Untitled

Report of CIOMS Working Group IV CIOMS publications may be obtained directly from CIOMS, c/o WorldHealth Organization, Avenue Appia. 1211 Geneva 27, Switzerland. Theyare also distributed by the World Health Organization, Distribution and Sales Unit, Avenue Appia, 1211 Geneva 27, Switzerland and are availablefrom booksellers through the network of WHO sales agents. A list ofthese agents may be obtained by writing to the above address.
Benefit-Risk Balance for Marketed Drugs: Evaluating Safety Signals Report of CIOMS Working Group IV The Council for International Organizations of Medical Sciences is greatly indebted to the members of CIOMS Working Group IV, onBenefit-Risk Balance for Marketed Drugs: Evaluating Safety Signals,and to the drug regulatory authorities and pharmaceutical companiesthey represented, for the efficient and expeditious way in which theybrought this project to its successful conclusion. Special thanks are dueto the co-chairs, Dr Win Castle and Dr Murray Lumpkin, for theircapable leadership, to Ms Susan Roden, the secretary of the Group,and to Dr Arnold J. Gordon, editor of the full report, who veryeffectively coordinated, collected and edited the contributions of itsindividual members. We thank also Ms Salpie Leylekian, Dr Gordon'ssecretary, for typing and collating the many drafts of the manuscript,and Dr James Gallagher for his assistance in the editing of the finalreport.
Copyright # 1998 by the Council for International Organizations of Medical Sciences (CIOMS) ISBN 92 9036 068 2 Printed in Switzerland Reprinted 2000, 2004 TABLE OF CONTENTS B. Content, Scope and Definitions . . . . . . . . . . . . . . . . . 12C. General Principles and Limitations . . . . . . . . . . . . . . . 15D. Factors Influencing Benefit-Risk Assessments . . . . . . . . . 16E. The Evidence and Analytical Approaches . . . . . . . . . . . 19 II. STANDARD FORMAT AND CONTENT 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 252. Description of the Epidemiology and Natural History 3. Purpose or Intended Outcome of the Treatment . . . . . 274. Evidence For Benefits: Degree Of Efficacy Achieved in Clinical Trials and Effectiveness in Practice . . . . . 28 5. Alternative Therapies: Comparison of Benefits . . . . . . 296. Alternative Therapies: Comparison of Tolerability, 7. Points to Consider In Evaluation 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 342. General Considerations in an Analysis . . . . . . . . . . . 353. Risk Profiles for Individual Drugs . . . . . . . . . . . . . 384. Risk Weighing for Individual Adverse Drug Reactions . . 415. Quantification of Risk . . . . . . . . . . . . . . . . . . . . 476. Suggested Sequential Overall Approach . . . . . . . . . . 51 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 522. Descriptive and Semiquantitative Methods of Analysis . . 56 5. Modifications to the Product or its Use, 6. Suspension of Product Licence or Investigational-Status Approval . . . . . . . . . . . . 68 7. Withdrawal of the Product From the Market . . . . . . . 698. After the Analysis: Communication of New or Reinforced Information to the Medical Profession or the Public . . . 70 A. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 71B. Basic Approaches To Good Decision-making Practices: IV. UNADDRESSED AND UNRESOLVED ISSUES . . . . . . . 79 OVERALL SUMMARY OF PROPOSALS . . . . . . . . . . . . 83 APPENDIX A. Membership and Activities of CIOMS APPENDIX B. Illustrative Case Histories of Benefit-Risk Evaluations . . . . . . . . . . . . 101 APPENDIX C. Model for Quantification of Risks . . . . . . . . . . 135APPENDIX D. Detailed Example of a Quantitative Benefit-Risk Evaluation: Dipyrone . . . . . . . . 137 APPENDIX E. TURBO Model for Benefit-Risk Analysis . . . . . . 151APPENDIX F. A Survey of Manufacturer and Regulatory Authority Decision-Making Practicesfor Post-Marketing Safety Issues . . . . . . . . . . 155 It is the hope of CIOMS Working Group IV that manufacturers and regulators will endorse and adopt the proposed systematic approach to theevaluation and reporting of the balance between benefits and risks of amarketed medicinal product with a suspected major safety problem.
Strategies are recommended for improving communication between, andthe sharing of action by, regulators and manufacturers, designed tostrengthen decision-making in the interest of public health. We also envisagethat those performing benefit-risk evaluation under these circumstances willsubscribe to the following key principles and practices: Although evaluation and decisions are made on behalf of the at-riskpopulation from a public-health perspective rather than that of theindividual patient, the needs and perspectives of the different stake-holders and constituencies affected will be carefully taken into account.
Although each instance of a safety issue is unique, all parties will adoptconsistent practices in analysis and reporting.
A benefit-risk evaluation should always be conducted relative to notherapy or to properly chosen comparator drugs and other treatments; tofacilitate comparison between alternatives, standard graphical risk-profile representations will be routinely used.
It will normally not be sufficient to evaluate only the effect of the newproblem (‘‘signal'') on the benefit-risk relationship; a re-examination ofthe entire safety profile, or at least of the most prominent/importantadverse drug reactions relative to other treatments, is recommended.
Deciding the appropriate action subsequent to a benefit-risk evaluationrequires ‘‘good decision-making practices''; in the assessment anddecision-making processes the basis and rationale of decisions shouldbe transparent and inter-agency cooperation should be encouraged.
The Working Group hopes also that research will be undertaken in several poorly explored or inadequately developed areas of significance tobenefit-risk weighing, including metrics for weighing the relative severity/seriousness of different adverse drug reactions; expansion of knowledge inthe natural history of diseases in target populations and in the aetiology andmechanisms of drug-induced diseases; personal perception/acceptance ofrisks relative to benefits (how much benefit is needed to accept what level ofrisk); approaches to better quantification of the benefit-risk relationship;and use of formal decision theory in the decision-making process.
In summary, the CIOMS Working Group foresees a harmonized approach to the analysis, reporting and decision-making steps involved ina re-examination of benefit-risk weighing when an important new safetyissue arises. It also envisages further research in this area to improve ourability to refine the results, conclusions and actions taken on behalf ofpatients.
Comments are invited and should be sent to Dr Zbigniew Secretary-General, CIOMS, c/o WHO, Avenue Appia, 1211 Geneva 27, Switzerland.
This is the fifth in a series of contributions from the CIOMS Working Groups on Drug Safety. This collaboration between representatives ofregulatory authorities and drug manufacturers, originating in 1986,represents a commitment to improving and standardizing internationalreporting of adverse drug reactions and other aspects of pharmacovigilance,with a vision of better understanding and communication of mattersrelating to the safety of medicinal products. Although the safety or risksassociated with medicines must always be judged in relation to theaccompanying benefits, there are no standard, systematic approaches tothe deceptively simple notion of weighing or balancing risks againstbenefits, especially when important new information on marketed drugsbecomes available. The prior efforts of the CIOMS Working Groups havebeen a natural evolution towards the development of guidance in thisdifficult area.
CIOMS I (1990) established the now widely adopted practice of using the ‘‘CIOMS I reporting form'' for standardized international reporting ofindividual cases of serious, unexpected adverse drug reactions. Project IAdeveloped a set of standard data elements for individual adverse-reactionreports and their specifications for electronic reporting, which contributedto a similar initiative under the auspices of the International Conference onHarmonization (ICH). CIOMS II proposed a standard for periodic safetyupdate reports, which has been adopted extensively since the publication ofthe report in 1992; it also served as a basis for the development of the officialICH guideline for such reports. CIOMS III (1995) developed proposals forinternational harmonization of the practical aspects of defining, creatingand modifying the sections of data sheets or package inserts that containsafety information. It elaborated the concept introduced under CIOMS IIof a manufacturer's Core Data Sheet for a product and the Core SafetyInformation (CSI) it contains.
CIOMS IV is to some extent an extension of CIOMS II and III. It examines the theoretical and practical aspects of how to determine whethera potentially major, new safety signal signifies a shift, calling for significantaction, in the established relationship between benefits and risks; it alsoprovides guidance for deciding what options for action should beconsidered and on the process of decision-making should such action berequired. This is not a scholarly treatise on benefit-risk analysis. Rather, itsuggests pragmatic approaches to reassessing the benefit-risk relationship,to producing a standard report describing the results of such an effort, andto good decision-making practices.
The comparative evaluation or weighing of benefits (positive effects) and risks (potential harm) of various medical options for treatment, prophy-laxis, prevention or diagnosis is essential. It is done during research anddevelopment on new medical products or procedures (such as surgery), orby a regulatory authority deliberating the approval or withdrawal of aproduct or some intermediate action, by a physician on behalf of a patient,or by the patient. Such weighing, whether implicit or explicit, is at the heartof decision-making in medicine and health care.
This apparently straightforward concept is expressed through such terms as benefit to risk ratio, benefit-risk difference, benefit vs. risk, therapeuticmargin, therapeutic index and others. Regrettably, in spite of common andfrequent use, none of these inexact expressions has been adequately definedor is easily quantifiable with a summary statistic. Moreover, althoughregulators and companies routinely make decisions driven by the balancebetween benefits and risks, however they are measured, there are nogenerally agreed procedures or regulatory guidelines for conducting andacting upon benefit-risk assessment.
Nevertheless, the European Union, for example, requires manufacturers to conduct ‘‘adequate ongoing monitoring and benefit/risk evaluationduring the post-authorization period'' with the aim of ensuring that ‘‘safetyhazards are minimized and the benefits of treatment maximized byappropriate action.''1 Other regulatory bodies make reference to benefit-risk assessment in relation to product development and approval, to theeffect of new adverse-reaction reports on a marketed product's profile (e.g.,Germany), and in relation to pharmacovigilance in general. The WorkingGroup is aware of only one detailed regulatory technical guide on benefit-risk evaluation; it deals with the special topic of carcinogenicity (ReviewersGuide for Risk/Benefit Assessment of Drugs with Carcinogenic Potential,Health Protection Branch, Drugs Directorate, Canada).
Notice to Applicants for Marketing Authorizations for Medicinal Products for Human Use in theEuropean Community, Chapter V, Pharmacovigilance: 5. On-Going Pharmacovigilance EvaluationDuring the Post-authorization Period, December 1994 (Brussels, III/5944/94) and revisions (inpreparation).
Various approaches have been suggested to benefit-risk assessment during the development of new medicines2, 3, 4, 5 but little has been publishedon marketed drugs, though they have much greater immediate conse-quences for the public health than drugs under development. An advisorypanel of the Canadian Public Health Association has prepared reports6, 7 onbenefit, risk and cost management of prescription drugs, including a reviewof benefit-risk methodologies and risk management strategies, as well asrecommendations on communication and awareness programmes for thepublic. Another particularly interesting presentation of ideas and views onbenefit-risk decisions in licensing applications, licensing changes, andpatient care, which included use of actual examples, is based on a 1985workshop of the Centre for Medicines Research (UK).8 Other recom-mended readings are found in the proceedings of two conferences dedicatedto risk perception, risk management and benefit-risk assessment9, 10 and in atreatise by the Royal Society (London) on analysis, perception andmanagement of risk.11 Over the years, limitations have been imposed on the use of some marketed drugs; between 1961 and 1994, 131 products were withdrawnfrom various markets for reasons related to safety.12 Although some of therelevant data are available, rarely are the process and rationale of thedecisions made known. In a recent case (ketorolac and gastrointestinal or Spilker, B. Incorporating Benefit-to-Risk Determinations in Medicine Development, Drug News andPerspectives, 7 (1), February 1994, 53-59.
Chuang-Stein, C.A. New Proposal for Benefit-less-risk Analysis in Clinical Trials, Controlled ClinicalTrials, 15: 30-43, 1994.
Methods and Examples for Assessing Benefit/Risk and Safety for New Drug Applications, Proceedingsof a DIA Workshop, Drug Information Journal, 27: 1011-1049, 1993.
Cocchetto, D. and Nardi, R.V. Benefit/Risk Assessment of Investigational Drugs: CurrentMethodology, Limitations and Alternative Approaches, Pharmacotherapy, 6: 286-303, 1986.
Benefit, Risk and Cost Management of Drugs. Report of the CPHA National Advisory Panel on Risk/Benefit Management of Drugs. Canadian Public Health Association, January 1993.
Tools Employed in the Measurement of the Risks and Benefits of Drugs — A Literature Review for theCanadian Public Health Association's National Panel on Risk/Benefit Management of Drugs. CurryAdams and Associates, Ottawa, Ontario, Canada, July 1991.
Walker, S.R. and Asscher, A.W. Medicines and Risk/Benefit Decisions, MTP Press Limited, London,1987.
Horisberger, B. and Dinkel, R., Editors. The Perception and Management of Drug Safety Risks.
Springer-Verlag, Berlin, 1989.
10 Dinkel, R., Horisberger, B. and Tolo, K. W. Editors. Improving Drug Safety — A Joint Responsibility.
Springer-Verlag, Berlin, 1991.
11 Risk: Analysis, Perception and Management. Report of a Royal Society Study Group. The Royal Society, London, 1992.
12 Spriet-Pourra, C. and Auriche, M. Drug Withdrawal From Sale, 2nd edition, SCRIP Reports, Richmond, UK, March 1994.
post-operative haemorrhage) the same data reviewed by differentMember States of the European Union led to different benefit-riskassessments and regulatory actions.13 Similarly, different actions weretaken in regard to the association of aspirin with Reye‘s syndrome. In theUnited Kingdom, manufacturers voluntarily withdrew paediatric aspirinfrom the home market and added new warnings in respect of otherpreparations; in other countries, including the United States, warningswere added to product information but no products were withdrawn.14 The concept of benefit-risk weighing and its difficulties are not unique to the development and use of medicinal products; much work on this topichas been done in relation to the environment, food additives and chemicals.
Useful models are also available from clinical practice, in which surgery andother modalities in addition to pharmaceuticals have been the subject ofbenefit-risk assessment.
It was against this background that the Working Group decided to focus its efforts in this complex subject on developing some guidance forregulators and manufacturers on assessing the balance between benefits andrisks of marketed products with a newly established or suspected majorsafety problem. The balance could also be altered by the appearance of afavourable or unfavourable change in benefits to patients or by animprovement in the expected risk profile. Such a change is relatively unusualbut its assessment is based on the same principles and methods.
Signals of potential safety problems with marketed medicines represent a broad spectrum of severity and impact on the public health. Most signalswill not merit the type of formal benefit-risk evaluation proposed here bythe Working Group, though the concepts are regarded as generally useful inany periodic or special evaluation of relative benefits and risks. Theappearance of a new but relatively insignificant adverse reaction mayrequire only a routine change in product information (data sheets, labelling,etc.). CIOMS Working Group III developed criteria and procedures forsuch a change, on the basis of strength of evidence and the concept ofthreshold for inclusion.15 13 Summers, K. Crisis Management — Handling a Collapse in Confidence, Scrip Magazine, June 1995, 53- 56. See also, Ketorolac PMS Study Published, SCRIP, No. 2103, February 16, 1996, p. 26.
14 Paediatric Aspirin Withdrawn in UK, SCRIP, No. 1112, June 18, 1987, p. 5. Commission on Reye's/ Aspirin, SCRIP, No. 1180, February 18, 1987, p. 27. ASA/Reye's Syndrome Swedish Warning, SCRIP,No. 1272, January 8/13, 1988, p. 12.
15 Guidelines for Preparing Core Clinical-Safety Information on Drugs, Report of CIOMS Working Group III, Council of International Organizations of Medical Sciences, Geneva, 1995.
When, then, is it necessary or appropriate to undertake the formal process developed here? The Working Group suggests that consideration begiven to whether there has been a ‘‘major shift'' in the balance betweenbenefits and risks. ‘‘Major'' and ‘‘shift'' cannot be defined to fit allmedicines and circumstances; it is always a matter of judgment to decidewhen the threshold for investigation and analysis has been reached. Thisdecision may depend on the unique characteristics of the medicine and itsintended use (e.g., urgency of its indications, availability of therapeuticalternatives) and on the nature of the adverse reaction and the ability topredict, detect, intervene or prevent it. In essence, the higher threshold forsuch a comprehensive re-evaluation than for routine or simple changes inproduct information is reached when a significant change is suspected tohave taken place in the generally understood benefit-risk relationship. Thechange will usually be related to the seriousness, severity or frequency of thereaction and may be expected to result in significant changes in productinformation or marketing status (e.g., new contraindications, substantialrestrictions on indications or treated populations, mandates for furtherresearch, or withdrawal of the product from use).
This broad topic raises many technical issues as well as complex ethical questions about the responsibility of a health authority or a company forassessing medicinal products, a process that ultimately affects an individualpatient. There are no set methods that are applicable to all benefit-riskassessments; each case will be more or less different, but for all cases thereare fundamental approaches and common points to consider, and thisreport attempts to summarize them.
Appendix A lists the members of the Working Group and summarizes its activities over the course of this project.
B. Content, Scope and Definitions In formulating its proposals the Working Group developed, reviewed and made use of actual case histories (Appendix B) taken from theexperience of companies and regulators in several countries: quinine and allergic haematological events (US) felbamate and blood dyscrasias (EU, US) dipyrone (metamizole) and agranulocytosis (US, Sweden) temafloxacin and renal impairment and hypoglycaemia in the elderly(US, EU) remoxipride and blood dyscrasias (Sweden) clozapine and agranulocytosis (Europe, US, South America) sparfloxacin and phototoxicity (EU, Japan, other countries) Chapter II, sections B, C and D, uses these examples and other material to illustrate basic principles and methodologies as well as to suggest ways ofdisplaying data, in connection with benefit estimation, risk estimation andbenefit-risk evaluation. Once an analysis is completed, companies andregulators separately or jointly must consider the various options for action,between the extremes of doing nothing and immediate withdrawal of aproduct from the market (Chapter II, section E). Chapter III covers thedecision-making process on the best options, follow-up and responsibilities;it includes a summary of the results of a small survey of regulators andcompanies, carried out by the Working Group, on decision-makingprocesses and the use of outside experts.
