drugs in toxicology studiesas appropriate; such data may revealpolymorphic or allelic variations in certainanimals that may have specific utility assentinels of human risk, and,

� to increase knowledge of species differencesin organ pharmacology and physiologythat would contribute to more accuratephysiologically-based pharmacokineticmodelling.

30


necessary to saturate liver clearancemechanisms in order to achieve systemicexposures within range of therapeutic exposuresin humans.

8.5 The regulatory requirement that the top dose oflow-toxicity drugs in a rodent bioassay shouldachieve a 25 fold multiple of the human AUC isimpractical in the majority of cases especiallywith relatively low toxicity drugs. High residuesof drug in the gut or high concentrationseliminated in bile or urine can cause pathologythat is secondary to overdosing.

8.6 The current method for identifying themaximum tolerated dose, although mandatedby current regulatory guidelines, should bediscouraged as it frequently results in secondarypathology not related to the primary mode ofaction of the drug. Pharmacokinetics andpharmacodynamic responses of the test speciesas well as consideration of human exposure todrug at therapeutic doses should be factored into dose selection in animal studies. (37).

8.7 Observations of toxicity in animals at themaximum tolerated dose elicit considerablevariation in the actions taken by regulatoryagencies in different countries, especially withregard to the scale and scope of investigationsrequired to absolve the drug from a highlynegative risk benefit assessment. This can resultin failure to progress a drug in developmenteither because of restrictions on the range ofdoses that can be tested or because of patientexclusions.

Recommendations

8.8 Representatives of the UK and EU Health andregulatory authorities and the pharmaceuticalindustry should convene to review regulatorycustom and practice over the last decade with aview to revising the requirements and guidelinesfor dose selection in toxicology studies andminimizing use of the current method foridentifying the maximum tolerated dose.

8.1 Dose selection is one of the critical elementsgoverning the successful completion of animalstudies and clinical trials. The exposure (Cmaxor AUC) produced by a toxic dose in animalsmay be greater or less than the exposureproduced by a therapeutic dose in patients. It isimportant that species differences in sensitivityto the primary and secondary actions of thedrug or to chemically mediated toxicity betaken into account, for the therapeutic indexmay be high in the animal species even thoughblood concentrations are less than in humans.Excessively high doses should not be used inanimals in an attempt to achieve a ‘safe’multiple of animal relative to human exposure.

8.2 Thus the requirement by regulatory guidelinesto use a maximally tolerated dose in animalstudies frequently results in toxicity at levels ofexposure that may be greater than or less thanexposures produced by doses in the humantherapeutic range. This may raise unfoundedconcerns about drug safety.


8.3 Insufficient attention is given to choosingdoses over the lower end of the dose range inanimal studies. This is the area of the doseresponse curve most relevant to human safetyand a greater emphasis should be placed inanimal studies to quantify drug-related effectsat or slightly above pharmacologically activedoses (37).

8.4 Selection of the uppermost dose in animalstudies is important otherwise there is a dangerof producing false positive or false negativeresults. Administration of oral doses that saturatedrug metabolism, protein binding or excretionmay cause false positive safety concerns. Suchexcessive doses can result in non-linear kineticsor produce toxic metabolites, neither of which isnecessarily relevant to human exposure withinthe therapeutic range. However, in rodents withhigh first pass clearance of orally absorbed drugfrom the intestine by the liver, it is sometimes

Chapter eight - Dose selection

31


9.1 Adverse drug events are defined as unintendedand undesirable effects of a drug that occur atdoses used in humans for prophylaxis, diagnosisor therapy. In the context used hereidiosyncratic adverse events are unexpectedand occur predominantly after approval of thetherapeutic or diagnostic product. Implicit is theassumption of failure, despite extensiveinvestigations, to predict that such events mightindeed occur in patients receiving the drug (65).

9.2 Unexpected adverse events related toexaggerated pharmacology, or interaction ofthe therapeutic chemical or protein with cellularmacromolecules, are typically rare, althoughthey may occur from time to time because ofunanticipated system failure. In other words,most are predicted in animal studies, studies invitro, or in patients in clinical trials.

9.3 A further category of unexpected adverse event,designated “idiosyncratic,” refers to drug-related reactions that appear to be hostdependent and typically occur at a frequency of1:1000 to 1:100,000 patients. They are usuallywithout precedent, either in animal studies orclinical trials, and the mechanism of thereaction, and the reasons for inter-individualdifferences in susceptibility, are poorlyunderstood.

9.4 Idiosyncratic adverse events present a diverseclinical picture. They have been associated withall systems in the body – the liver, skin, kidney,haemopoietic system, and the immune systembeing most frequently affected. The same drugmay be associated with a different set ofsymptoms in different patients. In cases of liverinjury, symptoms range from mild,asymptomatic changes in serum transaminasesto life-threatening liver failure.

9.5 It is often assumed, but without proof, thatthese idiosyncratic adverse reactions havean immunological basis. Certainly, a secondexposure may be more severe and cause death.

9.6 All classes of drugs with diverse mechanisms ofaction have been found to cause idiosyncratic

adverse events. Generally speaking suchidiosyncratic reactions are not observed fordrugs given at doses less than 10 mg per day,indicating that some form of chemical stress isinvolved in the pathogenesis.


9.7 Experience so far would suggest that there areno validated standardized tests either in vitro orin vivo, of animal or human origin, with which topredict idiosyncratic drug toxicity. At presentmost idiosyncratic adverse events are detectedonly after approval to market a drug, and onlyafter very large numbers of patients, muchgreater than in clinical trials, have been exposedto the drug.

9.8 The challenge is the prediction of idiosyncraticdrug reactions in terms of 1) the chemistry of thedrug and, 2) the genotype or phenotype of theindividual patient in a clinical trial. This begsthe question as to when in the course of drugdevelopment is it possible to predict the risk ofhost-dependent drug toxicity. Intuitively onewould expect a test system to contain a geneproduct that is a determinant of hostsusceptibility to this kind of low frequencyevent. In the first instance it requires a detailedunderstanding of the cellular and molecularmechanisms involved. It is important todistinguish between those studies that aim todefine the pathogenesis of the reaction andthose whose goal is to develop empirical testsystems.

9.9 The investigation of idiosyncratic adverse drugreactions to date has largely been undertakenby a small number of academic groups. Newprocesses are required that will supportcoordination and collaboration betweenindustry, medical practitioners, regulatoryauthorities and academia in order to achieve thefollowing outcomes.

9.10 Understanding generic and compoundspecific mechanisms: although the detailed

Chapter nine - Idiosyncratic adverse events

molecular events resulting in idiosyncraticadverse reactions vary between drugs there mayexist common aspects of the relevant pathogenicmechanisms that could be identified on the basisof detailed cellular and molecular investigations.New experimental in vitro and in vivo systemsare required for mechanistic studies. Thedevelopment of ex vivo systems incorporatingaffected patients’ cells and transgenic animalsprovide promising approaches.

9.11 Availability of biological samples forresearch: there is a need to establish acoordinated approach to identifying patients,recording and characterizing idiosyncratic drugreactions, and making available patient dataand clinical samples that will facilitateexpedited investigations of mechanisms ofaction. Funding is required to initiate andfacilitate data and sample collection. Ethicalissues need to he identified, addressed, andresolved.

9.12 Availability of clinical and experimentalexpertise: serious adverse drug reactions leadto immediate discontinuation of drug treatmentand may lead to withdrawal of the drugaltogether. If investigations to study mechanismsare to be successful they must start without delay.This calls for a high level of efficiency andeffectiveness so that timely management of suchevents can be introduced.

9.13 Establishment of a national centre forresearch and information: consolidationof clinical and experimental capabilities andcoordination of operations within a singlecenter of expertise would greatly facilitate basicresearch into this important area of publichealth. Such a centre would encourage a greaterunderstanding of idiosyncratic adverse healtheffects associated with drug exposure as well asserve as a central repository for data, datamanagement and clinical samples. A NationalCentre for Safety Assessment could serve thisfunction (see above).

9.14 The centre would provide both clinical andlaboratory based education in the field andwould resource a teaching programme inmolecular pharmacology and toxicology alliedto human medicine that could be incorporatedinto the undergraduate medical curriculum.

32


9.15 Application of new technologies: there isa need to develop further those modernpharmacogenomic and toxicogenomicplatforms to facilitate a detailed understandingof the heritable and environmental factors thatpredispose to adverse drug reactions. The ‘omictechnologies are powerful tools but they requirespecific fine-tuning for application to particularhuman health problems, especially those arisingfrom, or associated with, idiosyncratic adversedrug reactions.

9.16 Patient assessment: there is a need tounderstand in greater detail the ways in whichacquired, environmental, and dietary factorsmay act, either alone or in concert, withheritable factors to determine inter-individualdifferences in susceptibility to adverse drugreactions. Thus techniques are required forsimultaneous investigation of patient phenotypeand genotype. It is well recognized that theseadvances are an essential step to achieving thegoals of personalized medicines in clinicalpractice.

Recommendations

9.17 To facilitate progress in these directions, there isa need to determine which drug toxicities werefailures of the system and which drug toxicitiescould not have been avoided with all presentknowledge available. This retrospectiveanalysis should guide investment of time andresources into areas of research most likely toenable us in future to avoid idiosyncraticadverse events.

9.18 There is also a need for clinicians, regulatorsand the general public to develop consensusregarding what level of risk is acceptable, andunder what circumstances, especiallyconsidering the benefits offered, even bystigmatized drugs, to the majority of patients. Itis important not to raise false hopes for, giventhe diversity of causality and the large numbersof patients who elect to take drugs for theirdiseases, there will always be a degree ofunavoidable risk.

9.19 There is a need for a change of mindset in orderto appreciate that idiosyncratic drug toxicity is a

33


general health problem rather than an obscuremedical problem caused occasionally by a drug,with all the responsibility to resolve andunderstand all its dimensions falling at the doorof the pharmaceutical industry.

9.20 To understand and prevent such idiosyncraticreactions fundamental research must have aclinical focus, which requires sufficientexperimental material, supported by the latesttechnologies for understanding drug action. Atpresent such funding is largely directed towardspreclinical research.

9.21 National and international databases, andtissue, serum and cell banks are required toprovide sufficient clinical material for research.The clinical picture with any adverse drugreaction is variable. Therefore, for idiosyncraticdrug toxicity, typically of extremely lowincidence, we are faced with the somewhatparadoxical situation that large numbers ofpatients are required who are accuratelycharacterised clinically, and stratified bybioinformatics.

9.22 In the light of these considerations the keyrecommendations are:

� to initiate develop and coordinate clinical andexperimental investigations of adverse drugreactions,

� to provide a vehicle for collaborations andsharing of samples as well as information withother centres in the US and Europe,

� to initiate a coordinated effort to be fundedjointly by the pharmaceutical industry, theMedical Research Council, the Departmentof Health, NPSA, and the National HealthService.

34


10.1 The principle features, and assumptions, of therodent bioassay for carcinogenicity weredetermined in the 1940’s and 50’s (66,67,68).The process was developed with only limitedmechanistic understanding of carcinogenesis,but with the knowledge that exposure to anumber of chemicals was associated with thedevelopment of cancers in man andexperimental animals. Analysis of these dataled to the belief that the “majority of all cancer”is caused by chemical or environmental factors(69,70). However, it should be noted that at thistime there was a tendency to assume thatenvironmental (which simply meant that theetiological factor was extrinsic) meantchemical. The logical conclusion from theseassumptions is therefore self evident: identifycarcinogenic potential and cancer avoidancewould be possible (70). As a consequenceconsiderable impetus developed to screenchemicals for carcinogenic potential. This isnow governed by extensive regulatoryrequirements for food additives, environmentalchemicals, agrochemicals and pharmaceuticals.

10.20 Inherent in the use of animals forcarcinogenicity bioassay is the assumption thathumans and animals behave in a similar way.In addition, two experimental concepts formthe scientific basis on which the animalcarcinogenicity bioassay is based.

10.21 The first is the empirical relationshipdeveloped by Druckrey (71):

Tumour incidencea dtn

Where ‘d’ is dose, ‘t’ is time to tumourincidence, and ‘n’ is a power term, usually 2, or3 or even higher.

10.22 The experimental work (mostly in skin usingpolycyclic aromatic hydrocarbons) which ledto this relationship, indicated that time was themost important parameter in determiningincidence, and that the tumour incidence wasdirectly proportional to dose. Thus, tumourincidence could be increased or the time totunour could be decreased by increasing dose,

although there was a minimum time beforetumours would develop. The secondimportant concept was that carcinogenesiscomprises two stages, tumour initiation andpromotion. This was developed as anoperational paradigm for chemicalcarcinogenesis in the skin, by Berenblum andShubik (72,73). Experimental analysis of skincarcinogenesis showed that this first required ashort or limited exposure to a chemical thatresulted in an irreversible change in the skin,which was called initiation. This needed to befollowed by prolonged exposure to a chemicalwhich could acts as a promoter of the initiatedcells, the effects of which were reversible forsome time. This stage was called promotion.The model had a number of requirements:chemicals that acted as promoters did not actas initiators; initiation had to precedepromotion; promotion could be delayed forsome time after initiation.

10.23 This general model has been extended toapply to a number of other cancer types. Inaddition, while first developed as anoperational model to describe experimentalobservations, it has since acquired amechanistic interpretation although this maynot be as well supported as the originaloperational description. Nonetheless, initiationis now generally taken to imply primarydamage to DNA leading to a critical mutationwhile promotion is taken to mean theepigenetic steps that allow expression of theprimary genetic lesion through the acquisitionof other heritable genetic changes (74,75).

10.24 The original experimental model has sincebeen modified. Chemicals that can act ascomplete carcinogens are required to act asboth initiators and promoters, although thecomplete process may take a considerableperiod of time and the two stages can stillsometimes be discerned experimentally.Compounds that do not cause direct damage toDNA may, however, increase tumourincidence because they “promote” cells thathave undergone some prior spontaneous

Chapter ten - Assessment of carcinogenic potential

35


mutation. It has been suggested that particularcauses of this may be oxidative damage toDNA; exposure to exogenous carcinogens;reduced repair capacity as animals age (76). Ithas also to be appreciated that this convenientexperimental model is too simplistic and ifmolecular biology has taught us anything it isthat carcinogenesis is a complex process (77,78).

10.25 The evolution in the design of thecarcinogenicity bioassay has been greatlyinfluenced by the outcome and experience ofthe NCI/NTP bioassay programme. Theprogramme was proposed by MichaelShimkin in the1960’s who saw the need for amore systematic investigation of chemicals forcarcinogenesis. The experimental work wasinitiated by John and Elizabeth Weisberger(68) who asked the simple question: “Howmany industrial chemicals, related in structureto reference rodent carcinogens such as 2-AAFand B[a]P, will also prove to be carcinogenic torodents”. The answer turned out to be themajority. As the NTP programme grew, otherchemicals, with no structural precedent forcarcinogenicity, were added to the growing listof chemicals nominated for bioassay. It turnedout that a similar proportion of these agentswere also carcinogenic, but that their tumourprofile was different to that of the earliercarcinogens (79).

10.26 A decade of mechanistic studies confirmedthat there are many ways for a chemical toincrease the tumour incidence in rodents, onlysome of which appear relevant to humans (80).This view of relevance is not universallyaccepted; some still consider that any increasein tumour incidence induced by a chemical, ineither rats or mice, is of immediate relevanceto humans. Indeed it has been argued that thevalue of the rodent bioassay lies in the fact thatthere is no assumption of mechanism of cancerdevelopment (81).

10.27 Although there has been much discussion ofthe relevance of the testing procedures theyare still based on a dual strategy:

1) assessment of genotoxic potential

2) assessment of carcinogenic potentialthrough life-time studies in rodents

10.28 The data from such studies may be supportedby investigative studies aimed at determiningmode of action and relevance to humanexposure (dose, metabolism, etc.)

10.29 Experimental data are then set againstpotential human exposure and a riskevaluation is carried out. This process maydiffer between pharmaceutical and othersituations such as environmental or work-placeexposure. In either case assessment is madeagainst potential benefit but the weighting maybe different in different situations.

Current Position on Genotoxicity Tests

10.30 Over the past 30 years there have been severalperceived roles for genetic toxicity assays inthe hazard definition/risk assessment process.Initially it was assumed that the Salmonellabacterial mutation assay could single-handedlypredict potential rodent carcinogens, and byimplication, potential human carcinogens.Considering that the rodent and humancarcinogens known at that time were potent,due mainly to the fact that they had beencapable of detection using limited rodentbioassays or small human epidemiologicalstudies, it was not surprising that the largemajority (� 90%) of them were also damagingto DNA (genotoxic) and mutagenic to bacteria.Further, the mutagenicity of such carcinogenscould usually be easily rationalised in terms oftheir chemical structures being reactive toDNA, either directly or following predictablemetabolism (structurally alerting).