Figure 1 provides a conceptual summary of the possible steps and processes of benefit-risk assessment. The material covered by the WorkingGroup begins with the box labelled Signal Evaluation, in which thesuspected risk is estimated (Chapter IIC), and, if the signal is confirmed, theprocess continues as shown in the Figure and elaborated in Chapters IIand III.
There are no standard, widely acknowledged definitions of the terms benefit and risk as applied to medicine and particularly to medicinalproducts. These and other terms, as used in the context of this report, aredefined and explained in the Glossary.
It is a frustrating aspect of benefit-risk evaluation that there is no defined and tested algorithm or summary metric that combines benefit and risk dataand that might permit straightforward quantitative comparisons ofdifferent treatment options, which in turn might aid in decision-making.
Rarely is it possible to express the relationship between benefits and risksquantitatively with simple units of measure (as a ratio or difference, forexample), although attempts have been made, as Chapter IID indicates.
Nevertheless, no matter how the relationship between benefits and risks of a given treatment is measured or described, the measurement ordescription should never be an absolute one. According to the circum-stances, the comparisons should apply to alternative medicinal products,non-medicinal modalities, or indeed no treatment. They may apply also todifferent doses or dosage forms of the same medication, whether for thesame or different indications, or even to a combination of pharmaceuticaland other options.
C. General Principles and Limitations The Working Group realized there were matters that were not appropriate to the purposes of this report or possible to resolve in its context, though allwere germane to the subject of benefit and risk. Chapter IV, Unaddressed andunresolved issues, summarizes those topics and demonstrates the breadth ofissues that the Group considered. In addition, the following special pointsshould be noted when reviewing this CIOMS IV proposal.
1. The comparative benefit-risk assessments to which this report refers relate to marketed medicinal products. Signals of adverse reactions ofmedical importance arise typically from market experience, usuallythrough spontaneous reports, but may surface also in large post-marketing or other studies. However, information from all sources mustbe used and the same principles apply when the signal for an importantsafety issue surfaces in a clinical development programme in countrieswhere the drug may not yet be on the market.
2. The public-health perspective taken here requires the collection and analysis of data from many and varied sources. One of the more difficultunresolved questions is how such an aggregate approach may beextrapolated and made relevant to a specific patient in specificcircumstances. In that sense, this approach addresses the population atrisk, not the patient at risk. The latter is the province of the physician (whatis the best treatment for an individual under specific circumstances?).
3. A signal may indicate a possible emergency (e.g., an imminent hazard to the population at risk), which calls for crisis management measures.
Although this adds to the pressure on the parties concerned, amethodical approach to a thorough benefit-risk evaluation as discussedin this report is still necessary.
4. Typically, a benefit-risk reassessment is prompted by the appearance of a particular sign, symptom, symptom-complex or diagnosis, and oneautomatically tends to restrict the investigation to the particular signal.
However, in weighing the benefits and risks of a drug against alternatives(or no treatment) in the light of a new, important risk, the impact of thenew information must be reviewed in the total context of the drug's use.
Thus, it will usually be necessary to review its entire benefit-risk profile inthe light of available treatment options (see Chapter IID).
5. The nature or source of the data (e.g., from clinical trials, spontaneous reports, observational studies, non-clinical experiments) will determinewhich of the different available methodological and analyticalapproaches will be used to estimate and analyse benefits and risks. This report places more emphasis on the estimation of risks than of benefits;the reader is referred to standard texts and articles, such as those cited,for the usual details.
6. In addition to problems that may be related to the intrinsic properties of the active ingredients, metabolites or excipients of a medicine, medicallyimportant risks may arise from accidental or intentional contamination orproduct defects (e.g., poisoning of Tylenol (acetaminophen/paracetamol)capsules in the United States; toxic contaminant in L-tryptophan in Japan;or inadequate potency or lack of sterility due to a manufacturing flaw).
Although the cause may not be apparent at the onset of the initial signal —and therefore it may appear that a shift has occurred in the benefit-riskprofile — once its source has been identified and corrective stepsundertaken, a formal benefit-risk assessment may no longer be indicated.
7. Although economic considerations (e.g., cost-effectiveness) exert an increasingly pervasive influence in the choice and use of medicines withinhealth care systems, the Working Group believes that they should notinfluence the types of benefit-risk assessment covered here. See D.4.
(below) and Chapter IV for more discussion on this point.
8. A benefit-risk assessment is only as good as the underlying data and their treatment. Although a manufacturer may have access to most or all of theextant data on its own product, it often has difficulty obtainingcomparable information on the other treatment options under considera-tion, particularly if the data are unpublished. Especially for medicallyserious conditions it is important that the concerned parties, usuallycompanies and regulators, cooperate as much as possible and promptlyexchange the needed information. Within the limits of proprietary andother confidential interests, therefore, regulators are expected to share allpertinent information with the companies concerned, toward theircommon goal of resolving the suspected problem. Similarly, transparencyin decision-making will not only enhance the credibility of the result butalso provide the experience and understanding much needed by everyoneengaged in this complex undertaking (see Chapter III).
D. Factors Influencing Benefit-Risk Assessment Even when the best and most complete data are available there can be various subjective and objective influences on the way benefit-risk is assessed,on the urgency with which the assessment is needed, and on the options fordecisions and actions considered at the end of the process. The followingexamples cover some prominent influences on the evaluation process.
1. Stakeholders and constituencies The main thrust of this report is to provide guidance for manufacturers and regulators, who bring a certain perspective based on their goals andresponsibilities but who are ultimately acting on behalf of others. Thus, theymust be aware of and sensitive to the possibly different perspectives thatvarious stakeholders, who may weigh the information differently, bring to abenefit-risk analysis. Patients, physicians, pharmaceutical companies, ethicscommittees, regulatory authorities, other public health bodies, insurers,consumer groups and others may have very different perspectives2. This isparticularly true of the relationship between benefits and risks; for example,two patients exposed to (and knowing about) the same benefits and risks mayhave different perceptions of, or thresholds for accepting, the risks, andthereby make different choices. An interesting analysis of this issue, withexamples of the use of triazolam (a benzodiazepine hypnotic) and isotretinoin(a retinoic acid for cystic acne), has been published6.
As discussed elsewhere in this report, it is usually advisable that some of the affected parties take part in the assessment.
2. Nature of the problem The speed and intensity with which a benefit-risk evaluation is conducted will depend on the medical seriousness of the suspected adverse drug reactionas well as on other considerations (see 3. below). A potential threat to life, forexample, clearly demands very rapid attention, with close and frequentinteraction between the regulatory authorities and the company. Althoughthe nature of a problem will usually dictate the time-course of action, it mustbe emphasized that all benefit-risk assessments are made in the face ofuncertainty. When a major new potential safety-problem arises, the need forurgent action to protect the public from a possibly serious hazard must beweighed against the need for additional data that might provide morecertainty or confidence in the analyses and conclusions (See Chapter IIE).
3. Indication for drug use and population under treatment A drug may be used for prevention, prophylaxis or treatment and possibly for diagnostic purposes. It may be given for a life-threateningcondition or a self-limiting disease; the acceptable risk for the former willundoubtedly be the higher. If it is an ‘‘orphan drug'' used in limitedpopulations or other circumstances for which there are no reasonabletherapeutic alternatives the threshold for acceptable risk may be relativelyhigh. Certain target sub-populations for the medicine may respond differently owing to ethnic differences in pharmacodynamics or pharma-cokinetics or to medical-cultural influences. Also a safety or other problemmay occur in association with an unapproved (unlabelled) use.
These and other examples remind us that comparative benefit-risk analyses can be highly situational, and may be especially difficult when asingle product falls into more than one of these categories (e.g., multipleindications or different dosage forms/routes of administration, withdifferentiable benefit-risk properties).
4. Constraints of time, data and resources The proposals in this report call for a systematic, thorough evaluation of all available data when a signal suggests that a product carries a potentiallymajor new risk.
As indicated above (II.C.4.), such an evaluation should take into account not only the new signal but also the overall risk profile of the productrelative to that of the appropriate comparator products. In practice,especially when regulatory needs demand urgent attention (e.g., underconditions of a Type II Variation or ‘‘Urgent Safety Restriction'' in theEuropean Union), such a comprehensive, comparative benefit-risk analysismay be difficult. This is especially the case if sufficient data on comparatordrugs or other modalities (e.g., surgery) cannot be obtained reasonablyquickly, if the data-bases are very large (e.g., hundreds or thousands ofadverse-reaction cases that might require review), and if the number ofpeople available is limited. It would then be advisable to hold discussionswith the authorities to decide on the scope and strategy of the evaluationand the written report.
5. Economic issues Some economists have suggested that ‘‘economic efficiency'' (difference between benefits and costs) should be a fundamental criterion for decisionson health and safety regulation, and have offered principles for the use ofbenefit-cost analyses16 or benefit-risk-cost analyses6. For example, evenbefore the widespread use of serotonin re-uptake inhibitors, Henry andMartin17 in a benefit-risk analysis of many antidepressants counted the lowcost of older drugs as part of their benefits.
16 Arrow, K.J., Cropper, M.L. Eads, G.C. et al. Is there a Role for Benefit-Cost Analysis in Environmental, Health and Safety Regulation? Science, 272-221-222, 1996.
17 Henry, J.A. and Martin, A.J. The Risk/Benefit Assessment of Antidepressant Drugs, Medical Toxicology, 2: 445-462, 1987.
Although economic impact on the allocation of medicines is receiving greater attention, the Working Group did not consider economic outcomesin terms of financial considerations in the types of benefit-risk assessmentcovered by this report. However, the Group did acknowledge and invokeimportant concepts and methods of economics science, such as trade-offsand weighings.
Once it is established that conventional benefit-risk evaluation shows no significant differences between different products, cost may be one basis forchoosing among them for a patient who responds equally to them. It mustalso be acknowledged, however, that in economically depressed areas amedicine with a less advantageous benefit-risk balance may be the onlyaffordable alternative. Willingness to accept risk may therefore depend oncost in many parts of the world, where health regulators and otherauthorities may indeed regard it as a normal aspect of decision-making.
E. The Evidence and Analytical Approaches The treatment of benefit evaluation and risk evaluation in Chapter II includes examples of the kinds and amounts of data and their interpretationthat should be brought to bear in benefit-risk assessment. Some generalconsiderations are appropriate here, along with some discussion ofcombining benefit and risk metrics.
1. Information and its limitations When a signal arises that leads to a re-evaluation of the benefit-risk profile, a manufacturer or regulator will attempt to gather as muchinformation as possible from animal, in vitro and human investigations orexposure. Evidence from both empirical and non-empirical sources willusually be available. For a marketed drug, much of the data come fromuncontrolled sources of surveillance (e.g., spontaneous reports) and areoften factually uncertain, incomplete and imprecise.
Controlled clinical trials and observational studies provide data that are more reliable, but some caveats are indicated. A clinical-trial data-basetypically available for benefit-risk evaluation is mostly the result of drugdevelopment programmes designed to demonstrate and assess efficacy.
Such programmes invariably include carefully selected populations oflimited size (usually below 10,000 patients); conditions are highly controlledand treatment periods relatively short; the object is to reach statisticallymeaningful efficacy-endpoints for the purpose of regulatory approval.
There are thus inherent limitations regarding generalizability of bothefficacy and safety to ‘‘real world'' populations, as well as inadequate powerto detect relatively rare but potentially significant adverse reactions. Theseand other difficulties with randomized clinical trials have been extensivelydescribed.18 Nevertheless, there are useful strategies for distinguishingbetween drug-related and non-drug-related adverse events.19 Structured post-marketing surveillance follow-up and ad hoc case- control studies are capable of examining larger populations in more‘‘naturalistic'' circumstances and over longer periods of drug exposure thanmost randomized controlled clinical trials. Some of these studies can beperformed with large automated data-bases of records collected foradministrative, clinical or other purposes, supplemented with other clinicalrecords as needed, to validate exposures and diagnoses. Researchers mayneed to anticipate bias in the selection of patients treated in any non-randomized studies, and to design the studies and analyses carefully.
However, because of their larger potential sizes and ‘‘real world'' settings,these studies may represent the only opportunity to quantify the risk of aparticular adverse event, to evaluate predictors of risk, and to estimate riskin populations typically excluded from clinical trials (e.g., pregnant women).
Nevertheless, such studies usually cannot provide primary evidence for causation and can indeed lead to false-positive associations.20, 21 An excellentexample of the inherent difficulties in this area is the ongoing debate aboutthird-generation oral contraceptives and venous thrombosis.22 Ideally, for comparisons between treatment alternatives, like types and amounts of data should be available for all the options under consideration.
In practice, this is rarely the case and there are frequent gaps in the quantityand quality of information. Moreover, in comparing pharmaceuticals withother options (e.g., surgery or other modalities) there are additionaldifficulties (different measurements and scales, and different medicalcultures and expertise). Finally, in deriving a measure of the relationship 18 Laupacis, A., Sackett, D. and Roberts, R. An Assessment of Clinically Useful Measures of the Consequences of Treatment, New England Journal of Medicine, 318: 1728-1734, 1988.
19 Hsu, P.-W., Pernet, A.G., Craft, J.C. and Hursey, M.J. A method for Identifying Adverse Events Related to New Drug Treatment, Drug Information Journal, 26: 109-118, 1992.
20 Feinstein, A.R. Meta-analysis: Statistical Alchemy for the 21st Century, Journal of Clinical Epidemiology 48: 71-78, 1995.
21 Milloy, S. Science Without Sense. The Risky Business of Public Health Research, Cato Institute, Washington, DC, 1995.
22 Hughes, S. More evidence in favor of third-generation oral contraceptives, SCRIP, No. 2279.
October 28, 1997, p.7.
between benefits and risks, value judgments are needed for giving weightor priority to the undesirability of adverse effects (e.g., ‘‘seriousness'')and to the desirability of beneficial effects. Such judgments can be highlysubjective and will vary among health professionals and patients.
It is important, therefore, to obtain data from all relevant sources, to document carefully all the information, and to record the assumptionsadopted and any limitations in the use of the data for analysis. If there areconsiderable uncertainties, such as those mentioned here, sensitivityanalyses may be helpful in determining the robustness of the assessmentor whether the conclusions drawn from various models differ in anysignificant respects.
2. Benefit-risk metrics If the benefits of the various options under consideration can be assumed to be equal, benefit-risk evaluation can rely on measures of relative risk.
Otherwise, in the absence of a readily available and quantitative relation-ship between benefits and risks, which is commonly the case, evaluationusually comes down to analyses and conclusions that rely on indirect,informal and unavoidably subjective processes. There are no acceptedgeneral methods for deriving a ‘‘benefit-risk ratio'' or another compositemetric, or for using such measures to compare relative merits of alternativetreatments. As ordinarily used, therefore, the benefit-risk ‘‘ratio'' comparesfiguratively, but not often quantitatively, the relative magnitudes of benefitsand risks. Some hold that it is more meaningful to speak of benefit-riskdifference, the net value a patient (or population) can expect from a therapy.
Selected attempts described in the literature and some new approachesconsidered by the Working Group are presented in Chapter IID.
At a minimum, an attempt should be made to quantify the incidence of the event in the treated population, if possible by age, sex and othercharacteristics of significance for the medicine's indication.
II. STANDARD FORMAT AND CONTENT OF A BENEFIT-RISK EVALUATION REPORT The Working Group recommends a standard outline for a written report to describe the results of a benefit-risk evaluation. The outline and check-listshown below are regarded as a reasonable approach toward such a standardfor a self-contained (‘‘stand-alone'') report. The items listed are meant toprovide a preview and abstract of the more fully developed themes insections IIA through IIE, which follow the outline and which discuss thebasis for the content of a report.
OUTLINE OF A REPORT Brief specification/description of the drug and where marketed Indications for use, by country if there are differences Identification of one or more alternative therapies or modalities,including surgery A very brief description of the suspected or established major safetyproblem B. Benefit evaluation Epidemiology and natural history of the target disease(s) Purpose of treatment (cure, prophylaxis, etc.) Summary of efficacy and general toleration data compared with: – other medical treatments – surgical treatment or other intervention C. Risk evaluation Weight of evidence for the suspected risk (incidence etc.) Detailed presentations and analyses of data on the new suspected risk Probable and possible explanations Preventability, predictability and reversibility of the new risk The issue as it relates to alternative therapies and no therapy Review of the complete safety profile of the drug, using diagrammaticrepresentations when possible (‘‘risk profiles''); when appropriate, focuson, e.g., the three most common and the three most medically seriousadverse reactions Provide similar profiles for alternative drugs When possible, estimate the excess incidence of any adverse reactionsknown to be common to the alternatives When there are significant adverse reactions that are not common to thethe drugs compared, highlight important differences between the drugs.
D. Benefit-risk evaluation Summarize the benefits as related to the seriousness of the target diseaseand the purpose and effectiveness of treatment Summarize the dominant risks (seriousness/severity, duration, incidence) Summarize the benefit-risk relationship, quantitatively and diagramma-tically if possible, taking into account the alternative therapies or notreatment Provide a summary assessment and conclusion E. Options analysis List all appropriate options for action Describe the pros and cons and likely consequences (impact analysis) ofeach option under consideration, taking alternative therapies into account If relevant, outline plans or suggestions for a study that could providetimely and important additional information If feasible, indicate the quality and quantity of any future evidence whichwould signal the need for a re-evaluation of the benefit-risk relationship Suggest how the consequences of the recommended action should bemonitored and assessed.