10.31 However, with more refined and detailedassessments of carcinogenicity came tworealisations. First, that not all agents defined asgenotoxic in vitro are carcinogenic to rodents,and secondly, that approaching 50% of newlydefined carcinogens were devoid of genetictoxicity. The latter problem has beenaddressed in two distinct ways. First, it wasassumed that these non-genotoxic carcinogensare, in fact mutagens incapable of detectionusing current assays. This has led to thedevelopment of ~100 new mutagenicity assaysfor the detection of alleged ‘crypto-mutagens’.Alternatively, there was the viewpoint that

such instances are representative of a growinggroup of non-genotoxic carcinogens whosedetection must rely on biological activitiesother than mutagenicity. This latter view is theone that currently has general acceptance andis the principle on which most testing and riskassessment are done today.

10.32 All candidate drugs are now routinelyassessed for genetic toxicity, as described laterherein. Except in cases of overwhelmingpotential benefit, in practice usually valuableanti-cancer agents, agents with intrinsicgenetic toxicity are not developed further. Allcandidate drugs are submitted to lifetimerodent carcinogenicity bioassays as part oftheir development, and irrespective of the factthat only agents devoid of genetic toxicity aresubmitted to bioassay, a large number of newcandidate drugs produce positive results whenbioassayed for rodent carcinogenicity, 40% inan analysis published in 2001 (82). This wouldbe an unacceptable rate of attrition – either theexisting genetic toxicity assays are lacking, orthe rodent bioassay is sensitive to aspects ofchemical toxicity independent of genetictoxicity and of limited or no relevance tohumans (see Section 10.44).

10.33 The term ‘non-genotoxic carcinogen’ is stillcontested in some quarters, but it is nowgenerally accepted that some chemicals canincrease cancer incidence in rodents andhumans by virtue of properties they possesbeyond simple DNA-reactivity. Perhaps themost striking and explicable examples of non-genotoxic carcinogenesis are the humancarcinogenicity of the immunosuppressiveagent cyclosporin A, and the carcinogenicityto the mouse uterus of the synthetic oestrogendiethylstilbestrol and the phytoestrogengenistein. The carcinogenicity of the last twochemicals is clearly related to their intrinsicestrogenic activity, as evidenced by the factthat equal carcinogenic incidences wereobserved when each was tested at equi-estrogenic doses, as defined by the mouseuterotrophic assay (83). However, mostapparent non-genotoxic carcinogens are notassociated obviously with an alternativemechanism of carcinogenic action. The mainchallenge facing safety assessment of drugs

36


thus becomes to recognize a range of possiblemarkers of non-genotoxic carcinogenesis,such that these can be used to supplementgenetic toxicity data when predictingcarcinogenic potential.

Current use of genetic toxicity assays

10.34 Due to the unavoidable and insolubleproblems faced when mutagenicity assayswere attempted to be used to predict all classesof chemical carcinogen, attention has turnedover past years to gaining internationalconsensus on three more justifiable andpractical uses for these assays, as discussedbelow.

10.35 For the definition of genetic toxicity:this is the most secure use of genetic toxicityassays, because used in this way they aredefining an activity per se, not providing data tobe used to predict another biological activity.The clearest consensus statement on thismatter is provided by the testing strategyreported by the International Program forChemical Safety (84). This strategy involvesassessment of chemical structure and genetictoxicity in vitro. The assays used in this phaseof testing involve bacterial mutation andmammalian cell mutation tests – the precisemammalian cell assays to be used being leftundefined. However, both the IPCS strategy,and that of most regulatory bodies, implies theneed for both gene mutation assays andcytogenetic assays. Great progress has beenmade on refining these assay protocols inorder to reduce the incidence of false results.

10.36 Chemicals devoid of genetic toxicity in vitroare assumed by most investigators not to posea hazard in vivo. However, some authoritiesrequire the conduct of a rodent bone marrowcytogenetic assay as part of preliminaryscreening. Agents found to be genotoxic invitro are either assumed to present a potentialhazard, or they are assessed in vivo to establishtheir genetic toxicity to rodents. The use ofrodent assays for unscheduled DNA synthesis(UDS) or DNA fragmentation (e.g., the Cometassay) means that most testing strategies endwith a statement of genetic toxicity (i.e.

37


damage to DNA), as opposed to mutagenicity(i.e. heritable change).

10.37 The recent introduction of transgenic rodentgene mutation assays (such as BigBlueTM ratsand mice, and MutaMouseTM mice) means thatthe mutagenicity of chemicals in any tissue of arodent may now be assessed. However, theseassays have not replaced the more establishedassays mentioned above due to uncertaintiesregarding their sensitivity, their relatively highcost, and the current absence of internationallyagreed regulatory guidelines for their conduct.Currently, these assays are most often used toassess mechanism of action of tissue-specificrodent carcinogens.

10.38 For the prediction of somatic and germcell mutagenicity: germ cell mutagenicity isa major toxic endpoint for chemicals. This isusually assessed only for males via the use ofthe rodent dominant lethal, heritabletranslocation, or specific locus assay. Theseassays are demanding of resources and cannotbe conducted routinely in most laboratories. Tothis end great reliance is placed on the fact thatto date, all rodent germ cell mutagens are alsoactive as somatic cell mutagens, as detectedusing the IPCS strategy. About 30 rodent germcell mutagens are currently known, all of whichare detected by the standard rodent bonemarrow micronucleus assay – and no examplesexist of a confirmed human germ cell mutagen.

10.39 For the prediction of genotoxiccarcinogenesis: had Bruce Ames entitledhis 1974 paper ‘Genotoxic carcinogens aremutagens’, instead of ‘Carcinogens aremutagens’ the situation would have not alteredover the succeeding years. However, the fewcarcinogens tested by Ames that were notmutagenic to Salmonellae (sodium saccharin,thiourea, etc) slowly grew to the point whereclose on half of all rodent carcinogens definedby the late 1990s were not mutagenic toSalmonellae. However, genotoxiccarcinogens, whose mechanism of actioninvolves perturbation of the genetic integrity ofthe host animal, as the direct result of theagent’s ability to alter the integrity of hostDNA, can be reliably detected using the teststrategy outlined above.

10.40 The problem of false positivepredictions of carcinogenicity: not allagents found to be genotoxic in vitro aregenotoxic in vivo. This usually requiresadditional experiments in rodents beforeconcluding on the genetic toxicity of an agent.Most sources of truly artifactual positive resultsin vitro have been eliminated by protocolrefinements (controlling pH, osmolarity etc),and it is probable that most instances of ‘falsepositive’ genetic toxicity results representfailures of the assays to anticipatepharmacokinetic (PK) or pharmacodynamic(PD) (e.g. host response to initiation) factorsassociated with the agent in rodents. This is aproblem common to all in vitro assays andcannot be remedied without recourse to animalexperimentation or the advent of accuratemodelling of PK and PD factors.

Current Bioassay Practice

10.41 The standardized version of the bioassayrequires that two rodent species are exposedfrom early adulthood to the test chemical for alarge proportion of their natural lifespan (75%).Groups of 50 or 60 animals are given the testcompound by an appropriate route. Theremay be 3 or 4 dose levels with one or morecontrol groups. The top dose is generally set atthe Maximum Tolerated Dose and lower dosesat geometric intervals. Earlier designs mayhave only two dose levels: the MTD and 1/2MTD. For pharmaceutical agents the lowerdose may be a dose level that results in anexposure, as measured by the AUC, that isclose to the exposure seen in patients attherapeutic doses. On occasion the top dosemay be set at some large multiple (e.g. 25)of the therapeutic exposure (AUC) (seeSection 8).

10.42 The current practice for animal bioassays isformalized in various national and internationalguidelines and for pharmaceuticals has beenharmonized under the ICH guidelines (85).

10.43 There has been much discussion over theutility of the rodent bioassay in the process ofcarcinogenesis assessment. The significance oftesting at the MTD has been questioned

because of the considerable physiologicaldisturbances that may occur at this dose level(80). Furthermore, the kinetics of thecompound may also be quite different to thatat lower doses, with saturation of metabolicpathways. In addition, extrapolation fromexperimental ranges to human exposure levelsoften requires assumptions regarding the doseresponse relationship that are, by definition,not verifiable by experimental evidence. Thecarcinogenic response may differ betweenspecies and between different strains: aflatoxinB1 causes liver cancer in rats when given in thediet at a few parts per million whereas themouse is non-responsive at 1000 times thislevel (86); dimethylbenzanthracene causesmammary tumour development in Sprague-Dawley rats but not in the Wistar strain (87).Such differences may reflect differences inmetabolism or differences in interaction with anumber of other risk factors such as hormonalstatus and are evident, as indicated, even forgenotoxic agents.

10.44 Species and strain difference in response areeven greater for non-genotoxic agents. Inmany cases increased tumour incidencecaused by non-genotoxic carcinogens inrodents, and in at least one case in the dog,have been shown to be strain, sex or speciesspecific and therefore, it is claimed, notrelevant for human risk assessment (87).Recently, detailed mode of action analysis withsystematic comparison of the key events and ofpharmacokinetics between experimentalspecies and humans has provided support forthis claim, in several instances (88,89).

10.45 The need for both the rat and the mouse inbioassays has been questioned regularly. Informulating its guidelines, the InternationalConference on Harmonisation of TechnicalRequirements for Registration ofPharmaceuticals for Human Use (ICH)evaluated available data on the testing ofpharmaceuticals for carcinogenicity in mouseand rat bioassays. The conclusion reached wasthat the mouse contributed little in theevaluation of carcinogenic potential ofpharmaceuticals to humans and that therewould be little loss of sensitivity by omittingstudies in the mouse (85). However, the

38


consensus position adopted was thatpharmaceuticals should be assessed forcarcinogenic potential in a rat bioassaytogether with an additional in vivo test forcarcinogenicity. The second test could beeither a short or medium-term in vivo rodenttest or a long-term test in a second rodentspecies. In practice this has meant either atransgenic mouse model or a mouse bioassay(90,91). In several subsequent studies, ithas again been concluded that themouse provides no added value in theevaluation of the carcinogenic potential ofpharmaceuticals (92).

10.46 For non-pharmaceutical agents humanexposure usually results from environmentalcontamination or work place exposure.Usually the exposure levels are much less thanthose set in the bioassay, and standardizedprocedures for estimating human risk havebeen developed, although these differ amongstregulatory authorities. Human risk can beestimated on the basis of identification ofcarcinogenic hazard and the extrapolation ofresponse based on the dose response curvesand estimates of human exposure. Someassessment of that risk is then made which maybe called “risk characterization” (93).

10.47 Opposing views have been taken on the utilityof using chronic animal data in assessinghuman risk. Thus Gio Batta Gori wrote in2001: “It should be apparent beyond doubtthat presently no science is available for thetranslation of chronic animal test data intoobjective forecasts of human cancer risk” (94).This contrasts with the more optimistic viewexpressed by Maronpot, Flake & Huff (95),who argued that useful data could be obtainedfrom animal studies. They argued that thereare sufficient similarities in the anatomy,physiology and carcinogenic process betweenman and animals for studies in animals tohave relevance for man. Their argument,however, was based on the identification ofhazard rather than an estimate of risk and ineither event for the majority of animalcarcinogens there is no corroborating data inman. There is excellent concordance betweenknown human carcinogens and animal data(81); however, the converse is not true; there

39


are many thousands of animal carcinogens forwhich there is no concordant human data (94).

10.48 There are two possible reasons for this lack ofconcordance: 1) the compounds are notcarcinogenic in man or 2) the conditions underwhich they are carcinogenic in rodents do notapply to human exposure: a further possibilityis that some of these compounds arecarcinogenic in man but the tumour incidenceis too low to detect in feasible studies.Nevertheless, it is unlikely that that we willever be allowed to expose groups of 50humans, held in cages separated by sex, feddiets in excess of need to doses of a test articlethat can be barely tolerated for between theage of 15 and 65!

10.49 These difficulties let to alternative models:mice were developed with geneticmodification that incorporated changesconsidered important in the carcinogenicprocess. It was thought that these animalswould give a more consistent response in ashorter time. Recent ICH guidance indicatesthat one long-term rodent bioassay supportedby additional evidence, such as that obtainedusing a transgenic mouse model, may beacceptable and give improved risk assessment.(91). However, it is unclear which mechanismsof carcinogenicity were expected to bedetected by each of these new models; thisallows constant retro-justification of the resultsobtained when testing 2-year NTP bioassaycarcinogens (96). Despite early hopes, none ofthe available transgenic rodent carcinogenicitybioassay models (97) has matured to bereliable and capable of general adoption.

10.50 The full expression of a cancer phenotyperequires changes in oncogenes, tumour-suppressor genes and so called stability genessuch as those involved in DNA repair and inmitotic recombination and chromosomalsegregation (78). If these general principleshold, then it would be more appropriate todefine the conditions under which thesealterations might exist in order to improve riskassessment. It is recognized, however, thatrodents, and their cells, may behave in adifferent way to humans and human cells,which is an additional confounding factor

when assessing human risk from animaldata (98).

10.51 What would be reasonable to expect from therodent bioassay and is there a better way ofassessing carcinogenic risk to man? If thequestion is: ‘does this chemical increase thetumour incidence in any of the available rodentbioassay models’ then recourse to all of thosemodels will be needed in order to define a non-carcinogen. If the question is: ‘does thischemical present an obvious cancer hazard toexposed humans’ then limited and focusedstudies can lead to an adequate answer.Balancing the adequacy of answers with therate of testing of chemicals is obviously a matterof judgment, but such judgment is rarelydiscussed.

Current Understanding of Carcinogenesis

10.52 Before discussing an alternative strategy, thereare a number of general points that should beconsidered:

� Pattern of tumour types and incidence differbetween animal species and between strains.Eighty percent of the tumours that arise inman are carcinoma while 80% of tumours insome strains of mice are sarcoma, althoughthe life-time risk of developing cancer isabout the same in man and rodents (30%).(70,98)

� Smoking is a major confounding factor intumour development in man

� Diet plays an important role in cancerdevelopment in man and animals(70,99,100)

� Standardised incidence rates show up to100- fold variation in certain cancers indifferent geographical regions (IARC WorldCancer Reports, 2003)

10.53 Cumulative data in both man and animalsshow that cancer results from the progressiveaccumulation of a number of traits related to alimited number of molecular functions:proliferation, differentiation and death (75,77).These functions may be modified byalterations in the function of three broad gene

classes: tumour-suppressor genes, stabilitygenes and oncogenes (78). There are over ahundred genes that may be associated withone or other of these broad classes; not all areneeded and the sequence of action and thedetail of their interaction are poorlyunderstood. Some steps in the carcinogenicprocess, and perhaps even the occurrence oftumours in an individual, are stochastic events,and hence not reliably predictable other thanon the basis of probability theory (101).

10.54 The complex nature of any assessment that hasto be made in assessing risk may be illustratedby the case of estrogens and the role they playin influencing the incidence of cancer atseveral sites in man (102). Epidemiologicaldata indicate that prolonged unopposedestrogen exposure may double the risk ofendometrial cancer (103). Obesity is also animportant risk factor in post-menopausalwomen and it has been postulated that theincreased risk is associated with the conversionof androstendione to estrone in adipose tissue(104,105). Similarly the risk of breast cancer isincreased in post-menopausal women inwhom estrogen levels are raised (106). Animalstudies have also convincingly shown thatestrogens play an important role in theinduction and promotion of mammarytumours (107,108). Furthermore, both medicaland natural Selective oEstrogen-ReceptorModulators (SERMs) appear to have aprofound affect on the development of breastcancer although the relationship is complexand may depend on genetic and otherpredisposing factors (109,110).

10.55 SERMs may exhibit different estrogenic andanti-estrogenic effects in different tissues and indifferent hormonal states (110). There is ageneral consensus that estrogens may play apromotional role in carcinogenesis byincreasing proliferative activity thus increasingthe risk of critical mutations occurring or“fixing” those that have already occurred (111).However, a number of oxidative metabolitesof estrogens have been reported, which bind toDNA, and that may result in mutation andinitiation of the cancer process in breast tissue(112). Against this, it is possible to inhibit thecarcinogenic response to estrogens with anti-

40


estrogens, without any change in covalentbinding to DNA (113).

10.56 The estrogen receptors are members of thenuclear-receptor superfamily that plays animportant role in cell differentiation andproliferation (114). As well as estrogenreceptors it includes androgen, retinoic acid,retinoid X, vitamin D, constitutive androstaneand per-oxisome proliferator-activatedreceptors (110) and it would not be surprisingif modification of these receptors influencedthe incidence of a variety of tumours in thelong term. However, the complexity of theirinteraction is further confounded by theactivity of transcriptional co-activators and co-repressors that may modify the response ofligand-receptor interactions in different tissues(115). In addition, such receptors can showmarked tissue and species difference in theirlevels of expression, e.g. PPAR� This makesextrapolation across tissues and betweenspecies difficult.