A. Introduction to a Report The introduction should briefly describe the medicine and the report in a way that ensures that reviewers will interpret correctly the scope of thedocument. Reference should be made not only to products covered but alsoto those excluded. It is also important to mention whether data are includedfrom other parties, such as co-marketers or licensees/licensors, and whetherit is known that they are preparing separate reviews.
One of the first steps in a comparative benefit-risk assessment is to specify the alternative therapeutic products. Ideally, for fair comparison,the alternatives will have the same indications and durations of treatment.
Comparisons will also try to match, or at least define and account for, theseverity of the disease treated even for the same indication, concomitantmedications, and age and sex distribution of the affected populations. Inaddition to alternative medicines, the comparison could include surgery,other treatment modalities or no treatment. Thus, the therapies selected forcomparative analysis should be specified, along with the rationale of theirselection. It may be necessary or desirable to obtain the advance agreementof the regulatory authority on the selections.
B. Benefit Evaluation A benefit-risk analysis logically begins with a discussion of the benefits, since the beneficial effects are the basis for use of a drug or vaccine. Thebenefits should be described and wherever possible quantified in a way thatis comparable to the quantification of risks (e.g., potential lives saved as aresult of treatment vs. potential lives lost as a result of adverse reactions).
Benefit (and risk) may be defined in terms of the individual being treated, ofnet benefits across individuals being treated or, as in the case of vaccines orantibiotics, of the net benefit to society.
The discussion of benefits should include consideration of the epidemiology and natural history of the disease being treated, the purposeor intended outcome of treatment, the evidence upon which the benefitshave been established, and the availability of alternative therapies. Thosepoints are discussed in detail below. The various points to be considered in abenefit evaluation are summarized at the end of this section.
The terms benefit and efficacy have customarily been regarded as synonyms; their expression has been derived from metrics used in clinical trials. However, this concept of benefit has been extended to embraceadditional measures such as quality of life, compliance with therapy,outcomes, and the notion of ‘‘evidence-based medicine.''1 Although thesemetrics have so far been little used in benefit-risk analysis, such measures asquality of life are likely to become increasingly important.2 For illustrative purposes, and because it could be misleading to consider benefit in isolation, some of the cases described here include discussions ofaspects of risk and benefit-risk (covered in more detail in II.C. and II.D.,respectively).
2. Description of the epidemiology and natural history of the target disease A description of the epidemiology and natural history of the target disease is usually helpful for putting both benefits and risks into perspective. Theepidemiology may be described in terms of the incidence or prevalence of thetargeted disease as well as of the specific populations at increased risk (e.g.,based on age or sex, renal or hepatic insufficiency, etc.). The natural history ofthe disease is important for differentiating self-limiting conditions (e.g., thecommon cold or trauma) from chronic and progressive disease states (e.g.,diabetes, coronary heart disease, cancer, AIDS) or underlying disease withintermittent acute exacerbations (e.g., asthma, multiple sclerosis, gout).
Tolerance of risk will generally be much greater for therapies that changethe natural history of acute or chronic conditions associated with significantmorbidity or mortality (e.g., HIV disease, coronary heart disease, orosteoporosis) than for those that provide symptomatic relief in otherwiseself-limiting conditions not known to be associated with significant morbidity.
The Lennox-Gastaut syndrome is a severe epileptic encephalopathy refractory to conventional treatment. In France there are 4000 new cases ofepilepsy a year in children, of which 30% are resistant to therapy and 1%are severe enough to be considered Lennox-Gastaut. The cumulative annualincidence of sudden death among patients with the syndrome is 1 in 500.
Felbamate (Appendix B) is an antiepileptic drug that has been found todramatically reduce morbidity and mortality in patients with the syndrome,for whom there is no alternative drug. Felbamate is associated with severeblood dyscrasias, including aplastic anaemia. Surgery has been used Taubes, G. Looking for the Evidence in Medicine. Science, 272: 22-24, 5 April 1996.
Tools Employed in the Measurement of the Risks and Benefits of Drugs — A Literature Review for theCanadian Public Health Association's National Panel on Risk/Benefit Management of Drugs. CurryAdams and Associates, Ottawa, Ontario, Canada, July 1991.
successfully in the treatment of the syndrome but it too carries some risk.
In this example, given the life-threatening nature of the disease, it may beappropriate to use a therapy known to be associated with significant risk,as long as the risk is lower than that of any available alternative,including no intervention.
3. Purpose or intended outcome of the treatment The threshold of acceptable risk will vary with the purpose and intended outcome of the treatment. It is important, therefore, to describe clearly itsintended impact on the defined natural history of the disease being treated.
From a regulatory perspective, this discussion will ordinarily be confined tothe approved indications.
Tolerance for a drug-associated risk will generally be low when the drug is intended to prevent disease in an otherwise healthy person; the degree ofacceptable risk will be influenced by the extent to which the disease and itsassociated morbidity and mortality can be prevented. Tolerance willgenerally be greater for a product intended to prevent serious or fatalcomplications of existing disease. For example, the tolerance for riskassociated with cholesterol-lowering agents will be low for their use inpatients without confirmed coronary heart disease, whose only risk factormay be moderately elevated serum cholesterol, but it will be somewhat higherin the case of patients with a history of myocardial infarction and at high riskof its recurrence. Since the natural history of HIV disease is such that virtuallyall of those infected could be expected to die of it before promising new drugsbecame available, the tolerance for risk associated with current drugs will berelatively high as long as there are no better alternatives.
Vaccines to prevent childhood diseases are instances of interventions in otherwise healthy populations, in which risk-tolerance will be low. The caseof vaccines is unique in that the benefit-risk evaluation will generally requireconsideration of the societal as well as the individual perspective. Hence, thedescription of the benefit would have to take into account the overall effectsof vaccination programmes on occurrence of disease in entire populations.
Likewise, choice of antibiotics would take into account the risk oforganisms becoming resistant, which would pose a risk to future patientsand alter the future beneficial effects of the product.
When a safety issue is so significant that it warrants a reassessment of the benefit-risk relationship, it usually refers to a potentially life-threateningadverse reaction. Hence, in terms of comparability, benefit-risk evaluationof medicinal products intended to reduce morbidity and mortality is likely to be relatively straightforward — i.e., comparison of morbidity andmortality with and without the intervention. However, when a drug isintended to be palliative (e.g., for post-surgical pain), to relieve chronicsymptoms (NSAIDs for chronic arthritis), to reduce the frequency orexacerbation of intermittent symptoms (treatment of chronic asthma), or toreduce the risk of treatable complications (post-surgical prophylacticantibiotic use), the benefits and risks are not so easily compared. In suchcircumstances, measures of quality of life take on greater importance.
In addition to the intended purpose or outcome, the intended place of the product in clinical practice must be made clear. Certain products may havesecond-line status owing to previously identified risks or limited efficacy/effectiveness or, as in the case of some antibiotics, because their use shouldbe curtailed to preserve their benefits for future patients by reducing theopportunity for resistance to occur. While certain risks may not beacceptable for first-line therapy when a safer product is available, the samerisks may be acceptable when there is no safer or effective alternative. In theexample above, felbamate is an accepted treatment for Lennox-Gastautsyndrome, where patients are by definition resistant to other drugs, but itwould not be acceptable as first-line therapy in uncomplicated epilepsy.
4. Evidence for benefits: Degree of efficacy achieved in clinical trials and effectiveness in clinical practice From a population-based perspective, and because of the societal implications, the number or proportion of deaths prevented or patientsspared or cured of a disease by the drug or vaccine is of paramountimportance. In general, the degree to which the product achieves such anintended outcome should be described. This would include a review of bothits efficacy (effect under ideal clinical conditions) and its effectiveness (effectunder usual conditions of clinical practice).
Although it is customary to express benefits of preventive therapies in terms of percentage reduction in mortality or another predefined endpoint(e.g., reduction of AIDS-related diagnoses with antiretroviral agents), sucha metric can be highly misleading unless accompanied by data on theabsolute values describing the treated and untreated (or active comparative)group. This is especially so when such benefit statistics are being comparedwith those of product-associated risks. For example, rather than statesimply that a product compared with no treatment achieved a 33%reduction in the incidence of some medical event, the absolute rates for thecompared groups should be given as well (e.g., a background fatality rate of30% compared with a treated rate of 20%, which represents a 33% reduction from the background rate or, expressed in another way, anabsolute reduction of 10%). It may also be informative to describe apreventive agent in terms of the number of individuals who would need tobe treated, and for how long, to prevent one morbid or fatal event.
For products intended for treatment of established disease, the measurable effect will vary according to whether the treatment results indecreased mortality, decreased morbidity, symptomatic improvement orimproved quality of life. For symptomatic or palliative treatment,measurement of benefit would likely be in terms of extent and duration ofsymptom relief or improvement.
Often, though less commonly in recent years, drugs are approved on the basis of surrogate markers of disease and disease outcome. For example,antihypertensive agents are generally approved on the basis of clinical trialsshowing efficacy in terms of lowering of blood pressure rather than reductionin cardiovascular events. In this case, the effect of the drug on morbidity andmortality will be determined by epidemiological data or clinical trials of otherantihypertensive agents rather than by direct evidence. Although this willgenerally be acceptable for long-established therapies, data from studies thatdirectly measure the impact of a particular drug on morbidity and mortalitywill always be considered stronger evidence of benefit than extrapolation ofresults from epidemiological studies or other interventions.
Any discussion of benefits should take into account the degree to which they have been demonstrated, the degree of certainty of results of clinical trials, andthe generalizability of clinical trials to the broader population targeted fortreatment. If quality of life, utilities, quality-adjusted-life-years, psychosocialfactors or functional status are used, the methods used to measure them and thesteps taken to validate the measurements should be defined.
In the United States, quinine was available over-the-counter for treatment of nocturnal leg cramps, a relatively benign and otherwise self-limitingcondition of unknown aetiology. Clinical trials did not support its efficacy:the analyses were flawed, or several confounding factors were present, orthere was no demonstrable difference between quinine and placebo. Hencewhen it became apparent that quinine may be associated with allergichaematological events, the lack of evidence of clinical benefit was adetermining factor in the decision to remove it from over-the-counter use.
5. Alternative therapies: Comparison of benefits Included in the comparison should be a discussion of alternative therapies and their relative efficacy and effectiveness, indication by indication if there is more than one. If there is no alternative therapy, theappropriate comparison would be with no treatment or non-drug modalitiesor both. In this case, a particular effort should be made to understand thenatural history of the disease being treated. If there are other viablealternative therapies, all of their benefits (and risks) should be compared withthose of the medicine in question. The validity of any such comparison shouldbe considered. Ideally, but unusually, one would know of clinical trials withhead-to-head comparisons of all alternative therapies. Instead of directcomparisons, data should be sought from studies carried out in comparablepopulations and using comparable methods of data collection and analysis.
Without such data, any valid comparison of therapies will be difficult. Anyattempt to make such a comparison should include a discussion of thelimitations and potential pitfalls of the comparative analysis.
Oral polio-vaccine is an example of a preventive product which, in the context of mass immunization, is responsible for the near eradication ofpolio from many developed countries. Polio is associated with significantmorbidity and mortality. The live attenuated oral vaccine is moreefficacious than the alternative, killed vaccine. However, the United Statesin recent years has had more cases of polio caused by the live attenuatedvaccine than by wild polio virus. This is why the recommendations for poliovaccination have recently been changed, from a course of three doses of thelive, attenuated oral vaccine to a course of two doses, followed by two dosesof killed vaccine. This practice is expected to prevent half of the eight to tencases of vaccine-induced polio annually in the United States. It is also anexample of how recommendations may be changed as new informationbecomes available to alter the benefit-risk balance.
Temofloxacin provides an example of a drug that was used primarily to treat infections of the urinary and respiratory tracts but was removed fromthe market because of unacceptable risks. Such infections can be life-threatening if left untreated, especially in susceptible populations, but thereare equally efficacious alternative drugs. Of course, the choice of alternativedepends on its relative risk, as well as benefits, as indicated in the nextsection.
6. Alternative therapies: Comparison of tolerability, convenience and preference Because it is likely to affect compliance and therefore overall effectiveness and quality of life, tolerability in relation to alternativetherapies should also be discussed. Other considerations when comparing relative benefits may include route of administration (e.g., oral orsubcutaneous), frequency of dosing, palatability, or other factors thatrelate to convenience or patient preference.
Schizophrenia is a serious chronic disease with a 1% lifetime prevalence; about half of the patients suffer chronic disablement andabout 10% commit suicide. Remoxipride, an antipsychosis agent of thebenzamide group, was approved in Sweden for the treatment ofschizophrenia as well as other types of psychosis in which delusions,hallucinations and thought disturbances are prominent symptoms andwhere the classical neuroleptics have been found to cause intolerable side-effects. In clinical trials no difference in efficacy was observed betweenremoxipride and haloperidol, thioridazine or chlorpromazine. Shortly afterits approval, remoxipride was removed from the market owing to a higherthan expected incidence of aplastic anaemia. However, the generaltolerability profile of remoxipride was better, with fewer drop-outs andless sedation, than that of haloperidol, and it caused less sedation andfewer autonomic side-effects than chlorpromazine or thioridazine. On thisbasis, some have argued that there is still a place for this drug in thetreatment of schizophrenia.
7. Points to consider in evaluation of benefits — A check-list The epidemiology and natural history of the target disease What is the incidence/prevalence of the disease or condition? Have any high-risk populations been identified? Is the disease self-limiting, or fatal or disabling, or one with considerable morbidity or is the condition being treated an asymptomatic riskfactor for subsequent disease? If the condition being treated (e.g., hypercholesterolaemia or hyperten- sion) represents a precursor or risk factor for another disease orcondition (e.g., coronary heart disease or stroke), how well does therisk factor predict the occurrence of the more serious disease? What is the potential impact (i.e., the benefit) of the therapy on the If the disease itself is the direct target of treatment, what is its associated morbidity and mortality? What is the potential impact of early intervention on the disease? What are the consequences of no intervention? Are there prognostic factors to be considered in relation to disease Purpose or intended outcome of the treatment Is the purpose of the treatment: to prevent disease (e.g., vaccine)? to prevent the recurrence of disease (e.g., antibiotic prophylaxis forotitis media)? to treat an acute condition (e.g., streptococcal pharyngitis)? to treat symptoms of a self-limiting condition (e.g., a decongestant)? to reduce the risk of a serious outcome (e.g., treatment ofhypertension, hypercholesterolaemia or osteoporosis)? to prevent progression of disease (thrombolytic therapy)? to treat chronic disabling symptoms (e.g., NSAIDs for chronicarthritis)? to reduce or delay morbidity or mortality among patients (e.g.,treatment of AIDS or cancer)? Are there prognostic factors to be considered in relation to the expected response to treatment? Is benefit more appropriately defined in terms of: the individual being treated net benefits across individuals being treated (e.g., cholesterol-lowering) net benefit to society (e.g., vaccines, antibiotics)? Is the product recommended as first-line or second-line therapy (second- line when other first-line therapies have failed)? Evidence of benefits: Degree of efficacy achieved in clinical trialsand effectiveness in clinical practice What is the evidence of efficacy (effect of the intervention under ideal clinical conditions)? What is the evidence of effectiveness (effect of the intervention under usual conditions of clinical practice)? For preventive agents: to what extent is the risk factor for the disease affected and what is theassociated reduction of disease risk? how many would need to receive the product, and for how long, toprevent one person from experiencing a morbid or fatal event? For disease treatment, what are the measurable effects of therapy, such as decreased mortality, decreased morbidity, symptomatic improve-ment, improved quality of life? For treatment of symptoms, to what extent and for how long are symptoms improved and in what percentage of patients? Do thedemonstrated effects represent only surrogate markers of disease anddisease outcome (e.g., blood sugar, blood lipids)? If so, what is theevidence that the surrogate measures are valid markers? How are beneficial effects measured, what is the quality of the available data, and what is the relevance of the measurements to outcomes thatpatients consider important? What is the level of certainty in interpreting results of clinical trials? Are the results of clinical trials generalizable to the broader population targeted for treatment? Alternative therapies If there is no alternative therapy, what is the effect of no treatment or a non-drug intervention? (See discussion of natural history.) If there are other viable alternative therapies: are data available on comparable populations from studies utilizingcomparable methods, so that the therapies may be validly comparedas to efficacy and effectiveness? if so, how do the alternatives compare in these respects with thetreatment in question? are there other important factors to be considered in relation toalternative therapies, such as tolerability, convenience or patientpreference? C. Risk Evaluation The benefits and risks of medicinal products are continuously evaluated during drug development. An unfavourable benefit-risk balance in thedevelopment phase usually results in cessation of research. Before market-ing, clinical trials are the exclusive source of data on efficacy in humans butthey also provide high-quality comparative data on the risks of the drugsbeing developed. Adverse reactions that are linked to the pharmacologicalaction of a medicinal product are predictable, usually dose-dependent in arather straightforward way, and likely to occur frequently (e.g., sympto-matic hypotension during treatment of hypertension, hypoglycaemia ininsulin-treated diabetic patients, or bradycardia with beta-blockers). Otheradverse effects, such as allergic and idiosyncratic reactions, are neitherpredictable nor dose-dependent and occur mostly in persons who may havea special sensitivity.
In contrast to the benefits of medicines, usually represented as one or more than one well-defined outcome, risks usually include a mixture ofadverse reactions of different types. The medical impact of an adversereaction is characterized by its frequency of occurrence, duration andintensity. Different reactions are not comparable unless they can beexpressed by a common health-outcome measure. Thus, drug-risk evalua-tion requires a multifactorial approach in order to determine the qualitativeprofiles of different adverse reactions (including the signal reaction ofconcern), their frequency of occurrence, and, if possible, the one or morehealth outcomes common to different reactions. Only such a compositeapproach allows a fair risk-comparison of therapeutic alternatives.
Although the quality and amount of available information on therapeutic alternatives may vary, all that is available, preferably as of adefined date, should be compiled. The absolute and relative risks are neverformally analysed with total knowledge of all their parameters. It istherefore crucial that the best estimates possible be obtained from allavailable information by means of analytical and statistical tools appro-priate to the specific circumstances. All qualitative and quantitativeassumptions must be explicitly documented. Risk evaluation usuallyunderestimates the actual risk of previously unidentified adverse reactionsassociated with marketed medicinal products, owing in part to under-reporting. It should be updated, therefore, when sufficient new informationbecomes available. Cut-off dates and key words (such as medical subjectheadings) used in a literature search must be carefully documented.
Potentially fatal reactions such as anaphylactic shock, agranulocytosis or acute renal failure are quite rare. They have been documented as occurringin far less than 1 in 5000 patients receiving therapy. Since pre-marketingclinical trials usually include at most a few thousand subjects, they areunlikely to reveal potentially fatal adverse reactions. After approval, whenmedicinal products are used in much larger populations, such rare reactionsmay become evident. Different methods can be used for the surveillance ofrisks; spontaneous reports and case series from the literature have beenamong the most common and useful in generating signals.
Once signals are identified, the overall risk attributed to a medicinal product must be reassessed and quantified, and compared with that for itstherapeutic alternatives. For this purpose, it is usually necessary to usepopulation-based data from which to determine the incidence of expectedcases over a designated observational period in a defined population ofpatients taking the product. The handling of such data for the evaluation ofdrug-attributed risks is the domain of pharmacoepidemiology. Usuallywhen reports of such cases are first published there are no epidemiologicaldata on the frequency of rare but potentially serious adverse reactions.
Epidemiological studies are important, however, and perhaps necessary, fortesting hypotheses on a new signal.
The methodology of risk evaluation is described below in terms of the basic elements and techniques that are expected to yield a reproducible andtransparent quantification and description of risks attributable to drugs.
2. General Considerations in an Analysis History and description of specific adverse drug reactions promptingthe benefit-risk evaluation It is necessary to begin with a history of how the new safety issue came to light and to follow with a thorough review of the evidence and whether theevidence is sufficient to ascribe causality. In addition to the nature, severity,associated outcome (morbidity and mortality) and duration of an event,many other factors have to be considered. They include any dependence ofthe temporal relationship on use of the drug (e.g., dose or duration oftreatment), physico-chemical characteristics of the drug (e.g., fat solubility,protein build-up, enzyme inhibition/activity, metabolism, relationship to fastacetylation and genetic polymorphism), the possibility of a class effect, apossible effect of concomitant treatment (including the occurrence of drug-drug or other interactions), the background incidence of the event, possiblecorrelations of such variables as demographics and concomitant diseases,and supporting evidence from clinical trials or animal studies.
Moreover, what were the circumstances under which the event became known? For example, a report of a cluster of cases, stimulated by a publicnotice or publication, should be put into proper perspective. In certainunusual circumstances, causes other than the chemical entity should beconsidered, such as when reports are inconsistent with the known safetyprofile of an established product, or there is an unusual lack of response to atreatment for life-threatening illness (an indirect safety issue), or there is nobiologically plausible mechanism for producing the reaction. Such causescould include a defective product or foul play, including tampering with,counterfeiting or sabotaging the product. Other causes may relate toexcipients, contamination or potency.
All these and other considerations relate to the strength of the evidence in assessing a signal, as described in detail in the report of CIOMS WorkingGroup III.1 Finally, great care must be taken in any attempt to extrapolate the results beyond the specific populations (perhaps even geographic locations) andalternative products under consideration.
Preventability, predictability and reversibility of the reaction Another aspect of risk evaluation relates to the events before and after a reaction occurs — in other words, the natural history of the reaction.
Although this may not affect directly the determination of the risk ofoccurrence of an adverse reaction, it can affect the subsequent comparativeanalysis of overall benefit-risk among alternative therapies and must beborne in mind. How difficult is it to treat the reaction? Is it easily reversible?What are its pharmacological properties in the presence of concomitantmedications or concurrent diseases? Could it have been prevented byappropriate labelling changes or education about dosing and administra-tion? Are there early warning signs or symptoms that could signal the needto discontinue treatment, especially when it is known that this can reversethe reaction? If signs and symptoms occur only at a late stage (e.g., in agranulocytosis), what are the prospects of detecting the reaction earlier by laboratorytesting? If early detection is possible and it can improve the outcomeprognosis, is it practical to maintain surveillance for the reaction? Can high-risk individuals be identified for more intense surveillance? What are theconsequences of not detecting the reaction early? Guidelines for Preparing Core Clinical-Safety Information on Drugs, Report of CIOMS Working GroupIII, Council for International Organizations of Medical Sciences, Geneva, 1995.
Prior to the approval of remoxipride in Sweden and the United Kingdom, clinical trials had demonstrated that it had a more favourabletolerability profile than other anti-psychosis agents. About two years afterthe first approval it was noted that seven cases of aplastic anaemia (and oneof severe thrombocytopenia — possibly early aplasia) had been reported.
An estimated 50,000 patients had been treated with the drug, for an overallincidence of about one case in 6,000 (5 in 10,500 patients in the UnitedKingdom and 2 in 16,500 in Sweden).
Spontaneous reports indicated that the risk of fatal and other serious adverse reactions to remoxipride was probably the same as that for otherneuroleptics. However, unlike the reactions associated with some alter-native drugs (e.g., agranulocytosis), it is uncertain whether early detectionof aplastic anaemia decreases the risk of a fatal outcome. Generally, by thetime abnormalities in peripheral blood are detectable it is too late.
The concept of risk driver and its effect on risk profiles The adverse reaction that dominates the overall risk profile (carries the most weight) is referred to as the risk driver or dominant risk. The existenceand documentation of the risk driver may predate the new prompt (signal).
For this and other reasons, therefore, in addition to the data on the reactionthat prompted the risk evaluation, all other data on adverse reactions to theproduct should be taken into account in describing and quantifying theoverall risk; this applies especially to other serious or frequent reactions. Toobtain as complete a picture as possible, complete adverse-reaction profilesattributed to each therapeutic alternative should be compiled. Standardgraphic displays for this purpose are suggested below. The circumstances andthe perspective taken can determine whether the relative-risk evaluationmight be restricted to a defined set of reactions, such as those that can be fatalor lead to hospitalization. If a risk driver can be identified, the risk evaluationmay need to cover only one, two or three reactions. For example, an NSAIDand dipyrone could be compared for the risks of upper-gastrointestinalbleeding, associated particularly with NSAIDs, and of agranulocytosis, withdipyrone, and the comparison could be weighted by statistics on mortality ordays of hospitalization within a defined observational period.
In general, however, it is suggested that as a first step the three most often reported and the three most serious adverse reactions be chosen asrepresentatives for the risk profile of each medicine in the comparison.
Risk evaluation across products Typically, risk is evaluated for a single medicinal product, without reference to its therapeutic alternatives. Comparative risk evaluation, which is crucial and more relevant, is more difficult for many reasons. One is thatknowledge of drug risks increases with exposure and duration of use;because therapeutic alternatives enter and penetrate the market at differenttimes the amount of information available about them may differ. Also,much of the information on serious adverse reactions may originate fromspontaneous reports, but the necessary information to determine withsufficient precision the frequency of drug-attributed risk is not readilyavailable (for the company conducting the analysis, this may be trueespecially of information on comparator agents). Another reason is thatdifferent types of adverse reaction are not easily comparable when theirclinical significance differs (e.g., duration and consequences). In addition,adverse reactions to medicines are only directly comparable if they can bedescribed with the same measure of severity.
Once the comparators are chosen, the three key elements of risk assessment — qualitative description of the adverse reactions, theirfrequency of occurrence, and a weighting in a common unit of the relativeimportance of the reactions — should optimally provide numericalexpressions for comparison.
3. Risk Profiles for Individual Drugs A basic principle As a simple and straightforward approach to the description and quantification of risks, the Working Group suggests the use of standardvisual presentations of data on adverse reactions in the form of ‘‘profiles,''usually as bar charts. Such profiles can be very useful for, as a preliminary step,illustrating patterns of data, from different sources, about the same or differentmedicines. The basis for this approach is found in previously published work2, 3on prescription-related profiles of adverse reactions for benefit-risk analysis.
The comparison of risk profiles presented side by side for different drugs can bequite useful. The key components of such a standard are: a common set of elements (categories of outcomes) a common structure a common presentation of well-defined data.
Speirs, C.J. Prescription-related Adverse Reaction Profiles and their Use in Risk-Benefit Analysis, inIatrogenic Diseases 3rd edition, P.F. D'Arcy and J.P. Griffin, editors, Oxford University Press, 1986.
Wiholm, B.E., Myrhed, M., and Ekman, E. Trends and Patterns in Adverse Drug Reactions to Non-Steroidal Anti-inflammatory Drugs Reported in Sweden. In Ramsford, K.D. and Velo, G.P., eds., SideEffects of Anti-inflammatory Drugs, Part 1, MTP Press, Lancaster, 1986, pp. 55-62.
Figure 1 is an example of such a display; it presents actual data from spontaneous reports on drug X.
Figure 1. Adverse-drug-reaction (ADR) Reporting Profile of Drug X.
Number Of Spontaneous Reports Over A Specified Period
skin and app.
Number of reports
The adverse-reaction terms are sorted into body systems (system-organ classes), a widespread practice which has been proven quite useful and hasbeen included in other proposals of the CIOMS Working Group (e.g.,CIOMS III). In addition, specific reactions may be displayed in this way,grouped collectively, for example, as reactions that were serious or resultedin hospitalization or were fatal, or in other appropriate categories; profilesspecific to system-organ classes may also be informative. The source of thesystem-organ class categories used (e.g., WHOART, COSTART, or,presumably in the future, MedDRA) needs to be specified and must notbe changed, especially as system-organ classes do not necessarily comprisethe same sets of preferred terms in the different terminologies now in use.
Structure and presentation of the data It is common to sort according to estimated frequency of occurrence of adverse reactions (e.g., in order of highest to lowest, from top to bottom ofthe chart), and CIOMS III proposed this method in regard to core safetyinformation. However, the sequence is likely to be different for everyproduct. Although different sequences still permit a qualitative comparisonof drug-attributed adverse-reaction profiles, it is much better to use adefined sequence, in order to establish a standard method of comparingdifferent medicines. The sequence chosen is immaterial; it should be easy toremember and kept unchanged for the sake of familiarity and consistency,so that it is recognizable at a glance.
It is well recognized that in themselves absolute numbers of reports (as given in Figure 1) are inadequate gauges of the magnitude of reporting,especially when comparing different drugs. However, for a given drug, thedistribution of reports among the different system/organ classes, or forspecific adverse reactions within those classes, is very useful and can beportrayed by using percentages rather than numbers (N) of reports for eachclass; the numbers could still be shown on the bars if desired.
Some differentiating characteristic (e.g., colour or shading) within each bar is useful to indicate subcomponents (serious or fatal cases, for example).
For any form of presentation, such information as units, periods of time covered, and other key elements integral to interpretation of the data shouldbe provided with the diagrams.
Specification of data sources Profiles can be very useful for illustrating qualitative adverse-reaction reporting patterns for the same or different drugs. To avoid combining datainappropriately, separate profiles should be established for data from different sources and of different quality, such as data from spontaneousreports, clinical trials or epidemiological studies. For example, sponta-neous-report data can only reflect reporting frequencies, which areinfluenced by many factors affecting the numerator (number of reports)and the denominator (number of drug users). Therefore, risk estimates fromspontaneous reports must not be regarded as true incidences or trueestimates of risk.4 In the absence of comprehensive information, other data sources might include official data sheets (e.g., in the European Union an updatedSummary of Product Characteristics, or in the United States a packageinsert) or published literature; when needed, data from regulatory registriesmay also be available (e.g., from the US FDA via the Freedom ofInformation mechanism or from the Medicines Control Agency's ADROITdata-base in the United Kingdom).
Different populations ( e.g., adult or paediatric), indications and dosage forms may also necessitate separate profiles. For convenient visual effect,side-by-side bar charts with negative and positive x-axes could be used tocompare data from different sources.
Comparisons of profiles for different drugs As previously discussed, only drugs with similar conditions of use and data of the same type and quality should be compared. The patterns shownin profiles are influenced by many other less obvious factors as well, whichmust be taken into account and discussed (e.g., proportion of new- asdistinct from repeat- prescription exposures; length of time on the market;special attention given to a drug, such as ‘‘black triangle'' status in theUnited Kingdom). When these and other influences are considered,4 the bar-chart profiles can still be useful in demonstrating different patterns ofreported observations, as in the example shown in Figure 2. A main objectiveof comparing profiles of different drugs is to determine which reactions foreach drug should be chosen for further comparison (dominant risks).
4. Risk Weighing for Individual Adverse Drug Reactions Ideally, for an impartial comparison of the risks associated with two or more drugs, there should be an agreed method of scoring (weighing) the Sachs, R. M. and Bortnichak, E. A. An Evaluation of Spontaneous Adverse Reaction Monitoring Systems,American Journal of Medicine, Supplement 5B, 81:49-55, 1986; Baum, S., Kweder, S. and Anello, C. TheSpontaneous Reporting System in the United States, in Strom, B. L., ed., Pharmacoepidemiology, 2ndedition, John Wiley and Sons, 1994, pp. 125-137; and WihoIm, B., Olsson, S., Moore, N. and Wood, S.
Spontaneous Reporting Systems Outside the United States, ibid., pp. 139-155.) drugs in Sweden (% of total repor
y
se reactions repor
ofiles of ad
Figure 2.
Anti-inflammatory erse Drug Reactions to Non-Steroidal iholm, B.E., Myrhed, M., and Ekman, E.T Lancaster: MTP Press, pp. 55-62, 1986. relative seriousness of different reactions. Whereas there are standardregulatory/administrative definitions for classifying a case as ‘‘serious''(death,hospitalization, etc.), little has been done on the question of howbest to summarize the risks. For example, if a drug caused aplasticanaemia in 1/10,000 patients and bronchial asthma in 1/1000, how shouldthese risks be viewed in comparison with those of a drug that causedanaphylaxis in 1/10,000, glaucoma in 1/10,000 and diarrhoea in 1/1000? Although prescribers would like to make these kinds of comparison each time they prescribe one of a class of drugs, there is no obvious metric ormethod for doing so. Prescribers need to rely on their perception of risk,which is usually based on personal clinical experience. Although there is anobvious need to facilitate risk comparisons, and the Working Group wassupportive of any new ideas, any method would have obvious limitations.
What methods are available? Some companies have lists of medical conditions to help them decide whether spontaneously reported cases aremedically serious. The lists are often stored in a computer so that suchserious cases are automatically flagged. Similarly, conditions that wouldalways trigger an evaluation can be listed; for example, the WHOCollaborating Centre for International Drug Monitoring (Uppsala,Sweden) uses a ‘‘critical terms'' list for recognizing potential signals.5 Inthe future, when all companies and regulatory authorities are likely to use asingle medical terminology for adverse reactions, it may be possible toassign to each preferred term a score indicating its medical seriousness,which would be useful in comparing the relative risks of drugs.
Thus, it is tempting to consider for adverse-event terms the development of a standard hierarchy or scale that would define levels of seriousness (e.g.,a ranking of such conditions as anaphylaxis, myocardial infarction,gastrointestinal bleeding, rash) that could be used as the basis (threshold)for initiating a new benefit-risk evaluation and for comparing risks ofalternative therapies.
In an informal test, 24 members of the Working Group scored independently 20 adverse-event terms, selected arbitrarily by one member,on a scale of 1 to 10 (10 being the most serious). The scale resulting fromsuch an exercise obviously depends on who is doing the weighing:perspectives of patients, physicians, and others may very well differ fromthose of the Working Group. Table 1 shows the 20 terms ranked by theirmean scores and the range of scores. Though the assessors varied in their WHO Adverse Reaction Terminology — Critical Term List, WHO Collaborating Centre forInternational Drug Monitoring, Uppsala, Sweden, January 1996.
opinions, adverse reactions could be broadly categorized according to themean score into: very serious (score 7.5+), including aplastic anaemia,congenital abnormality, anaphylaxis, and disorders of cardiac conduction;serious (5-7.5), including seizures, gastrointestinal haemorrhage, asthma,glaucoma, premature labour and depression; fairly serious (2.5-5),including confusion, hypotension, raised liver function tests, waterretention, swelling of the eyelids, urticaria, maculopapular rash, andmuscle stiffness; and least serious, including diarrhoea and headache.
Table II shows the assessors' consistency — how reproducible the results were when 14 members of the Group repeated the scoring exercise withoutaccess to their original assessments. On average, each re-evaluated scorediffered by one point in either direction.
Table I: Adverse drug reaction seriousness scores: Rank order as assessed by 24 Working Group members Congenital abnormality Cardiac conduction disorder Raised liver function tests Swelling of eyelids Maculopapular rash Overall average range Table II: Adverse drug reaction seriousness scores: Re-scores from 14 Working (either direction) Congenital abnormality Cardiac conduction disorder Raised liver function tests Swelling of eyelids Maculopapular rash Overall average range * Based on 24 scores (10 scoring once and second score of the 14 scoring twice).
A similar survey conducted among a group of 221 general practitioners in the north of Yorkshire, England, ranked 63 adverse-reaction-terms in aranking similar to that of the CIOMS Working Group.6 Unlike theWorking Group survey, however, a statistical method was used to produce aweighted index value, based on the amount of variability among thephysicians. They did not repeat the exercise to test for consistency withinobservers.
Kind, P. and Dolan, P. Determining the Severity of Adverse Drug Reactions: The Views of FamilyDoctors. A survey commissioned by Hoechst Marion Roussel. See Proceedings of the Annual Meeting ofthe International Society of Technology Asssessment in Health Care (ISTAHC), Barcelona, May 1997.