10.57 The epidemiological data, observations ofcancer in man and experimental data supportthe view that cancer morbidity rates may beinfluenced by environmental factors. It is nowgenerally considered that the term‘environmental factors’ should not be taken toequate only to man-made chemicals, but togross aspects of diet, to infectious agents and tochemicals that are either natural in origin orproduced by cooking practices. Nonethelessthe view that identification of specific chemical-risk for carcinogenicity is the primary means ofpreventing cancer still largely persists andremains the cornerstone of present testingpolicies including the testing of drugs.

10.58 It is unlikely that an overtly genotoxicchemical (other than perhaps an anti-cancerdrug) would be developed as a pharmaceuticalin the present regulatory environment. Theproblem that we face is that there are multiplepotential epigenetic mechanisms that mightlead to increased cancer and some of thesemay be species and organ specific. In both ratsand mice tumours that have arisen because ofnon-genotoxic mechanisms and have beenpreceded by hyperplasia have usuallyoccurred at sites subject to a high background

41


incidence (116). In these cases it has beenargued that the process is not relevant inassessing risk to man because of a thresholdabove therapeutic exposure or because of amechanism that is not applicable to man (117).Some of these are listed in table 1.

10.59 Without prior knowledge of outcome in manit seems prudent to make some assessment ofpotential carcinogenicity in man fromexperimental data. The question is what is theappropriate experimental system. Thebioassay has been criticized because of costand of the uncertainty in whether theobservations that are made are relevant toman: induction of liver tumours is common inrodents but this is not a common tumour typeof man in western industrialised countries. Ithas been proposed that better mechanisticunderstanding would improve predictivityand risk assessment and this is encapsulated inthe concept of evaluating mode of action of acarcinogenic effect (118). In some instances,as indicated, it may be sufficient to show thatthe process, as observed in rodents, has norelevance to man: examples of this are theinduction of �2u-globulin-associated kidneytumours in male rats and thyroid tumours thatresult from perturbation of thyroid hormonestatus (117,118,119). In other instances, themode of action may prove relevant, but clearthresholds for key events in humans can beidentified (89). However, these notableexamples are the exceptions and in most casesin which mode of action has clearly beendefined it is generally for potent genotoxiccarcinogens. Detailed analysis for butadiene,vinyl chloride and benzene has been reportedby Albertni et. al. (119), and the analysis

clearly shows how this kind of data may beused to further characterize the dose responseand assist in the extrapolation of the risk toman.

10.60 Additional features that have been associatedwith experimental carcinogenesis are:increased oxidative stress; hormone orhormone-like effects; perturbation in theimmune system and changes in apoptotic ratesin preneoplastic lesions (altered foci in theliver). While it is probable that some of thesefactors are important in the genesis of cancer inman, the magnitude of the risk is uncertain aseven in animals dose-response relationshipsare poorly documented and, in man, largelyunavailable other than in the case ofhormonally mediated disease (120).

10.70 Detailed analysis of the relationship of hepaticcarcinogenesis to prechronic liver lesions (121)for over 80 chemicals in the NTP bioassayprogramme, showed that hepatocellularhyperptrophy gave a good prediction ofcarcinogensis in both rat and mice. This wasimproved if increase in liver weight wasincluded, however, in both cases the falsepositive rate was high and markedly so whenliver weight was added. Interestingly theconcordance between results of the Salmonellamutation assay and carcinogenicity in the liverwas poor, supporting the view that themajority of rodent hepatocarcinogens actedthrough a non-genotoxic mechanism.

10.71 These data are of interest in that they clearlydemonstrate that the induction of tumours inrodents is complex and that the relative risk ofeach of the potential contributing mechanisms

Liver Enzyme induction altered foci Adenoma/carcinomaKidney alpha-2u-globulin Hyperplasia Adenoma/carcinomaMammary Prolactin elevation Hyperplasia AdenocarcinomaThyroid TSH elevation Hyperplasia Adenoma/carcinomaPancreas Trypsin inhibitors Hyperplasia Adenoma/carcinomaStomach Proton pump/H2 antagonist Hyperplasia Carcinoid tumoursTestis LH elevation Hyperplasia Leydig cell tumoursSalivary �2-agonist Hyperplasia Adenoma/carcinomaMuscle �2-agonist Hyperplasia Leiomyoma

Table 1. Epigenetic modes of action in rodent carcinogenesis

is difficult to quantify. Nonetheless it is clearthat epigenetic processes are relevant inassessing risk for cancer in man. Detailedmode of action analysis can help in evaluationof the potential risk of rodent hepaticcarcinogens to humans. We have previouslyalluded to the complex issue of hormonallymediated cancer but even in the case ofsmoking-induced pulmonary cancer in manepidemiological data would suggest thatepigenetic mechanisms also play an importantrole (120).

Recommendations

10.72 There are many safeguards built in to the drugdiscovery processes that mitigate the risk ofcausing cancer in humans. Drugs are designedto interact with well-characterized, disease-modifying molecular targets, the dose isknown, and the exposure is controlled,treatment is under medical supervision, andthere is always intended benefit to offset risk.The strategies and methods used to date toavoid introducing drugs that might pose acarcinogenic hazard to humans appear to havemet their objective. It appears timely thereforeto propose improvements to current testingstrategies in line with current knowledge andunderstanding. Modifications to the testingstrategy in the short term would include atiered approach to assessing carcinogenichazard as follows:

� overtly genotoxic new drug candidateswould be identified, with the exception ofDNA-reactive drugs for cancer, andeliminated from development prior to theconduct of the conventional rodentbioassay. This would be accomplished bystructure-activity considerations and by abattery of recommended genotoxicity tests.

� the bioassay, or any suitable in vivo test forcarcinogenicity, would be directed atdetecting carcinogenic potential other thanby direct genotoxicity. Given concern overthe interpretation of the conventionalanimal bioassays, it is proposed to limit thisto a single study conducted in a suitable ratstrain. The reasons are as follows:

42


� there is generally considerably more bio-chemical and metabolism data in the rat

� the particular sensitivity of the mouse tothe development of hepatic tumours andthis often confounds interpretation

� conducting a long-term rodent studywould allow the detection of novelcarcinogenic mechanisms that requirethe whole animal: additional assessmentwould then be made on a case-by-casebasis.

� additional assessment of risk wouldbe conducted on a tissue by tissue basislooking at:

� proliferative, antiproliferative andapoptotic drivers

� oxidative stress, genetic damage andrepair

� general effect on the immune system.

� hormone or hormone like effects

� pertubation in metabolic function

� the use of genetically modified animalsshould be on a for-cause basis in order tofurther characterise mechanism.

10.73 Further modifications to testing strategies inthe longer term could be introduced instepwise fashion based on experience,accumulated knowledge and the evolution ofnew and advanced methods for assessingcarcinogenic hazard, for example:

� as understanding of the modes of action ofnon-genotoxic carcinogens increases, it mayprove possible to develop more focused,shorter-term studies both to detect and toexclude potential human carcinogens andthus eliminate the need for a lifetime studyin rodents. This would be based on analysisof the type of precursor effects listed above,augmented as appropriate with novelbiomarkers identified in multi-parametricassays (e.g. proteomics). Compoundsnegative in such an assay would beconsidered to pose no carcinogenic risk tohumans. Compounds that were positivemay be tested further, for example in a

43


lifetime bioassay, to confirm the findings.The decision to proceed with a lifetimerodent bioassay would be based onconsiderations of dose, duration ofexposure, clinical indication and potentialbenefits.

� this information should be collected in asystematic way with a clear emphasis ondose response and threshold effect.Genomics, proteomics and other of thenewer technologies may be used in order toadd additional mechanistic understandingrather than for screening. These data wouldbe used to assess mode of action in respectto a positive bioassay result and secondly forextrapolation to man.

� data from animals show that for genotoxiccarcinogens the site of carcinogenicity maydiffer between species. Furthermore, somenon-genotoxic modes of action appear siteand species specific and so extrapolation

across species is difficult, although siteconcordance for human-relevantcarcinogens will be greater, because of thespecificity of the mechanism. This isconfounded by the relative ease oftransformation of animal cells whencompared with human cells thatfurther magnifies the complexity for theextrapolation of risk. We therefore proposethat greater emphasis should be given toidentifying conditions that would enhancethe risk for the common human tumours asit is among this group that marginal changesin risk would have the greatest affect. In manboth obesity and inflammation may affecttumour incidence at a number of sites. Themechanism is not understood but may bemediated by cytokines or hormones or bychanges in local metabolism leading tooxidative injury, the very targets at whichmany drugs under development are aimed.