Agreed scales or values for adverse reactions would contribute to the standardization of risk quantification; the issue constitutes an importantarea of research. Such a system would be enhanced if duration and/orincidence of a reaction, as well as its severity, could be factored into acomposite metric.
A member7 of the Working Group has proposed a very interesting model for quantifying risks along these lines; it is presented in Appendix C tostimulate thinking and further work. It incorporates quantitative criteria forseriousness and incidence.
5. Quantification of Risk Incidence of the reaction To put the newly identified risk into perspective, it is important to quantify it in terms of incidence. Precise quantification will usually bedifficult in the post-marketing environment, in which most new safetysignals arise from spontaneous reporting, with its associated uncertainties asto numerators (reported cases) and denominators (patient exposures).
However, risk can often be approximated in terms of magnitudes of 10, assuggested in the CIOMS III report: 51% (common or frequent); 51 per1000 but <1 per cent (uncommon or infrequent); 51 per 10,000 but lessthan 1 per 1000 (rare); <1 per 10,000 (very rare).
When possible, attempts should be made to determine whether the incidence is affected by the existence of any apparent high-risk groups.
These might be defined by, for example, dose or duration of treatment, useof other drugs (e.g., drug interactions), presence of other diseases (e.g., renalfailure), or special populations defined by demographics or ethnicity. Inprinciple, one of the most important functions of risk evaluation is toidentify individual patients at increased risk of serious adverse reactions.
Although some mechanisms are fairly well understood (enzyme inhibitionprocesses; drug interactions), the pharmacological and biological basis ofdrug-induced diseases (e.g., role of pharmacogenetics) is relativelyunexplored. Work in this area is needed and should be encouraged.
So far, we have been referring to the ‘‘reported'' (usually crude) incidence of a given adverse reaction. It is more meaningful, though, to speak of‘‘excess incidence'', namely the difference in incidence between patientsexposed and not exposed to the treatment over a defined period of Christian Be´nichou, MD, Synthe´labo (Paris, France) observation. This rate is also referred to as ‘‘drug-attributable incidence''.
However, ‘‘attributable'' need not signify that all the cases are causallyrelated to treatment — its meaning in this context should not be confusedwith its common meaning when used to ascribe causality (as in‘‘attributable to the drug''). The incidence in unexposed patients notreceiving alternative treatment (i.e., untreated) would represent thebackground (natural) incidence of the event in the diseased population.
Calculation of excess incidence from comparative clinical trials or cohortobservational studies is straightforward; in case/control studies, however,it is somewhat more complicated, as explained in standard references8.
The next level of comparison refers to incidence relative to alternative therapies. As mentioned above, such comparative assessments of risk are moredifficult, for many reasons. For example, knowledge of drug risks increaseswith exposure and time of use; a new drug may be associated with a relativelylower therapeutic response as well as a higher rate of adverse-reactionreporting. Thus, different timing and period of marketing, as well as extent ofmarket penetration, influence the type and amount of information availableabout alternative therapies. Ideally, data for determination of comparative (aswell as excess) incidence will come from clinical trials, comparativeobservational cohort studies, and population-based case-control studies. Inthe absence of such relatively high-quality data, crude estimates may begenerated from other data, usually spontaneous-report statistics for differentdrugs during comparable post-marketing periods. It must be emphasized yetagain, however, that because of the many confounders and potential biasesassociated with comparative spontaneous-reporting statistics4, estimatesacross drugs can be misleading and great caution must be exercised in theiruse. It is the judgment of the CIOMS Working Group that only underexceptional and rare circumstances should the use of such comparisons formthe sole basis of restrictive regulatory action.
It is also useful to keep in mind the obvious but often under-appreciated distinction between absolute and relative risk. For example, an event mayoccur with an excess incidence of 50%, or perhaps at an incidence 50%greater than that of a comparator. However, the meaning of such anapparently high risk can vary considerably as regards public health impactaccording to whether the actual incidence of an event is, for example, 1/10,1/1000 or 1/100,000, producing an absolute excess per million patientstreated of 100,000, 1000 or 10 patients, respectively.
For example, Kelsey, J.L., Thompson, W.D. and Evans, A.S. Methods in Observational Epidemiology,Oxford University Press, 2nd edition, 1996, or Armitage, P. and Berry, G. Statistical Methods inMedical Research, Blackwell Scientific Publications, London, 2nd edition, 1987.
Two of the illustrative cases provide simple examples of incidence considerations; see Appendix B for details.
Felbamate, a drug uniquely suited to the treatment of Lennox-Gastaut syndrome, is associated with a relatively high incidence of aplastic anaemia,about 1 in 4000 patients treated (rare), with a case-fatality rate of 28% (fromspontaneous reports). The evidence for causality is based on the incidenceand nature of the spontaneously reported reactions. Thirty-six cases ofaplastic anaemia had been reported at the time of a benefit-risk evaluation(estimated overall exposure was 120,000 patients). All had pancytopeniawith severe thrombocytopenia, and most had severe bleeding and infection.
Bone-marrow biopsies were consistent with aplastic anaemia. Felbamatewas the only drug used in 11 cases. There were two positive rechallenges.
Median time to onset was five months. There were no clear risk factors. Inaddition, there were 16 cases of hepatitis, of which nine were serious andfatal. Thus, total drug-induced mortality was of the order of 1/6000 patientstreated. The only alternative therapy, surgery, was considered to havecomparable morbidity and mortality. In contrast, the overall case fatality ofuntreated Lennox-Gastaut is even higher — about one in 500 cases.
Within 120 days after the introduction of temofloxacin (a fluoroquino- lone) to the market in the United States, the company and the Food and DrugAdministration became aware of several cases of serious adverse experiences.
These included such life-threatening reactions as haemolytic anaemia,thrombocytopenia, renal impairment and hypoglycaemia. Hypoglycaemiaoccurred predominantly in elderly patients, but none of the other reactionswas associated with a specific high-risk group. A comparison of spontaneousreports for three other fluoroquinolones did not reveal a similar problem.
With estimates of patients exposed (IMS data) for denominators, thereporting rate for temofloxacin was estimated as 108 per 100,000 prescrip-tions, compared with 13, 25 and 20 per 100,000 for the other fluoroquino-lones, and the product was subsequently taken off the market. This examplehighlights the observation that the usefulness of spontaneous reportsincreases with the severity and rarity of the adverse reaction, and when thereare large apparent differences in risk between drugs.
Practical approaches to weighting adverse reactions Only by attaching a specific weight to each adverse reaction is it possible to construct a common unit that makes it possible to qualitatively comparedifferent reactions. As already indicated, the ideal common unit would takeinto account the duration, severity and consequences (outcome) of thedifferent reactions. Depending on the circumstances, however, and in the absence of a composite standard, such markers of severity or consequenceas death, life-years lost, days of hospitalization, and quality-of-life scoresare examples of units that can be used when the requisite data are available.
If mortality is chosen, the case fatality rate is needed. For days ofhospitalization, data from diagnosis-related hospital statistics or surveysmay be needed. Quality-of-life changes, days of hospitalization, or case-fatality rates are often country-specific, and this must be considered in theanalysis and interpretation of the data.
Estimated total drug risk Once one or more standard units are selected as markers for comparison (such as case-fatality rate or days of hospitalization), the weighted excessincidence rate can be calculated for each reaction by multiplying the excessincidence rate by the reaction-specific weight factor.
To obtain the weighted overall excess incidence of the combined adverse reactions associated with a particular drug, the sum of the individual valuesprovides an aggregate (e.g., for the three most serious reactions, if this is themodel chosen for comparative analysis). The chosen comparators can thusbe compared on a common basis.
Finally, to obtain the incremental drug-attributable risk, the aggregate values for each drug must be subtracted from those of an arbitrary referencestandard drug. The net risk corresponds to the magnitude of the incrementaldrug risk. For a practical example, see Appendix D.
Validity of the estimates To test the robustness of the risk estimation, sensitivity analysis can be performed by including the precision of the risk or weighting measure. Forexample, if excess incidence estimates vary across studies, the lowest andhighest estimates could be included in the sensitivity analysis. Thisprocedure will provide information on the strength of the overallcomparative assessment — i.e., the extent to which the overall results varyby the range of risk estimates or other assumptions.
Great care must also be taken in any attempt to extrapolate the results beyond the specific populations (including geographic locations) andtreatment alternatives under consideration. If total risk estimates areextrapolated beyond the specific population from which the resultsoriginated, then all assumptions must be stated.
6. Suggested Sequential Overall Approach On the basis of the concepts and processes described above for arriving at a fair comparison of the risks associated with competing alternativetherapies, the Working Group suggests the following general methodolo-gical approach for the evaluation of drug-attributed risks. It is designed tominimize the potential for bias in the evaluation of risk and to maketransparent the methods used.
(a) Define the perspective of the risk evaluation (e.g., consumer, health authority, pharmaceutical firm, public health). In the present context,this is usually a public-health perspective.
(b) Review the relevant indication and use in the target population for the subject drug (to include recommended dose, duration, age group, etc.).
(c) Specify comparator drugs (i.e., drugs that could replace the drug of interest) and alternative therapies or the no-treatment option. Provideall pertinent information as in (b).
(d) Determine and display the profile of adverse reactions for the subject drug and the specified alternatives. Risk evaluation should not berestricted to one reaction; initially, profiles should represent allreactions, but detailed analysis may be limited to a subset, especiallyif a dominant risk (risk driver) can be identified.
(e) Obtain the background incidence of each reaction, usually from the literature, and for each drug select the reactions with the greatestcontribution to the overall risk (dominant risks). These can form part ofthe subset for detailed analysis (e.g., the three most serious and the threemost common reactions).
(f) For the specific reactions define a common outcome measure and determine values.
(g) Estimate the excess incidence of each reaction for each therapy; drug- attributable incidence rates are the relevant risk measure.
(h) To permit comparison across reactions, adjust the excess incidence of each reaction with the weighting factor to provide the estimatedweighted excess incidence rate of each reaction.
(i) Sum the weighted excess-incidence rates across reactions for each therapy to obtain the total drug-attributed risk for each treatment.
(j) Calculate the net difference between drug-attributed risks to provide a measure of incremental drug-risk relative to one of the alternativeschosen as an arbitrary reference.
(k) Ascertain the validity of the results by performing a sensitivity analysis, especially with the risk driver. Consider extrapolation of the resultsfrom the data used in the analysis to the total target population.
(l) Update the evaluation when relevant new information becomes available — from published studies, for example. Risk evaluation is acontinuing process and additional data or new reactions may come tolight as the number of patients using a drug increases.
The dipyrone case illustrates many of the concepts and steps discussed above (see Appendices B and D for details). In summary, the drug has beenassociated with agranulocytosis since the first case was reported in 1935.
Evidence of causality was based on the incidence and nature of the reportedevents as well as on the observation of cross-sensitivity with aminopyrine,another drug known to be associated with agranulocytosis. The Interna-tional Agranulocytosis and Aplastic Anaemia Study (IAAS), a large, multi-country, population-based case-control study, reported an overall annualincidence of community-acquired agranulocytosis of 3.4 per million, with afatality rate of 10%. The excess incidence of mortality for dipyrone was 0.10per million short-term users a week. The combined excess incidence ofmortality for agranulocytosis, aplastic anaemia, anaphylaxis and seriousupper-gastrointestinal complications was 0.11 per million a week fordipyrone, with gastrointestinal bleeding accounting for 69% (i.e., thedominant risk). For NSAIDs, however, the mortality per million users aweek was 1.66 for aspirin, 1.50 for diclofenac and 0.43 for ibuprofen; forthese drugs, serious gastrointestinal bleeding contributed 99% of the excessmortality. Agranulocytosis, the reaction that triggered the benefit-riskevaluation in this case, is associated with significant morbidity andmortality, though less so than before modern antibiotics became available.
However, when mortality due to other types of adverse drug reaction istaken into account, the risk associated with many alternatives to dipyrone isconsidered to be as high or higher.
D. Benefit-Risk Evaluation Evaluation of benefits (Chapter IIB) and risks (Chapter IIC) for a single therapy is only apparently a simple matter; when it refers to varioustherapies it is particularly difficult. In general, the net benefit-risk profile forsociety must reflect not only how many patients are expected to benefit froma therapy and to what extent, but also the uncertainties associated with its benefits and risks. Ideally, the relationship between benefits and risks can bedemonstrated by a simple graph (Figure 1). It indicates that at the extremes(high risk, low benefit; low risk, high benefit) judgments are relatively easybut that between the extremes there is more uncertainty, and decision-making can be more complex and difficult.
Rawlins1 has classified benefit-risk assessments as formal, comparative and informal (judgmental) analyses. Formal analysis is a scientifically-deductive process, requiring quantitative comparisons of benefits and risks;the comparisons provide a numerical expression of the trade-off. Aninformal analysis is an inductive or subjective process, calling for personaljudgment in assessing the relevant data on therapeutic options. Formalanalyses, especially in recent times, have incorporated quantitative modelsbased on studies of cost-effectiveness, cost-benefit and cost-utility, as well asthe tools of meta-analysis and of decision-analysis theory.
For many benefit-risk assessments, however, neither the benefits nor the risks are easily or appropriately compared quantitatively (e.g., a drug forsymptomatic relief of a common, self-limited disease but with anassociated risk of a rare, life-threatening adverse reaction). In such cases,the final assessment will be largely qualitative, even somewhat subjective;the level of risk tolerance is directly related to the perceived degree ofclinical benefit.
The comparative method compares qualitatively the product in question with similar products to determine whether the benefits and risks appearsimilar. However, the data and standards for benefit-risk assessment on the‘‘similar,'' usually older, drugs may not be satisfactory by today's standards.
In informal analysis, judgment can be influenced by perceptions and standards of medical practice, including different moral attitudes and otherintangible considerations. Probably most benefit-risk decisions have beenmade on a relatively informal basis, which has led Rawlins1 to state that‘‘invariably risk-benefit assessment is based on the fallibility of humanjudgement.'' The Working Group believes, however, that the better approaches available today should be pursued; this chapter discusses various attemptsdescribed in the literature and some new techniques. Although it may notyet be possible to truly quantify a benefit-risk relationship, certain tools canstrengthen the analysis and reduce reliance on judgment alone. Details ofsome approaches are presented in Appendix D.
Rawlins, M.D. Risk-Benefit Decisions in Licensing Changes, in Medicines and Risk/Benefit Decisions,Editors S.R. Walker and A.W. Asscher, MTP Press Limited, Lancaster, 1987, pp.137-141.
Figure 1: The Benefit-Risk Spectrum
the worst balance for restricted use only ay require further study acceptable balance excellent balance * Courtesy of W.Amery, personal correspondance A serious obstacle to the use of a composite metric for a quantitative analysis is that benefits and risks are usually measured and expressed bydifferent units and parameters. Besides, benefits reflect the degree to which aproduct decreases the risks associated with a disease; thus, it evinces a ‘‘risk-risk'' dichotomy. Various attempts have been made to derive a summarystatistic, mostly in the context of clinical trials. A simple example is theglobal assessment given by a treating physician to reflect a patient's overallresponse to treatment; this is commonly used in trials in Japan (referred toas yuyosei).2 In their evaluation of mass chest X-ray programmes, Payne and Loken3 defined a benefit-risk ratio as the number of efficacy events needed for eachoccurrence of a serious adverse event. From 40 million X-rays, 39,250 casesof curable tuberculosis, lung cancer, leukaemia and other diseases weredetected but also 723 somatic or genetic deaths, or a benefit-risk ratio ofabout 50:1. Chang and Fineberg4 investigated treatment strategies in the use Sakuma, A. Subjectivity in Clinical Medicine, Abstracts, Drug Information Association Second AnnualBiostatistics Meeting, Tokyo, Japan, August 30, 1995.
Payne, J.T. and Loken, M.K. A Survey of the Benefits and Risks in the Practice of Radiology, CRCCritical Reviews in Clinical Radiology and Nuclear Medicine, 6: 425-439, 1975.
Chang, R.W. and Fineberg, H.V. Risk-Benefit Considerations in the Management of Polymyalgia,Medical Decision Making, 3: 459-475, 1983.
of potentially toxic doses of steroids to control blindness in patientswithpolymyalgia rheumatica syndrome. They constructed an ‘‘incrementalbenefit-risk ratio'', which represented the number of additional cases ofsevere toxicity for each case of blindness averted in changing from onestrategy to another. In another approach, from available data on dose-benefit (dose-response) and dose-injury relationships, Andrews5 used fittedcurves to search for the optimum radiotherapeutic dose for patients withcancer.
In addition to the conventional measures of efficacy (morbidity and mortality under ‘‘ideal'' conditions) and effectiveness (under more ordinaryconditions), other potential influences of medicines may be considered: qualityand quantity of life (e.g., quality-adjusted life-years, or QALY) and use ofhealth-care resources (cost efficiency). As mentioned above, these are examplesof metrics that can lead to more formal quantitative approaches, particularlywhen used in conjunction with estimations of utility values for states of health.
There have been several instances of the application of multi-attribute utility-function approaches in weighing benefits and risks.6, 7 However, as theWorking Group has stressed elsewhere (Chapters 1 and 7), althoughmeasurements of quality and quantity of life and economic evaluation arebeing used increasingly in the licensing and use of drugs8, 9, there is noagreement on what constitutes proper methods in benefit-risk assessment.
Therefore, until techniques for such formal quantitative approaches are tested,standardized and validated for a number of diseases and populations, theirapplication is limited, particularly in the present context.
However, such difficulties should not be deterrents to the use of various quantitative or semi-quantitative techniques. Some general approaches aredescribed here to stimulate more thinking and needed research in thisnascent area. Like most quantitative models, their sensitivity, reproduci-bility and applicability depend heavily on the ‘‘weights'' or ‘‘scores'' appliedto the various measures of benefit and risk; such models must evolve and be Andrews, J.R. Benefit, Risk and Optimization by ROC Analysis in Cancer Radiotherapy, InternationalJournal of Radiation Oncology, Biology and Physics, 11: 1557-1562, 1985.
Eriksen, S. and Keller, L.R. A multiattribute-utility-function Approach to Weighing the Risks andBenefits of Pharmaceutical Agents, Medical Decision Making, 13:, 118-125, 1993.
Schumaker, G.E. Multiattribute Evaluation in Formulary Decision Making as Applied to Calcium-channel Blockers, American Journal of Hospital Pharmacists, 48: 301-306, 1991.
Benefit, Risk and Cost Management of Drugs. Report of the CPHA National Advisory Panel on Risk/Benefit Management of Drugs. Canadian Public Health Association, Janaury 1993.
Tools Employed in the Measurement of the Risks and Benefits of Drugs — A Literature Review for theCanadian Public Health Association's National Panel on Risk/Benefit Management of Drugs. CurryAdams and Associates, Ottawa, Ontario, Canada, July 1991.
tested by collaboration of experts from health authorities, manufacturersand others.
Irrespective of method of analysis, some basic principles and approaches must be considered for a sound benefit-risk evaluation.
All available relevant data on benefits and risks should be assembled andconsidered for all the therapies compared.
Whenever possible, use the same, or at least comparable, parameters andoutcome measures for all the therapies.
Indicate clearly all assumptions and rules adopted regarding inclusion orexclusion of data and their use, and include discussion on any valuejudgments in weighing the merits of different sources and types of data.
Any quantitative or qualitative model should be regarded as providing ameasure of the ‘‘intrinsic'' property of the medicine, reflecting for a givenmedicine the benefit-risk relationship at the population level.
It is useful to remember, however, that the results as applied in practice must be tempered by ‘‘extrinsic'' factors, or patient characteristics, whichmay not be fully understood but can influence the therapeutic value of amedicine to an individual, such as10: Are high responders at more or less risk than patients achieving lowerdegrees of benefit? Is the patient resistant to alternative therapies and, therefore, ‘‘must''have the subject drug? Are there factors that put a particular patient at higher or lower risk ofan adverse reaction and that are not necessarily described in prescribinginformation? How sensitive is the degree of risk to the needed duration of therapy (e.g.,a few or many weeks)? 2. Descriptive and Semi-Quantitative Methods of Analysis Benefits and risks can be described in relation to the intensity (seriousness or severity) of the treated disease or adverse reaction, itsduration or chronicity, and, especially in the case of a reaction, its incidencein the treated population.
10 Dr. W. Amery, private communication.
Thus, adverse reactions are characterized by their: Benefits may be evaluated and described for a target disease in the light chronicity (e.g., acute, chronic, or duration of disease) extent of control or cure.
Because, presumably, all treated patients have the disease in question (thus, ‘‘incidence'' is 100%), the incidence metric for adverse reactions doesnot apply directly to benefits in this context, except for the use of a drug inprophylaxis or vaccination (what is the incidence of disease with andwithout treatment?).
On another level, however, it may be appropriate in context to describe a disease in terms of its incidence in the population as a whole (e.g.,numerically as rate of occurrence per 100,000 people per unit of time). In asimplified approach to weighing of benefits and risks, various metrics anddisplays summarizing comparative-benefit and comparative-risk data alongthe above lines can be examined informally and judgments made regardingthe relative merits of alternative therapies. The challenge is to categorize theinformation in a way that reduces the statistics to summary proxies.
One way to reduce the complexity of an overall analysis is to express for each medicine its benefits and risks in terms of the above descriptors, and tofocus on the selected adverse reactions, including the signal-generatingevent. If needed, benefits may also be expressed separately for each of adrug's indications. A comparison with background (no treatment) is alsopossible.
A number of drugs may be associated with the same or multiple different adverse reactions of different severity and frequency. Moreover, the signalevent leading to a new benefit-risk evaluation may occur when the datasheet already contains a more ‘‘serious'' adverse reaction recognized as thedominant risk (‘‘risk -driver'') in the overall risk profile. Also, it is likely thatthere is better evidence for causality and actual incidence of the mostcommon and serious reactions.
Therefore, to make a fair comparison of risks between therapies, the ‘‘principle of threes'' can accommodate most situations: choose the three most serious and three most frequent adverse reactions for each drug in thecomparison.
For purposes of verbally describing benefit and risk properties, the following graduated terms are examples that could be used to describe eitherthe disease (seriousness, duration) under treatment or an adverse reaction(seriousness, duration, incidence): To aid in reconciling the relationship between the risk and the treated disease, a tabulation of the grades is informative. A hypothetical example isshown for chronic bronchitis treated with an antibiotic that causes amedically unimportant but fairly common reaction; the disease, itsalteration by the antibiotic, and the nature of the adverse reaction aredepicted separately: Nature of the disease (chronic bronchitis) Effects of the drug on the disease * Medicines used for prophylaxis will have a score in this row. Antibiotics have an important effect on acute attacks, but less effect on the chronic course.
Adverse drug effects Especially when comparing drugs, this descriptive approach cannot take into account the possibly varying types and quality of data used as the basisfor filling the cells of the cross-tabulation. Obviously, judgement is neededin the categorization scheme, but it is suggested that the conceptualframework may be used as a simple, albeit crude, algorithm to maintain aconsistent logic in benefit-risk evaluation. It can also contribute to moretransparency in decision-making. This approach has been presentedpreviously.11 To illustrate its applicability, the classification grid has beenapplied retroactively to some of the illustrative case examples (Appendix B);a numerical scale is used for the qualitative terms (low=1, medium=2,high=3; no effect/no influence=0).
Aplastic anaemia is the dominant risk for felbamate, an anti-epilepticdrug, which is the only effective treatment for Lennox-Gastaut epilepsy,a disease with a high mortality: 1 in 500 (seriousness score at least 2,probably3).Itisalsointractable(duration3)andrare(400casesayearinFrance; incidence 1). The drug reduces dramatically the morbidity andmortality of Lennox-Gastaut epilepsy (disease seriousness reduction/effectivenessscore 3,andduration3).Felbamatehasnoinfluenceontheincidence of the disease (score 0), since it suppresses rather than preventsor cures it. More value judgement has been used in scoring theeffectiveness of the drug than in scoring the disease parameters becauseof the limited information given in the case description.
Blood dyscrasia with aplastic anaemia ‘‘drives'' the risk profile(seriousness score 3, as aplasia is often fatal); the condition isgenerally persistent (score 2) and the incidence is estimated at 1 in4000 treated patients (score 2). Some might give incidence a scoreof 3, since the ‘‘acceptable'' incidence for serious adverse reactionsis often perceived as being very small compared with diseaseincidence. The results are easily displayed on a grid: 11 Edwards, I.R., Wiholm, B.E., and Martinez, C. Concepts in Risk-Benefit Assessment, Drug Safety, 15 (1): 1-7, 1996.
3=high, 2=medium, 1=low, 0=no effect/influence This shows simply and clearly how the assessor has judged the effect offelbamate in a serious disease, and that the dominant adverse-reactionrisk profile is marginally worse than the disease, but only in incidence.
In this particular case, for a disease with no other effective treatment,the absolute comparison of benefits and risks led to the conclusionthat such an effective drug as felbamate could be used for this specialpopulation but was not suitable for the general treatment of epilepsy.
Obviously, a certain amount of value judgement is needed to produce the grid scores; it is expected that as more quantitativeinformation becomes available there will be less need of value-weighing. Nevertheless, with such an approach it should be clear thatsmall changes in scores limited to a range of 0 to 3 may give quitedifferent results, and the sensitivity to such changes in values must betaken into account. A broader range of scores to allow for finerdifferentiation can be an advantage (see TURBO model, below).
Forthismildanalgesic,agranulocytosis is considered thedominantrisk.
As the dipyrone grid shows, with this simplified approach the scores 3=high, 2=medium, 1=low, 0=no effect do not allow an obvious judgment to be made regarding the balancebetween effects of the disease and the particular adverse drug reaction.
However, it is useful to examine a typical NSAID, say aspirin, todemonstrate the importance of comparisons before reachingconclusions or making decisions. Aspirin is also a mild analgesiccomparable in efficacy to dipyrone but with a very differentdominant risk: gastrointestinal bleeding.
3=high, 2=medium, 1=low, 0=no effect The higher incidence of gastrointestinal bleeding associated withaspirin, and degree of seriousness and duration of illness similar tothose of agranulocytosis, possibly give dipyrone a better riskprofile than aspirin! Formerly used to treat leg cramps, quinine is associated withthrombocytopenia(thr) and arrhythmia(arr) as risk-drivers. Thisexample demonstrates the presentation of data when there is morethan one risk factor. Note that a sensitivity analysis can be done byweighting to some extent the seriousness of either reaction.
3=high, 2=medium, 1=low, 0=no effect The lack of convincing efficacy data had shown that quinine wasan ineffective drug for this relatively trivial condition. Even withthe comparatively low scores for the adverse reactions for thisindication, the benefit-risk balance is unfavourable.
It is worth emphasizing again that, for each parameter in the grid, the quantitative detail and the qualitative judgements should be describedclearly. Instead of the simple grading used here, more refined numericalscores could be used, and even weighted. However, the scheme andmethodology must be thoroughly documented and explained. A focus onthe three most serious and three most frequent reactions for the subject drugand comparators, as already suggested, can be accommodated. Thedifferent medical options may be compared by inspecting the grids for thevarious adverse reactions (and benefits), or the scores may be added acrosseach drug and the totals compared to provide a crude relative ranking.
3. Quantitative Approaches When sufficient data on both benefits and risks are available, a more quantified version of the descriptive grid technique discussed in the previoussection could be useful. In accordance with the ‘‘principle of threes'' foradverse reactions, the following example illustrates the proposed method: 40% cure rate for acute-on-chronic bronchitis(a disabling disease) Adverse reactions: Skin rash (in 20% of treated patients; lasts aboutthree days after drug is stopped); stomach upset (in10%; lasts up to one day afterwards); diarrhoea (in5% of cases; may last up to three weeks afterwardsand is prostrating in 0.05% of cases); agranulocy-tosis (in 0.005%, with 10% fatality).
For ranking (scoring) the benefit and risk properties, a convenient (but arbitrary) gradation can be chosen, similar to that used in the grid exercise(in this case, low=10, medium=20, high=30). Scores are then calculated asfollows, with the judgmental values shown: cure rate x seriousness of the disease xchronicity/duration of the disease 0.4 x 30 (disabling disease) x 20 (acute-on-chronic) Common adverse reactions (Mean score 13.3) 0.2 x 10 (‘‘low'' seriousness) x 10 (‘‘low'' duration) = 20 Stomach-upset score 0.1 x 10 x 10 =10 0.05 x 10 x 20=10 Rare adverse reactions (Mean score 0.078) Prostrating diarrhoea score 0.0005 x 20 x 20=0.2 Agranulocytosis score 0.00005 x 30 x 20=0.03 Fatal agranulocytosis score 0.000005 x 30 x 30=0.0045 An overall ‘‘benefit-risk ratio'' (B/R) can be calculated by using the meanof all adverse-reaction scores [(13.3 + 0.078) /2 = 6.69]:B/R = 240/6.69 = 35.9 Alternatively, especially to compare two or more drugs, separate B/R values for common and rare reactions could be calculated (respectively, 240/13.3 or 18, and 240/0.078 or 3077); individual reactions may also becompared in this way between drugs. Although it is contrary to expectationthat the B/R for rare serious reactions is the greater value in the examplegiven (a consequence of the scales chosen and the arithmetic), comparisonbetween drugs is the key process.
Obviously, this technique is highly dependent on the scales and weightings used and it requires testing. Further development could incorporate suchbenefit metrics as mortality rates (measure of lives saved or deathspostponed), years of life gained, and quality-adjusted-life years (QALYs)gained; risk metrics, such as drug-attributed loss of quality-adjusted-life-years (DALQALYs), may also be applied.
For a drug-specific actual example incorporating global excess mortality risk as the principal metric, see Appendix D, which describes the benefit-riskevaluation of dipyrone compared with that of four other drugs and fourdifferent medically serious adverse reactions.
A similar quantitative and graphical approach to comparative benefit- risk analysis has been proposed by Dr Willem Amery, which he calls the TURBO model (Transparent Uniform Risk/Benefit Overview).12 Scores areassigned to the risks (R-Score) and the benefits (B-Score) and are thencombined into an overall TURBO or ‘‘therapeutic'' score; the score can beregarded as a measure of the ‘‘intrinsic'' property of a drug, reflecting thebenefit-risk relationship at the population level for a given medication;different indications can, of course, be associated with different benefit-riskbalances, usually because of different associated benefits. The ultimate goalis to place each drug in the benefit-risk spectrum (Figure 1) by use of thederived scores. See Appendix E for details.
4. General recommendations It is recognized in regard to benefit-risk evaluation that there is no structured and harmonized approach that could serve as a standard.
Techniques suggested here and elsewhere require extensive developmentand testing. Particular attention should be paid to the approaches outlinedin Appendices D and E. However, there are some general lessons to beshared in the form of ‘‘dos and don'ts'' when attempting to put the results ofa benefit-risk evaluation into proper perspective.
(a) Avoid the expression ‘‘benefit-risk ratio''. Until adequate quantitative approaches are validated, the term conveys a misleading mathematicalcredibility and has little meaning relative to impact on the publichealth.
(b) Avoid the use of relative expressions of benefits and risks in isolation; they do not reflect the true medical impact, which is better expressed inabsolute values. For example,a 33% relative risk reduction may meana decrease from an incidence of 30% to 20%, or 3% to 2%, or 0.3%to 0.2%.
(c) Be wary of uncritical use of overall expression of risk or benefit. Benefit and risk are rarely evenly distributed over time, the population treated,or the use of a drug (e.g., indication- or dose-dependency).
(d) Always frame the issue and the results in the proper therapeutic context, as in the following examples: treatment of a fatal disease with drugs that can cause fatal adversereactions ( e.g., treatment of acute leukaemia).
12 The CIOMS Working Group is grateful to Dr Amery for sharing his ideas, as yet unpublished. He welcomes comments and suggestions to aid in further development and testing of the concepts presentedhere. (Janssen Research Foundation, Turnhoutsweg 30, B-2340 Beerse, Belgium).
prevention of a serious, possibly fatal, condition vs. risk of a serious,possibly fatal, adverse reaction (aspirin or ticlopidine in strokeprevention vs. risk of gastrointestinal or cerebrovascular haemor-rhage or, in the case of ticlopidine, agranulocytosis) palliative treatment that improves quality of life vs. a serious,sometimes fatal reaction (disease-modifying anti-rheumatic drugs inrheumatoid arthritis) relief of symptoms in non-fatal, acute, self-limiting disease (aspirinfor ‘‘flu'', with risk of gastrointestinal bleeding; ‘‘cure worse than thedisease?'') prevention of risks to an individual or others, as in vaccination (e.g.,children vaccinated against rubella to protect pregnant women).
Finally, it is worth reiterating that the development of properly validated quantitative models for comparative benefit-risk evaluation is in its infancyand decisions are being made on a relatively informal basis. However, thesemi-quantitative and quantitative methods described can add a straight-forward descriptive and transparent approach to the assessment of therelative merits of different therapies.
E. Options Analysis To the extent possible, the risks and benefits of a medicinal product are assessed within the time-frame set by a company or regulatory authority.
After a thorough assessment, all the options available should be carefullyconsidered (options analysis) and decisions on the actions to be takenshould follow. The aim is to optimize the use of the product by maximizingits benefits and minimizing newly detected risks.
The options available to regulators and manufacturers are similar but not necessarily identical, and they are not mutually exclusive; they are: maintenance of the status quo (i.e., no change) ‘‘watching and waiting'' (i.e., monitoring subsequent experience) intensive additional data-gathering/new research (clinical or non-clinical) modifications to the product or its use or to the productinformation restriction of product availability suspension of product licence or investigational-status approval withdrawal of the product from the market (voluntary bymarketing authorization holder or mandatory by authorities) communication of new or reinforced information to the medicalprofession or the public Clearly, the options and decisions of a manufacturer and of a regulatory agency are interrelated and interdependent. In either case, wheneverpossible, the pros and cons of each option should be described in thecontext of the specific situation, including an impact analysis outlining theexpected consequences of adopting one or other option.
Each option is now discussed in detail.
1. Maintenance of the status quo If review of the evidence indicates no ground for concern — i.e., the suspicion of a problem raised by the original signal is unfounded — nochange is warranted in the dissemination and use of the product.
2. Watching and waiting If there is insufficient information for a definitive determination of whether there has been a change in the benefit — risk balance, the regulatormay choose not to require immediate corrective action but, rather, with themanufacturer to monitor carefully the continued use of the product. It mayalso be prudent for the manufacturer to gather additional data for periodicreview, analysis and submission.
3. Intensive additional data gathering/initiation of new research When a potential safety problem has been identified, the manufacturer will gather and assess as much data as possible to make the best possiblebenefit-risk evaluation. A decision may be needed on whether obtainingadditional data would help in the evaluation. The practicality of obtainingsuch data depends partly on how long it would take to gather the additionalinformation, and on the nature and magnitude of the safety concern(whether it represents a potential imminent hazard to the users, forexample). With these points in mind, the manufacturer should considerwhether additional studies (clinical and non-clinical), additional cases of a particular adverse event, comparative data on drugs of similar therapeuticuse or pharmacological class, or additional expert opinion would help indetermining the benefit-risk balance. In the temofloxacin and dipyroneexamples, additional comparative safety data were gathered to examine therelative risks of safety problems with the fluoroquinolone and analgesicdrug classes, respectively.