44


There has been continuous improvement in allaspects of drug safety assessment over the pastseveral decades. However, given experiences ofunexpected drug withdrawals in recent years, it istimely to review regulatory processes and industryprocedures with a view to seeking improvements.Although there appears not to be a need for radicalchanges to the overall process, there are ways inwhich new methods and different ways of working,

particularly at the interface between regulatoryagencies and industry, could effect improvements inrisk assessment and drug safety. If these changes areintroduced incrementally they will have a significantimpact on the ability of pharmaceutical companiesand regulators to deliver safer medicines to patients.We hope that some of the recommendationspresented in this review may provide further impetusfor discussion of the issues.

Conclusions

45


1. Bakke OM, Manocchia M, de Abajo FJ, Kaitin KI,Lasagna L (1995) Drug safety discontinuations in the United Kingdom, theUnited States, and Spain from 1975 through 1993: aregulatory perspective. Clin Pharm Ther 58, 108–117

2. Fung M, Thornton A, Mybeck K, Wu JH,Hornbuckle H, and Muniz E (2001) Evaluation of the characteristics of safety withdrawal ofprescription drugs from worldwide pharmaceutical markets –1960 to 1999. Drug Information Journal 35, 293–317

3. Friedman MA, Woodcock J, Lumpkin MM, Shuren JE, Hass AE, and Thornpson LJ (1999) The safety of newly approved medicines. Do recent marketremovals mean there is a problem?JAMA 281, 1728–1734

4. DiMasi JA, Hansen RW, and Grabowski HG(2003) The price of innovation: new estimates of drug developmentcosts. J Health Econ 22, 151–185

5. Dickson M, and Gagnon JP (2004) Key factors in the rising cost of new drug discovery anddevelopment. Nature Drug Discovery 3, 417–429

6. Stephens MDB (2004) Drug products withdrawn from the market for safety reasons. In: Stephens’ Detection of New Adverse Reactions:Fifth Edition, pp1–91. Eds. John Talbot and PatrickWaller. John Wiley and Sons Ltd.

7. Towler C (2002) Risk management: the new paradigm. In: Risk Management: the role of regulatory strategiesin the development of new medicines, pp11–18. Eds.Anderson C, McAuslane N, and Walker SR. Centrefor Medicines Research International, Institute forRegulatory Science, Epsom, UK

8. Lumley CE, Walker SR, Staunton N, and Grob PR(1986) The under-reporting of adverse drug reactions seen ingeneral practice. Pharmaceutical Medicine 1, 205–212

9. Pirmohamed M, James S, Meakin S, Green C,Scott AK, Walley TJ, Farrar K, Park BK,Breckenridge AM (2004) Adverse drug reactions as a cause of admission to hospital:prospective analysis of 18,820 patients. Brit Med J 329, 15–19

10. Silcock J and Pritchard C (2003) To Heal and harm: an economic view of drug safety. Office of Health Economics, London

11. OHE (2003) Compendium of Health Statistics. London: Office of Health economics

12. Workshop review (2002) Risk Management: The role of regulatory strategies in thedevelopment of new medicines. Eds. Anderson C, McAuslane N and Walker SR.Centre for Medicines Research International,Institute for Regulatory Science, Epsom, UK

13. Raine J (2002) Visions from a European perspective: the MCA. In: Risk Management: The role of regulatorystrategies in the development of new medicines, pp 43–50. Eds. Anderson C, McAuslane N, andWalker S R. Centre for Medicines ResearchInternational, Institute for Regulatory Science, Epsom, Surrey

14. Lonngren T (2002) Factors influencing risk management in the EU. In: Risk Management: The role of regulatorystrategies in the development of new medicines,pp59–65. Eds. Anderson C, McAuslane N and Walker SR. Centre for Medicines ResearchInternational, Institute for Regulatory Science, Epsom, UK

15. Roses AD (2002) Genome based pharmacogenetics and the pharmaceuticalindustry. Nature Reviews: Drug Discovery, 1, 541–549

16. Lindpainter L (2002) The impact of pharmacogenetics and pharmacogenomics ondrug discovery. Nature Reviews: Drug Discovery 1, 463–469

References

17. Lipinsky C and Hopkins A (2004) Navigating chemical space for biology and medicine. Nature 432, 855–865

18. Robinson DE, Pettit SD and Morgan DG (2003) Use of genomics in mechanism based risk assessment. In: Toxicogenomics pp194–203. Eds. Inoue T andPennie WD. Springer-Verlag, Tokyo

19. Ulrich R and Friend SH (2002) Toxicogenomics and drug discovery: will new technologieshelp us produce better drugs?Nature Reviews: Drug Discovery 1, 84–88

20. Hughes TR, Marton MJ, Jones AR, Roberts CJet al. (2000) Functional discovery via a compendium of expressionprofiles, Cell 102, 109–126

21. Dobson CM (2004) Chemical space and biology. Nature 432, 824–828

22. Thomas RS et al (2001) Identification of toxicologically predictive gene sets usingcDNA microarrays. Mol. Pharmacol. 60, 1–6

23. Jones AK, Buckingham SD, and Sattelle DB(2005) Chemistry-to-Gene screens in Caenorhabditis Elegans. Nature Reviews: Drug Discovery 4, 321–330

24. Spitsbergen JM and Kent ML (2003) The state of the art of the Zebrafish model for toxicologyandtoxicologic pathology research – advantages and currentlimitations. Toxicol. Pathol. 31, Suppl. 62–87

25. MacGregor JT, Casciano DA and Muller L (2000) Strategies and testing methods for identifying mutagenic risk. Mutat. Res. 455, 3–21

26. Searles DB (2005) Data integration: challenges for drug discovery. Nature Rev Drug Discovery 4, 45–58

46


27. Kilty C, Doyle S, Hassett B and Manning F(1998) Glutathione-S-transferases as biomarkers of organ damage:Applications of rodent and canine GST enzymeimmunoassays. Chem Biol Interact 111–112, 123–135

28. Han WK, Bailly V, Abichandani R, Thadhani Rand Bouventre JV (2002) Kidney injury molecule-1 (KIM-1): a novel biomarker forhuman proximal tubule injury. Kidney Int 62, 237–244

29. George DK, Ramm GA, Walker NI, Powell LWand Crawford DH (1999) Elevated serum Type IV collagen: a sensitive indicator of thepresence of cirrhosis in hemachromatosis. J Hepatol 31, 47–52

30. Newsholme SJ, Thudium DT, Gossett KA,Watson ES, and Schwartz LW (2000) Evaluation of plasma von Willebrand factor as a biomarkerfor acute arterial damage in rats. Toxicol Pathol 28, 688–693

31. Arnal JF, Dinh-Xuan T, Pueyo M, Darblade Band Rami J (1999) Endothelium derived nitric oxide and vascular physiologyand pathology. Cell Mol Life Sci 55, 1078–1087

32. Drab M, Verkade P, Elger M, Kasper M, Lohn M,Lauterbach B et al (2001) Loss of caveolae, vascular dysfunction and pulmonarydefects in caveolin-1 gene-disrupted mice. Science 293, 2449–2452

33. International Life Sciences Institute – Health and Environmental Sciences Institute(2002) Subcommittee on the development and application ofbiomarkers of toxicity. November 4–5, Washington DC

34. US FDA (2004) Draft Guidance for Industry: Pharmacogenomic datasubmissions.

35. Gehring PJ, and Blau GE (1977) Mechanisms of carcinogenesis: dose response. J Env Path & Toxicol 1, 163–179

47


36. Moloney D (1993) Toxicity tests in animals: extrapolating to human risks. Env Hlth Perspectives 101, 1–24

37. Morgan DG, Kelvin AS, Kinter LB, Fish CJ et al(1994) The application of toxicokinetic data to dosage selection intoxicology studies. Toxicol Pathol 22, 112–123

38. Capen C (2001) Overview of structural and functional lesions in endocrineorgans of animals. Toxicol Pathol 29, 8–33

39. Goodman JI and Watson RE (2002) Altered DNA methylation: secondary mechanism involvedin carcinogenesis. Ann Rev Pharmacol Toxicol 42, 501–525

40. Swenberg JA, Richardson FC, Boucheron JA,Deal FH, et al (1987) High- to low- dose extrapolation: critical determinantsinvolved in the dose response of carcinogenic substances. Env Hlth Perspectives 76, 57–63

41. Green S (1995) PPAR: a mediator of peroxisome proliferator action. Mutat Res 333, 101–109

42. Thompson CB (1995) Apoptosis in the pathogenesis and treatment of disease. Science 267, 1456–1462

43. MacGregor JT, Farr S, Tucker JD, Heddle JAet al (1995) New molecular endpoints and methods for routine toxicitytesting. Fund Appl Toxicol 26,156–173

44. Aardema MJ and MacGregor JT (2002) Toxicology and genetic toxicology in the new era oftoxicogenomics: impact of “-omics” technologies. Mutat Res 499, 1–16

45. Chandler M and Spain M (2004) Pitfalls to using small animals in preclinical testing arebeing eliminated. Preclinica 1, 245

46. Tucker GT, Houston JB, and Huang S-M (2001) Optimizing drug development: strategies to assess drugmetabolism/transport interaction potential – towards aconsensus. Brit J Clin Pharmacol 52, 102–117

47. Bjornsson TD, Callaghan JT, Einolf HJ, Fischer Vet al (2003) The conduct of in vitro and in vivo drug-drug interactionstudies: A Pharmaceutical Research and Manufacturers ofAmerica (PhRMA) perspective. Drug Met Disp 31, 815–832

48. Ayrton A and Morgan P (2001) Role of transport proteins in drug absorption, distributionand excretion. Xenobiotica 31, 469–497

49. Roberts SA (2001) High-throughput screening approaches for investigating drugmetabolism and pharmacokinetics. Xenobiotica 31, 557–589

50. Park BK, Kittteringham NR, Pirmohamed M,and Tucker GT (1996) Relevance of induction of human drug-metabolizingenzymes: pharmacological and toxicological implications. Brit J Clin Pharmacol 41, 477–491

51. Rostami-Hodjegan A and Tucker GT (2004) ‘In silico’ simulations to assess the ‘in vivo’ consequences of‘in vitro’ metabolic drug-drug interactions. Drug Discovery Today: Technologies 1, 441–448

52. Mussen F (2002) The use of models for benefit-risk assessment. In: Risk Management: the role of regulatory strategies in the development of new medicinespp19–27. Eds. Anderson C, McAuslane N and Walker SR. Centre for Medicines ResearchInternational, Institute for Regulatory Science, Epsom, Surrey, UK

53. Kennedy T (1997) Managing the drug discovery/development interface. Drug Dev Today 2, 436–444

54. Monro A and Mehta D (1996) Are single-dose toxicology studies in animals adequate tosupport single dose of a new drug in humans?Clin Pharmacol Ther 59, 258–264

55. Greaves P, Williams A, and Eve M (2004) First dose of potential new medicines to humans: howanimals help. Nature Rev Drug Disc 3, 226–236

56. Olson H, Betton G, Robinson D et al (2000) Concordance of the toxicity of pharmaceuticals in humansand animals. Toxicol Pathol 32, 56–67

57. Zbinden G (1991) Predictive value of animal studies in toxicology. Reg Tox and Pharmacol 14, 167–177

58. McClain RM (1994) Mechanistic considerations in the regulation andclassification of chemical carcinogens. In: Nutritional Toxicolgy. pp273–304 Eds. Kotsonis FN, Mackey M and Hjelle J. Raven Press, NY

59. Cohen SM (1998) Urinary bladder carcinogenesis.Toxicol Pathol 26, 121–127

60. Kerns WD and Schwartz LW (Eds.) (2004) Drug-induced vascular injury – a quest for biomarkers. Report of the Expert Working Group on Drug-Induced Vascular Injury. Tox Appl Pharmacol,

61. Louden C and Morgan DG (2001) Pathology and pathophysiology of drug-induced arterialinjury in laboratory animals and its implication on theevaluation of novel chemical entities. Pharmacol Toxicol 89, 158–170

62. Danoff TM, Campbell DA, McCarthy LC, Lewis KF et al (2004) A Gilbert’s syndrome UGT1A1 variant conferssusceptibility to tranilast-induced hyperbilirubinemia. Pharmacogenomics Journal 4, 49–53

63. Roses AD (2002) Genome-based pharmacogenetics and the pharmaceuticalindustry. Nature Reviews: Drug Discovery 1, 541–549]

64. Hetherington S et al (2002) Genetic variations in HLA-B region and hypersensitivityreactions to abacavair. Lancet 359, 1121–122

48


65. Pirmohamed M, Breckenridge AM,Kitteringham NR and Park BK (1998) Adverse drug reactions. Brit Med J 316, 1295–1298

66. Berenblum I (ed) (1969) Carcinogenicity testing. A report of the panel oncarcinogenicity of the cancer Research Commission of UICC,Vol 2, International Union Against Cancer, Geneva

67. Weisburger EK (1981) Techniques for carcinogenicity studies. Cancer Res. 41, 3690–3694.

68. Weisburger EK (1983) History of the bioassay program of the National CancerInstitute. Prog. Exp. Tumor Res. 26, 187–201.

69. Epstein SS (1979) The Politics of Cancer. Anchor Books, New York

70. Roe FJ (1989) What is wrong with the way we test chemicals forcarcinogenic activity. In: Advances in Applied Toxicology. Ed. Dayan ADand Paine AJ. Ch 2, pp. 2–17. Taylor and Francis Ltd,London.

71. Druckrey H (1967) Quantitative aspects in chemical carcinogenesis. In: Potential Carcinogenic Hazards from Drugs. Ed.Truhaut R. UICC Monograph Series Vol 7, pp.60–78. Springer-Verlag, New York.

72. Berenblum I and Shubik P (1947) A quantitative approach to the study of the stages ofchemical carcinogenesis in the mouse skinBr. J. Cancer Res. 1, 383–391

73. Berenblum I and Shubik P (1949) The persistence of latent tumour cells induced in the mouse’sskin by a single application of 9,10-dimethyl-1,2-benzanthracene. Br. J. Cancer Res. 3, 384–386.

74. Foulds L (1954) The experimental study of tumor progression: a review. Cancer Research 14, 327–339.

49


75. Hanahan D and Weinberg RA (2000) The hallmarks of cancer. Cell 100, 57–70.

76. Evans MD, Dizdaroglu M and Cooke MS (2004) Oxidative DNA damage and disease: induction, repair andsignificance. Mutation res. 567, 1–61

77. Hahn W and Weinberg RA (2002) Modelling the molecular circuitry of cancerNature Reviews: Cancer 2, 331–341

78. Vogelstein B and Kinzler KW (2004) Cancer genes and the pathways they controlNature Medicine 10, 789–798

79. Fung VA, Barrett JC and Huff JE (1995) The carcinogenesis bioassay in perspective: application inidentifying human cancer hazard. Environ. Health Perspect. 103, 680–683.

80. MacDonald JS and Scribner HE (1999) The maximum tolerated dose and secondary mechanisms ofcarcinogenesis.In: Carcinogenicity. Ed. Kitchin KT. Ch 3, pp.125–144. Marcel-Dekker Inc, New York.

81. Huff J (1999) Value, validity and historical development of carcinogenesisstudies for predicting and confirming carcinogenic risk tohumans. In: Carcinogenicity. Ed. Kitchin KT. Ch 2, pp.21–123. Marcel Decker Inc, New York.

82. Snyder RD and Green JW (2001) A review of the genotoxicity of marketed pharmaceuticals. Mutat Res 488, 151–169

83. Newbold RR, Banks EP, Bullock B andJefferson WN (2001) Uterine adenocarcinoma in mice treated neonatally withgenistein. Cancer Research 61 4325–4328.

84. Ashby J, Waters M D, Preston J, Adler I D,Douglas G R, Fielder R, Shelby M D, Anderson D,Sofuni T, Gopalan H N B, Becking G and Sonich-Mullin C (1996). IPCS Harmonisation of methods for the prediction andquantification of human carcinogenic/mutagenic hazard, andfor indicating the probable mechanism of action of carcinogensMutat. Res., 352, 153–157

85. ICH (1995) Proceedings of the Third International Conference (1995)on Harmonization, Yokohama, Japan, December, 1995.

86. Busby WF and Wogan GN (1984) Aflatoxins, Chemical Carcinogens, Volume 2, Ed. Searle C.American Chemical Society Monograph 182, pp.945–1136. American Chemical Society, Washington,DC.

87. King, RJB (2000) Chemical and Radiation Carcinogenesis.In: Cancer Biology 2 ed. Ch 7, pp. 96–119. PrenticeHall, Edinburgh

88. Cohen SM (1998) Cell proliferation and carcinogenesis. Drug Metabolism Reviews 30, 339–357.

89. Meek ME, Bucher JR, Cohen SM, Dellarco V,Hill RN, Lehman-McKeeman LD, Longfellow DG,Pastoor T, Seed J and Patton DE (2003) A framework for human relevance analysis of informationon carcinogenic modes of action. Critical Reviews in Toxicology 33, 591–653.