Circumstances will determine whether other, parallel, action may be needed pending the results of the new research.
4. Modifications to the product or its use or to the product information A regulatory agency and the manufacturer may introduce variations that could include changes to the prescriber or consumer information (datasheets, etc.), restriction of product use or supply, and formulation/manufacturing changes.
Changes to prescriber or consumer information The types of product information change which might be consideredinclude: the addition of new risk information to the sections coveringADRs, contraindications, warnings, precautions or interactions; changesin wording or emphasis to clarify or further specify adverse reactions;and restriction of indications, or in some cases, removal of information.
The felbamate case illustrates restriction of indications as a means ofmaintaining a product on the market while attempting to limit the risk ofimportant adverse reactions in a particularly susceptible population.
In certain cases, it may be appropriate to add a therapeutic recom-mendation for treating adverse reactions associated with the product.
The specific changes in product information for professionals or patientswill invariably be a joint decision of regulators and manufacturers. Aswith any changes, the new information should be placed and worded soas to maximize understanding by health professionals and any desiredchanges in prescribing habits.
Restriction of product use and supply Selective restriction of the availability of a product is sometimesconsidered an option to reduce the risk of an adverse reaction and tomonitor safety more efficiently. For example, the placement of a drug ona narcotics scheduling list is often considered in the case of products with significant potential for abuse. In the case of quinine the Food and DrugAdministration considered (but rejected) reclassifiying it from non-prescription to prescription status. Options available in some Europeancountries, as well as in the USA under certain circumstances, includerestriction of product distribution to hospitals or other specificinstitutions (institutional selectivity) and restriction of prescribing tospecialists in the field (professional selectivity). Such options wereconsidered in the case of felbamate in the European Union.
An option for investigational products that have a major safety problemand that have been licensed but not yet marketed is to limit distributionto patients enrolled in a ‘‘monitored release'' programme (Europe) or ina compassionate use/named patient programme (North America/Europe), or to patients who have signed an informed-consent document.
The latter has been used as well for a product already on the market (seethe felbamate and clozapine cases).
In some countries the cost of medications is fully reimbursed by anational health-insurance plan. Some regulatory agencies can imposelimits on reimbursement in order to restrict product use to a particularpatient population or to limit its duration.
Changes in formulation or in manufacturing It may be necessary or advisable to change the formulation or appearanceof a product or the manufacturing process so as to minimize or eliminatean identified risk (or a reduced benefit) related to the physico-chemicalproperties of a medicament. Examples are: a change in an excipient (or itselimination, in the case of a dye, for instance) shown to be responsible foran adverse reaction; a change in composition (e.g., lower strength of atablet); a change in a delivery system (e.g., from capsule to tablet to avertoesophageal insult); a change in particle size or crystalline form toovercome bioavailability or drug-delivery problems that influenceunfavourably the benefit-risk balance. Child-proof packaging may alsobe a means of reducing the likelihood of safety problems.
5. Suspension of product licence or investigational-status approval ‘‘Temporary suspension'' of a marketed product is an option in many European countries. It is considered when the magnitude of a safetyproblem is still to be fully determined. The suspension may be short-term orlong-term, and when the safety problem is resolved the manufacturer mayresume marketing without having to submit another new-drug-applicationdossier.
While products are still under investigation (pre-approval), most European countries and the USA can opt to suspend clinical trials (‘‘clinicalhold'') and can make the suspension permanent. ‘‘Clinical hold'' gives themanufacturer and the regulatory agency time to gather more informationand to review the safety issue in detail before additional patients are exposedto the product.
A related option is the suspension of reviews and decisions on marketing- application dossiers. In the case of felbamate, for example, all registrationprocedures in the European Union were ‘‘frozen'' in July 1994 after the firstreports of aplastic anaemia.
6. Withdrawal of the product from the market Although withdrawal is probably the least desirable option for either the regulatory agency or the manufacturer, it must be considered when itbecomes clear that a product's safety risks outweigh its benefits. In theinstance of quinine, a thorough review showed the weakness of the efficacydata and the substantial risk of severe adverse reactions.
Occasionally, it is necessary to consider and exercise the withdrawal option even when the benefit-risk balance has not been clarified, and when specificmeasures for reducing risk have not been identified. The dipyrone instanceillustrates this. Dipyrone was withdrawn from the market in the USA andSweden after reports of fatal agranulocytosis. Further epidemiological studyshowed that the excess risk of death from this reaction was 0.10 per milliondipyrone users a week. The addition of anaphylaxis, aplastic anaemia andgastrointestinal bleeding brought the total risk of death to 0.11 per millionusers a week, compared with 1.66 for aspirin and 1.50 for diclofenac. Yearslater, after resubmission of a new-drug application containing data from newstudies, the drug was re-approved for marketing in Sweden.
Another example is clozapine, which was also withdrawn from some markets after reports of agranulocytosis in Finland. It was subsequently re-approved after submission of new data, with a restricted indication forschizophrenia refractory to other therapy; in addition, mandatory white-blood-cell monitoring of patients is required wherever the drug is marketed.
In the most extreme instance of product withdrawal, the regulator may judge that the newly identified risk is an ‘‘imminent hazard'' to patients undertreatment or who might receive the drug. The regulator should then carefullyconsider alerting professionals and the public and doing an immediate recall,on the basis of the following suggested criteria and questions: How severe would be the harm that the drug might cause pending thecompletion of a customary review and procedure for its withdrawal orrestriction? How likely is the drug to cause such harm to users during a lessimmediate administrative process? What risk would immediate withdrawal pose to current users, given theavailability of other therapies and the need for patients to adjust to them? Are there other approaches to protecting the public health? What is the best judgment, given the information available at the time, ofthe likelihood that the drug will be withdrawn on the completion of themore customary administrative process? Product withdrawal, whether voluntary by the manufacturer or mandated, necessitates consideration of a product recall to health-careprofessionals and patients on a regional or global basis.
7. After the analysis: Communication of new or reinforced information to the medical profession or the public Critical to the protection of the public health is the duty to inform concerned parties when a new safety risk arises or there has been asignificant change in the benefit-risk balance of a product. There are severalways of doing so, once decisions on action are made. They include changesin the wording of the standard safety sections of prescribing information orpatient information leaflets, the addition of recommendations on thetreatment of adverse reactions, and restriction of indications. Newinformation might also include reinforcement on the appropriate use of aproduct, on dosage reduction schedules, on use of alternative therapies, oron the appropriate patient population, or ‘‘how to'' instructions on productadministration. The best methods of disseminating the information and thebest audience for its receipt should also be carefully considered. Methodsinclude: ‘‘dear doctor/health-professional'' letters, the use of patient leafletsand advertising campaigns to health professionals or consumers, journalpublications (scientific or lay press), and educational programmes/educational materials for health professionals or consumers via print,video, audio or computer (electronic) media.
This important step in the overall process is covered in more detail in Chapter III: Decision-making.
III. DECISION-MAKING After the technical aspects of evaluation and their limitations have been addressed, one or more of the various options enumerated and analysedmust be selected and implemented. The Working Group believes in regardto the selection of options that some fundamental concepts underlie gooddecision-making practices. As in most critical public health situations,decisions must be made on the basis of available evidence, tempered by priorexperience, political context and professional judgment - i.e., decision-making in the face of degrees of uncertainty.
A central concept applied in such judgments is ‘‘acceptable risk'' — the level of risk a population or group is prepared to accept in exchange forperceived benefits. Likewise, decision-making will be influenced by the levelof ‘‘acceptable uncertainty,'' namely the extent to which the case for anunfavourable shift in the benefit-risk balance has been proven. As a generalpremise, the less society is willing to accept risk in any given situation (e.g., aproduct used for prophylaxis in otherwise healthy people), the less certaintydecision-makers will require before they act on a new signal representing amajor safety hazard. The fundamental question of ‘‘acceptable to whom,''which entails the study of risk perception and risk toleration, remains asubject for continued research. For this discussion, however, the manu-facturer (responsible for making medications available to people who willexperience their associated benefits and risks) and regulators (responsiblefor protecting and advancing the health of the public) act on behalf of thetarget populations, who may be considered to have consented implicitly tosuch decision-making.
The Group observed that though details on decisions had been set out in formal legislation and regulations, little was known of the actual processesbrought to bear in making these decisions. A small survey of the members ofthe Group and selected outside companies concerning some of theirdecision-making practices revealed a broad range of approaches (or none)to the seeking and harnessing of professional expertise. It included the use ofpublic and private forums; the development and application of explicitcriteria; views about the involvement of stakeholders; public meetings andproceedings; and transparency of the process and rationale of the decisionsmade. The main results are presented below; for details see Appendix F.
The survey findings indicated a need to establish standards in benefit-risk decision-making. Several basic concepts are described below. It is suggested that by adopting the principles outlined in the next section decision-makerswill be in a position to learn progressively to improve decision-makingpractices and their consequences.
Basic approaches to good decision-making practices Decision-making means determining the actions to be taken, who should take them, and the order and methods of taking action; it also entailsdecisions on the best means of monitoring and follow-up, and ofcommunicating the appropriate information to the parties concerned.
It is suggested that decision-making should follow three principles: objectivity, equity and accountability.
Fundamental to all regulatory processes is that they must rise above impassioned arguments and partisanship and create a context in whichdecisions are based on the data available and are made without bias. TheWorking Group recognizes that precipitous decisions and actions may betaken sometimes in the face of urgent external pressures (e.g., legal, media-generated, consumer-driven), even when there is no clear and imminenthazard. Under such circumstances, however, attempts should be made togenerate support for scientifically-based, adequately informed decision-making, which can assure the public that it is being well served.
Objectivity relies on several attributes or qualities of people and The evidence base: This report contains suggestions on methods of representation, array and analysis of data. As far as possible,information obtained from all relevant sources and by a variety ofmethods — observational, epidemiological and experimental —should be gathered and analysed and factored into the decision.
Expertise: The best minds available from all relevant sectors must be brought to bear on these public-health decisions.
The survey showed that all regulators used publicly established and recognized committees for this purpose,and that the committees are consistently composed of leadingscholars and academics in the field. The use of ad hoc technicalconsultants who bring special expertise to the committee on anindividual decision, as the survey indicated, is to be commended.
The manufacturer: In general, the sponsor of a drug thatpresents an apparent change in the benefit-risk balance under-takes an analysis independently, before or at the same time asthe regulator. For this purpose, only three of the 11 surveyedcompanies had a standing external board to advise on benefit-risk issues; however, all reported the use of external consultantsto assist in difficult situations.
The consumer: The manufacturer and the regulator, with theirprofessional experts and advisors, must act on behalf of thepopulation and individual patients. Some countries are explor-ing ways in which patients, consumers or consumer advocatescan take part directly in decision-making (in our survey, onlythe United States regulator indicated that consumers orconsumer advocates were included on some of its expertadvisory panels). The Group understood the difficulty inachieving a balance between, on the one hand, public under-standing and early representation of consumers in decision-making and, on the other, the risk of raising false alarms andconfusing the public about difficult scientific and medicalissues. In this regard, there is need of means of guarding againstpremature or excessive public reaction.
Avoidance of bias and conflict of interest: Because often eminent experts have been associated with the discovery and development of newmedicines or are associated with direct market competitors, reviewersand referees must declare any relationships with the sponsors or theircompetitors which might compromise their ability to render anobjective judgment. Such conflicts of interest must be revealed on allrelevant occasions and discussed openly. If bias is probable, an expertmay be asked to testify but not to make a recommendation or to vote.
‘‘Scientific'' decision-making: None of the participant regulators uses computer-assisted algorithms to support regulatory decisions. Theyare used in other areas of public policy, however, and when usedappropriately have the distinct advantage that they can render explicitotherwise implicit or arbitrary assumptions and weightings of data.
The Working Group recommends that consideration be given to theuse of these powerful decision tools in this area.
Explicit predetermined criteria: When a drug is inherently more efficacious than others in its class, it would seem logical that itshould be permitted to bear greater toxicity. But are there qualitativeor quantitative criteria for how much better a drug should be topermit acceptance of higher toxicity? Or how much more toxic may it be and still have an ‘‘acceptable'' benefit-risk balance? Before tacklinga specific situation, it may be helpful to discuss the mattertheoretically and attempt to establish bounds of acceptability —namely ‘‘criteria'' against which a drug must perform. The survey ofparticipants revealed no development or use of such explicit criteria,and the Working Group knew of no such pre-established criteria ingeneral use in pharmaceutical decision-making. However, they areused for comparable benefit-risk purposes — in relation to environ-mental contamination, for instance — and their possible applicabilityshould be explored.
A corollary of objectivity, equity demands that all drugs and other therapeutic interventions be treated fairly by ensuring comparability inrequirements for evidence, analytical approaches, professional expertise anddecision-making rules. These standards can be described in the followingterms.
Choice of comparator: In assuring equity, there must be evidence that the benchmark product or intervention is in general use in thecircumstances in which the new benefit-risk balance will prevail.
Often, however, these comparator products have been on the marketfor many years and were approved under standards less rigorousthan today's. The complete data that would allow direct comparisonsmay not be available, therefore. Nevertheless, every effort should bemade to use the available evidence to render the best comparisonspossible.
Open process: Equity requires that the major stakeholders, at least the manufacturer/sponsor, take part in regulatory decision-making. Allsurveyed companies reported that the regulatory authority hadconsulted them and that they discussed benefit-risk matters with theregulator. The survey summarized the extent to which companiesparticipated in the most recent major safety issue handled by theregulators: – submitted a written report – consulted for testimony – took part directly in hearing or meeting – took part in the discussion Replication/consultation: Though not included in the regulatory survey, the question of interagency coordination was raised in each of thespecific illustrative cases reviewed by the Working Group (Appen-dix B). In none did different regulators considering the issues inparallel engage in full, open, coordinated dialogue. Neither was itusual for them to review or use another agency's summary ofevidence and conclusions (with the exception of review by theEuropean Agency for the Evaluation of Medicinal Products (EMEA)of data from individual Member States). While the need forindependence is understandable and differe¨nt national circumstancesmay call for different considerations, better cross-national commu-nication is strongly recommended.
Due process: Our survey showed that all regulatory authorities have a provision for manufacturers to appeal against a decision. In theEuropean Union the formal appeal process ordinarily applies in allMember States. Such appeals can necessitate multiple steps throughthe approving agency and often adjudication by an independentcommission or even the courts. In other words, other appointedrepresentatives of society at large may eventually determine ‘‘accept-ability.'' Specificity: No two drugs are identical; no two benefit-risk profiles will match exactly; no two institutional or health-care environmentalsettings are the same. Decision-making must be specific to thecircumstances. Therefore, in scientific, evidence-based decision-making, the criteria applied and the relative ‘‘weights'' with whichthey are brought to bear on the decision must be specified. It isparticularly important to indicate why one or another criterion mightnot be relevant to a specific benefit-risk decision.
Sensitivity: All societal benefit-risk decisions are by their very nature made in the face of uncertainty. Often this uncertainty can beexpressed as confidence intervals around point estimates for specificparameters, but a certain degree of subjectivity may be needed. Giventhese ambiguities, the assumptions underlying the decision need to betested. What if we are wrong? How important is it considered to beright for each of the parameters? Would a wrong assessment of asingle parameter change the decision? In other words what is the levelof uncertainty and does it exceed the ‘‘acceptable uncertainty'' fordecisions in this group of drugs or risks? If the uncertainty is too great,additional data may be required before a meaningful decision can bereached.
The extent to which society at large will come to appreciate thelimitations of science and technology in benefit-risk decision-makingis crucial to its acceptance of decisions made in the face of inevitableuncertainty.
Transparency: The process of decision-making, the participants and stakeholders concerned, the content of their arguments, the criteriaapplied to the decision, the results of sensitivity analyses — all need tobe expressed in a way that will enable others to consider theapplicability of the decision in different settings. This becomesespecially important in collaborative circumstances (e.g., the Eur-opean Union) in which decisions of an individual member state affectmaterially all the other states.
It is very difficult to translate these issues into a framework for publicunderstanding in controversial or emotional circumstances, particu-larly when there is considerable media interest. CIOMS WorkingGroup IV envisages a time when decision-makers will be able topresent their findings and the extent of uncertainty to the public. In sodoing, they will overcome the expectation of complete knowledge andcertainty, which does a disservice to all parties concerned.
Decisions are made at a specific time on the basis of the data then available. Circumstances change, however, and the effects of actions takenare rarely completely predictable. Therefore, expected outcomes (includingthe behaviour of health-care professionals and patients, as well as public-health impact) must be specified or estimated. In addition, as part ofdecision-making, criteria should be established for determining andassessing the effectiveness of the actions chosen. A recent example of suchan examination is the measurement of physician compliance with changesmandated in the use of terfenadine with contraindicated concomitantdrugs.1 After an action other than complete withdrawal of a drug, data must continue to be collected as part of the ongoing monitoring of safety.
Manufacturers and regulators have generally relied for this purpose on Burkhart, G.A., Sevka, M.J., Temple, R. and Honig, P.K. Temporal decline in filling prescriptions forTerfenadine closely in time with those for either Ketoconazole or Erythromycin. Clinical Pharmacologyand Therapeutics, 61: 93-96, 1997.
Thompson, D. and Oster, G. Use of Terfenadine and contraindicated drugs, JAMA, 275: 1339-1341,1996.
passive collection of data, mainly spontaneous reports. However, activesurveillance or a survey may be a better way to generate data to test forthe expected outcomes. The data should be periodically analysed todetermine whether (1) the outcomes are those desired, (2) newinformation dictates further review and a possible modification of theoriginal action, or (3) there are sufficient data to consider undertaking anew benefit-risk assessment.