90. ICH (1997) Guidance for Industry. S1B Testing for Carcinogenicity ofPharmaceuticals.

91. Van der Laan J-W and Spindler P (2002) The in vivo rodent test system for assessment of carcinogenicpotential.Regulatory Toxicology and Pharmacology 35, 122–125.

92. Van Oosterhout JP, Van der Laan JW, De Waal EJ, Olejniczak K, Hilgenfeld M, Schmidt Vand Bass R (1997) The utility of two rodent species in carcinogenic riskassessment of pharmaceuticals in Europe. RegulatoryToxicology and Pharmacology 25, 6–17.

93. Grasso P and Price S (1999) The cancer bioassay. In: Carcinogenicity. Ed. Kitchin KT. Ch 1, pp. 1–19. M Dekker Inc, NewYork.

94. Gio Batta Gori (2001) The costly illusion of regulating the unknowable risk. Reg. Toxicology and pharmacology 34, 205–212.

95. Maronpot RP, Flake G and Huff J (2004) Relevance of animal carcinogenesis findings to humancancer prediction and prevention. Toxicologic Pathology.32, 40–48.

96. Pritchard JB, French JE, Davis BJ and HasemanJK (2003) The role of transgenic mouse models in carcinogenidentification. Environmetnal Health Perspectives 111, 444–454.

97. Ashby J (1997) Identifying potential human carcinogens – the role ofgenetically altered rodents. Toxicologic Path. 25, 241–243.

98. Rangarajan A and Weinberg RA (2003) Comparative biology of mouse versus human cells: modelinghuman cancer in mice. Nature Reviews: Cancer 3, 952–959

99. Leakey JEA, Seng, JE and Allaben WT (2004) Influence of body weight, diet and stress on aging, survivaland pathological endpoints in rodents: implications fortoxicicty testing and risk assessment. Regulatory Research Perspectives. 4(2), 1–28

100. Bingham S and Riboli E (2004) Diet and cancer-the European prospective investigation intocancer and nutrition. Nature reviews: Cancer 4,206–215

101. Root-Bernstein RS and Bernstein MI (1999) A simple stochastic model of development and carcinogenesis. Anticancer Research 19, 4869–4876.

102. Henderson BE and Feigelson HS (2000) Hormonal carcinogenesis. Carcinogenesis. 21, 427–433.

50


103. Weiss NS and Szekely DR and Austin DF(1976) Increasing incidence of endometrial cancer in the UnitedStates. N Engl. J Med. 294, 1259–1262

104. Potischman N, Hoover RN, Brinton LA, SiiteriP, Dorgan JF, Swanson CA, Berman ML, Mortel R,Twiggs LB, Barrett RJ, Wilbanks GD, Persky V andLurain JR (1996) Case-control study of endogenous hormones and endometrialcancer. J.Natl. Cancer Inst. 88, 1127–1135.

105. Calle EE and Kaaks RF (2004) Overweight, obesity and cancer: epidemiological evidenceand proposed mechanism. Nature Reviews: Cancer 4, 579–591.

106. Zeleniuch-Jacquotte A, Shore RE, Koenig KL,Akhmedkhanov A, Afanasyeva Y, Kato I, Kim MY,Rinaldi S, Kaaks R and Toniolo P (2004) Postmenopausal endogenous oestrogen, androgen, and SHBG and breast cancer long-term results of a prospectivestudy. Br. J. Cancer 90, 153–159.

107. Dao TL (1981) The role of ovarian steroid hormones in mammarycarcinogenesis. In: Hormones and Breast Cancer. Eds. Pike MC,Siiteri PK and Welsch WC., (Banbury Report No 8)pp. 281–295. Cold Spring Harbor Laboratory Press,New York.

108. Allred CD, Allred KF, Ju YH, Clausen LM,Doerge DR, Schantz SL, Korol DL, Wallig MA andHelferich WG (2004) Dietary genistein results in larger MNU-induced, eostrogen-dependent mammary tumors following ovariectomy ofSprague-Dawley rats. Carcinogenesis 25, 211–218

109. Powles TJ (2004) Anti-oestrogenic prevention of breast cancer-the make orbreak point. Nature Reviews: Cancer. 2, 787–794

110. Sporn MB and Suh N (2004) Chemoprevention: an essential approach to controllingcancer. Nature Reviews: Cancer 2, 537–543.

51


111. Feigelson HS and Henderson BE (1996) Estrogens and breast cancer. Carcinogenesis 17, 2279–2284.

112. Rogan EG, Badawi AF, Devanesan PD, MezaJL, Edney JA, West WW, Higginbotham SM andCavalieri EL (2003) Relative imbalance in estrogen metabolism and conjugationin breast tissue of women with carcinoma: potentialbiomarkers of susceptibility to cancer. Carconogenesis 24, 697–702

113. Liehr JG, Sirbasku DA, Jurka E, Randerath K,Randerath E (1988) Inhibition of estrogen-induced renal carcinogenesis in maleSyrian hamsters by tamoxifen without decrease in DNAadduct levels. Cancer Research 48, 779–783.

114. Mangelsdorf DJ, Thummel C, Beato M,Herrlich P, Schutz G, Umesono K, Blumberg B,Kastner P, Mark M and Chambon P (1995) The nuclear receptor superfamily the second decade. Cell, 83, 835–839.

115. Katzenellebogen BS and Katzenellebogen JA(2002) Defining the “S” in SERMSs. Science 295, 2380–2381.

116. Hardisty JF (1985) Factors influencing laboratory animal spontaneous tumourprofiles. Toxicologic pathology 13, 95–104.

117. Ettlin RA and Prentice DE (2002) Unexpected tumour findings in lifetime rodent bioassaystudies-what do we do. Toxicology Lett. 128,17–33.

118. Sonich-Mullin C, Fielder R, Wiltse J, BaetckeK, Dempsey J, Fenner-Crisp P, Grant D, Hartley M,Knaap A, Kroese D, Mangelsdorf I, Meek E, RiceJM and Younes M (2001) IPCS conceptual framework for evaluating a mode of actionfor chemical carcinogenesis. Regulatory Toxicology and Pharmacology 34,146–152.

119. Albertini R, Clewell H, Himmelstein MW,Morinello E, Olin S, Preston J, Scarano L, SmithMT, Swenberg J, Tice R and Travis C (2003) The use of non-tumor data in cancer risk assessment:reflections on butadiene, vinyl chloride, and benzene. Regulatory Toxicology and Pharmacology 37,105–132

120. Clayson DB and Kitchin KT (1999) Interspecies differences in response to chemical carcinogens. In: Carcinogenicity. Ed. Kitchin KT. Ch 29, pp.837–880. M Dekker Inc, New York.

121. Allen DG, Pearse G, Haseman JK andMaronpot RR (2004) Prediction of rodent carcinogenesis: an evaluation ofprechronic liver lesions as forecasters of liver tumours inNTP carcinogenicity studies. Toxcicologic Pathology 32, 393–401

52


2-AAF: 2-acetyl aminofluoreneACTH: adreno-cortical trophic hormoneADME: absorption, disposition, metabolism,

eliminationAUC: area under the concentration-time

curveB(a)P: benzo[a] pyreneCmax: maximal concentrationCNS: central nervous systemCYP: cytochrome PDNA: deoxyribonucleic acidEPA: Environmental Protection Agency

(USA)EPAR: European Public Assessment ReportEU: European UnionFDA: Food and Drug Administration (USA)FMO: Flavine-containing monoxygenaseHESI: Health and Environmental Sciences

Institute (USA)HIV: human immuno-deficiency virusHLA: human lymphocyte antigenIARC: International Agency for Research on

Cancer (France)ICH: International Conference on

Harmonisation of Technical

Requirments for the Registration ofHuman Pharmaceuticals

ILSI: International Life Sciences InstituteIPCS: International Program on Chemical

SafetyMTD: maximum tolerated doseNCI: National Cancer Institute (USA)NIEHS: National Institute of Environmental

Health Sciences (USA)NTP: National Toxicology Program (USA)PB-PD/TD: physiologically-based pharmaco/

toxico-dynamic (models)PK: pharmacokineticPPAR: peroxisome proliferator-activated

receptorQSAR: quantitative structure-activity

relationshipR&D: Research and DevelopmentSERM: selective (o)estrogen receptor

modulatorSULT: sulphotransferaseTNF: tumour necrosis factorUDS: unscheduled DNA synthesisUGT: UDP glucuronosyl transferaseUK: United Kingdom

Abbreviations

safermedicines report - acmedsci.ac.uk › file-download › 34854-51e517b083b77.pdf · as unique...

Documents