One country surveyed (Canada) reported a ‘‘decisional impact analysis'' in which the broad range of expected consequences of regulatory actionsmust be stated as part of the final decision.
AND UNRESOLVED ISSUES Owing to the complexity of an overall benefit-risk assessment, several issues remain unresolved, although they were brought to the Group'sattention. Some were outside the Group's objectives and may serve as newtopics for CIOMS or other groups, while others involve geographical orregional cultural or medical practice issues.
Issues outside the scope of the report Although it was recognized that pharmaco-economic data may be important in weighing benefits and risks, economic impact was notdiscussed. The Group believes that economic considerations per se shouldnot influence the benefit-risk assessments covered in this report. Likewise,it did not consider the ultimate economic consequences of a benefit-riskassessment and any subsequent action. Risk and benefit evaluation wasconsidered from the perspective of patients' medical needs. Pending thefull development and validation of methods of assessing therapeutic costs,the cost of treating adverse events, loss of quality of life, and otheroutcome-related variables, the use of such methods is consideredpremature.
The Group did not address in any detail impact analysis of the decision options available after conducting a benefit-risk assessment. Also beyond itsobjectives were methods of assessing the success of communication andother action initiated with professionals and patients, and their subsequentbehaviour.
Benefit-risk assessment and decision-making can be subject to strong cultural and ethnic influences; different countries or populations may weighdifferently the same evidence for benefits and risks. This applies especiallyto the definition or perception of ‘‘acceptable'' risk.
Once regulatory/manufacturer decisions are made on the basis of a benefit-risk assessment, the results must be communicated to the partiesconcerned. Although this report suggests points to consider for this activity,it was beyond the scope of the project to develop specific recommendationson how, to whom, by whom and when such communication should bedirected. The Group recognizes the need for improvement and guidelines inthis area, and a separate CIOMS initiative is dealing with some of theseissues.
Formal decision-making theory has been increasingly applied to the practice of medicine1, but the Working Group did not consider elaborationof its techniques and their application to benefit-risk evaluation.
Unresolved issues How severe or serious or frequent should a risk be to trigger a benefit- risk re-assessment? There are quantitative approaches for determiningwhether a number of case-reports exceeds background risk2, and judgmentis always required, but more-systematic approaches would be valuable.
Some believe that a metric is needed that would combine severity and incidence of reactions. How could such an index be constructed andvalidated? Is it possible to establish an algorithm for benefit-risk assessment? This would require an expert system but all the values are not yet known andinferential rules are not validated.
A decision to withdraw a drug from the market may appear acceptable at the population level but individuals who might benefit from it might still optfor its use. This and other considerations affect the ethical aspects ofwhether benefit-risk assessments can and should be conducted for a patient,a population or a special group.
Risk perception and risk communication are central to the weighing of benefits and risks. However, it is extremely difficult to factor these into theassessment and decision-making processes described. Although assessmentis made and action taken from a public-health perspective on behalf ofindividuals, how much risk is a patient, given a choice, willing to take forhow much benefit? How relevant to everyday medical practice are very rarebut medically serious adverse reactions, and how should such informationbe communicated to a patient? These issues need considerable research. Anexample of the dilemmas faced, without resolution, relates to the morbidity Schumaker, G.E. Multiattribute Evaluation in Formulary Decisionmaking as Applied to Calcium-channel Blockers. American Journal of Hospital Pharmacists, 48: 301-306, 1991.
Eriksen, S. and Keller, L. Robin. A Multiattribute-utility-function Approach to Weighing the Risksand Benefits of Pharmaceutical Agents. Medical Decisionmaking 13: 118-125, 1993.
Sox, H.C., Jr., Blatt, M.A. Higgins, M.C., and Manton, K.I. Medical Decisionmaking. Butterworth,Stoneham, Massachussets, 1988.
Begaud, B., Moride, Y., Tubert-Bitter, P., Chasterie, A. and Haramburu, F. False-positives inSpontaneous Reporting: Should We Worry About Them? British Journal of Clinical Pharmacology,38:401-414,1994.
Edwards, R., Lindquist, M., Wilholm, B.-E. and Napke, E. Quality Criteria for Early Signals ofPossible Adverse Drug Reactions, Lancet, 336:156-158, 1990.
statistics associated with Stevens-Johnson syndrome and toxic epidermalnecrolysis as they affect the choice of various therapeutic options.3 By influencing public opinion, lay and scientific media often influence considerably a government or a company's opportunity to make rational,scientifically-based decisions. Ideally, health professionals should beinformed before decisions and information are communicated to the public.
However, a dialogue with the media is needed on these issues, which oftenhave ethical implications.
The elaboration of the many inadequately defined or poorly studied parameters affecting benefit-risk analysis or of the factors that influence itwould be very beneficial. Among the more obvious are better measures ofexposure of patients to drugs and of patient compliance as well as better useof the resulting data, the extent and impact of off-label use or misuse ofmedicines, dose-effect relationships (for desired and undesired effects),background/natural-history data for signal events in at-risk populations,and mechanisms of drug-induced diseases. Their elaboration might lead tomeans of identifying patient-groups or particular patients predisposed toserious adverse reactions.
Roujeau, J.-C., Kelly, J. P., Naldi, L. et al. Medication Use and the Risk of Stevens-Johnson syndromeor toxic epidermal necrolysis, New England Journal of Medicine, 333:1600-1607, 1995 and 334:922-923,1996.
V. OVERALL SUMMARY The key ideas and proposals of the CIOMS Working Group are presented in succinct form as ‘‘take-away messages'', in the order in whichthe subjects were discussed in Chapters I through IV, General Principles The approach proposed is intended to be applicable only in the case of a suspected major safety problem.
Typically, signals arise from market experience, but information from all sources must be used in benefit-risk assessment.
The CIOMS IV approach addresses the population at risk, not the patient at risk.
If crisis-management techniques are needed, a thorough benefit-risk assessment is still necessary.
Although, typically, a benefit-risk assessment is prompted by a new, important risk, the impact of the new signal must be reviewed in the totalcontext of the profile and use of the drug.
Problems may be intrinsically related to active ingredients, metabolites or excipients of products, and these are to be distinguished from accidentalor intentional contamination or defects.
Although economic considerations increasingly affect health-care systems, they should not influence the types of benefit-risk assessmentcovered here. Nevertheless, willingness to accept risk may depend on cost inmany parts of the world, where authorities may need to consider expense intheir decision-making.
As much cooperation and exchange of needed information as possible should take place as promptly as possible between the parties concerned,usually companies and regulators.
The need for urgent action to protect the public must be weighed against the need for additional data that might provide better conclusions.
Although the main thrust of this report is to provide guidance to manufacturers and regulators, the affected parties should be engaged in thedecision process when possible.
Introduction to a Report Briefly introduce the drug so that the report ‘‘stands alone'' and will not be misinterpreted.
Where and for how long has the drug been marketed? What are its indications by country? Are data from co-marketers or licensees included? Alternative therapies or modalities, including surgery, should be indicated at this stage, along with the reasons for the manufacturer'schoice. Ideally, fair comparison will require the same indications andduration of treatment for the alternatives. Also an effort should be made tomatch the severity of the disease treated, concomitant medications, and ageand sex distribution of the affected populations.
Benefit Evaluation (A more extensive check-list of points to consider in evaluating benefits is found at the end of Chapter II B.) A benefit-risk analysis should begin with a discussion of a drug's First, the epidemiology and natural history of the target disease should What is its incidence/prevalence? Are there high-risk populations? Is the disease self-limiting, fatal or disabling; does it have considerablemorbidity; or is the condition being treated an asymptomatic risk-factorfor subsequent disease? If the disease is the direct target of treatment, what is its associatedmorbidity and mortality? Comparative benefit-risk evaluation of a drug can be highly situational; it varies according to its intended application and the circumstances inwhich it is to be used, such as: prevention or treatment? life-threatening or self-limiting disease? to prevent disease progression or to cure? to treat chronic disabling symptoms or to reduce or delaymorbidity/mortality? as an orphan drug in a limited population with no therapeuticalternative? occurrence of adverse reactions with unapproved use? therapeutic response different in sub-populations? first-line or second-line therapy? Benefit may be defined with regard to the individual or to society (e.g., the net benefit of a vaccine to society).
What is the drug's absolute efficacy, viz., the number of deaths prevented or patients cured for symptomatic treatment? Over what length of time andin what percentage of patients is the drug effective? If it is designed to prevent disease, to what extent is the disease risk-factor affected and what is the associated reduction of disease risk? Where surrogate measurements are used, what is the evidence that they are valid markers? Take into account the extent to which benefits have been demonstrated and the quality of the data.
Are there any negative studies? Is the supportive scientific literature of high quality? Have such measurements as quality-adjusted-life-years beenvalidated? What is the degree of benefit achieved in relation to alternative therapies or interventions and their relative efficacy and effectiveness? Even if there is alternative therapy, what is the effect of no treatment? As tolerability is likely to affect compliance, the comparison of the drug with alternatives should also be discussed, with regard to route ofadministration, frequency of dosing, palatability and other factors relatedto convenience.
Unlike the benefits of a drug, which are usually distinct or well-defined, risks usually comprise a mixture of adverse reactions.
The medical impact of an adverse reaction is characterized by its frequency of occurrence, duration and intensity, and different reactions arenot directly comparable unless they can be expressed by a common measure.
Formal analysis of absolute and relative risks is never made with total General considerations in an analysis Begin with a history of how the new issue was identified.
Provide a thorough review of the evidence associated with the new safety issue, including temporal relationship, physicochemical characteristics,possibility of a class effect, possible role of interactions, backgroundincidence of the event, and the possibility of a subgroup at risk.
Discuss the strength of the evidence (as described in detail in the report of CIOMS Working Group III).
Address the preventability, predictability and reversibility of the The adverse reaction that dominates the overall risk profile (carries the most weight) is referred to as the risk driver or dominant risk.
In addition to data on the signal adverse reaction, all other adverse- reaction data on the product should be considered in describing andquantifying the overall risk; this applies especially to other serious orfrequent reactions.
Typically, a single medicinal product is evaluated for risk without reference to its alternative therapies, but comparative risk assessments,which are crucial and more relevant, are more difficult, for many reasons.
For comparators, three key elements, namely qualitative descriptions of the relevant adverse reactions (including their duration), their frequency,and a weighting of relative importance (‘‘seriousness''), should optimallyprovide an opportunity for numerical comparison.
As a first step, the three most often reported and the three most serious adverse reactions can be chosen as representatives for the risk profile of eachmedicine in the comparison.
For risk weighing for individual adverse reactions, there is no obvious metric or methodology.
What methods are available? Some companies have lists of medically serious conditions; the WHO Collaborating Centre for International DrugMonitoring uses a ‘‘critical terms'' list.
Agreed scales or values for adverse reactions would contribute to standardization; the issue is an important area for research.
The Working Group suggests the use of standard visual presentations, usually as bar graphs, to represent risk profiles of individual drugs,classified by subgroup (e.g., serious, fatal, or other adverse-reactioncategories). Units, periods of time covered and other key elements integralto interpretation should be clearly indicated on any form of presentation.
To avoid combining data inappropriately, separate profiles should be established for data from different sources, such as spontaneous reports andclinical trials.
In the absence of comprehensive information, other data sources might include official data-sheets or data from regulatory registries (e.g., from theUnited States Food and Drug Administration through the Freedom ofInformation mechanism or from the ADROIT data-base of the UnitedKingdom Medicines Control Agency).
Different populations, indications and dosage forms may also necessitate separate risk profiles.
Quantification of risk: To put the newly identified risk into perspective, itshould be quantified in terms of incidence, but this is usually difficult todo precisely.
Excess incidence of reaction (‘‘drug attributable incidence'') is a more meaningful measure. Calculation of within-drug excess incidence fromcomparative clinical trials or cohort observational studies is straightfor-ward, but case/control studies are somewhat more complicated.
It is more difficult to compare various alternative drugs for incidence of adverse reactions, in part because of their usually different marketinghistories.
The distinction between ‘‘absolute'' and ‘‘relative'' risk must be kept in mind. The significance of a high relative risk (e.g.,1.5) depends considerablyon whether the actual incidence of an event is 1 in 10 or 1 in 100,000.
Estimated total drug risk: Once standard units are selected as markers forcomparison, the weighted excess incidence rate can be calculated for eachadverse reaction by multiplying the excess incidence rate by the weightfactor specific to the adverse reaction.
To obtain the weighted overall excess incidence of the combined reactions of interest, the sum of the individual values provides an aggregate(e.g., for the three most serious reactions).
To obtain the incremental drug-attributable risk, the aggregate values must be subtracted from a reference standard treatment (or no treatment).
Validity of the estimates: To test the robustness of the risk estimation,sensitivity analyses can be performed. Great care must also be taken inany attempt to extrapolate beyond the specific populations; if total riskestimates are extrapolated all assumptions must be stated.
Suggested sequential overall approach: The Working Group suggests a12-step, general methodological approach to risk assessment.
Benefit-Risk Evaluation Comparison of benefits and risks, even of a single product, is a deceptively simple concept; comparison of the trade-offs between benefitsand risks among various products is particularly difficult.
One problem is that benefits and risks are usually expressed in different parameters and units. Some simple attempts have been made to combinebenefits and risks, such as: global assessment made by a treating physician to reflect apatient's overall response to treatment a benefit-risk ratio, defined as the number of efficacy events neededfor each occurrence of a serious adverse event an ‘‘incremental benefit-risk ratio,'' i.e., the number of additionalcases of severe toxicity for each therapeutic success.
In addition to efficacy (morbidity under ‘‘ideal'' conditions) and effectiveness (under ordinary conditions) other measures might includequality of life, quantity of life and cost efficiency. However, more research isneeded to test their validity for use in comparative benefit-risk assessment.
Certain basic principles apply in benefit-risk assessment: All available relevant data on benefits and risks should be assembled Be transparent in the way data are presented Comparable outcome measures should be used for comparison Explore whether it is possible to determine whether a specific subgroup is at particular risk Adverse reactions are characterized by their: Benefits can be assessed in terms of the effects of the drug on the target disease as regards its: extent of control of the target disease The incidence metric for adverse reactions does not apply to benefits except for prophylaxis or vaccination (what is the incidence of the diseasewith and without the preventive agent?).
Among different approaches recommended, use the above descriptors to express benefits and risks associated with each product, focusing on thethree most serious and three most common reactions, including the signal-generating event.
Judgment is needed in any categorization rubric of benefits and risks (e.g., categorization by levels of seriousness, duration, incidence); theconceptual framework for such a scheme may be used as a simple algorithmto maintain a consistent logic.
It may be possible to represent benefits and risks quantitatively and Use of the expression ‘‘benefit-risk ratio'' or any single ‘‘overall'' expression of risk or benefit is discouraged.
It is emphasized that the approaches discussed involve a consistent logic but cannot always serve as standard methodologies. There is no consensuson a single method; indeed different approaches are likely to be needed forindividual situations.
The results should always be framed in the proper therapeutic context and presented in terms of both relative and absolute expressions of benefitand risk.
Among the options for action available to regulators and manufacturers are the following: maintaining the status quo ‘‘watching and waiting'' (i.e., monitoring subsequent experience) additional data-gathering/initiation of new research (clinical ornon-clinical) modification of product information, distribution or manufacture restriction of product availability suspension of product licence withdrawal of the product from the market Whenever possible, the advantages and disadvantages (pros and cons) of each option should be described in the context of the specific situation. Alsoinclude an impact analysis which outlines the anticipated consequences.
Central to the interests of all, whichever the chosen option, is the need for clear and effective communication.
Decisions must be made on the basis of the available evidence, tempered by prior experience, political context and professional judgment.
A central concept is ‘‘acceptable risk,'' but the fundamental question is: ‘‘acceptable to whom.'' In decision-making there should be three broad principles: objectivity, equity and accountability.
Objectivity relies on: the evidence base avoidance of bias and conflict of interest Equity demands comparability in terms of: choice of comparator open process (involvement of all relevant parties) replication/consultation (e.g., interregulatory dialogue) due process (a provision to appeal a decision) specificity to the circumstances sensitivity (i.e., what is the level of uncertainty) transparency (i.e., the results expressed in a way which will enableothers to understand the rationale of the decision).
Accountability requires that expected outcomes are defined and criteria specified for assessing the effectiveness in practice of the option chosen.
Following an action other than complete drug withdrawal, collection of data must be continued in order to monitor the expected outcomes.
Work is encouraged on development of more powerful decision tools and better measurement criteria for a medicine's performance.
Unaddressed and Unresolved Issues Issues outside the scope of the report Economic considerations per se Impact analysis of the decision Benefit-risk evaluations and decision-making in relation to cultural and ethnic influences Specific recommendations on how, to whom, by whom and when communication on the results and decisions should be directed Decision-making theory and its application to benefit-risk evaluations.
Unresolved issues How substantial must a risk be to trigger a benefit-risk assessment? A metric combining severity and incidence of adverse reactions An algorithm for conducting benefit-risk assessment The ethical aspects of whether benefit-risk assessments can and should be conducted for a patient, a population or a special group Risk-perception and risk-communication: how best to factor these elements into the assessment and decision-making processes Dialogue with the media on their role, which often raises ethical Finally, the elaboration of many influences and parameters would be beneficial: e.g., better measures of patient exposure and patient compliance and better use of the data, the extent of off-label use or misuse of medicines,and data on the natural history (background incidence) of signal events inthe populations at risk.
Key terms are defined or described here. Most have no universally accepted definitions, though some regulatory authorities have adopted‘‘official'' definitions for some. Consequently, the definitions and descrip-tions are annotated with explanations to ensure that the meanings providedare understood with regard to their use in the context of this report.
Whenever possible, literature or other citations are given for previouslyestablished definitions. The absence of inverted commas for referenceddefinitions indicates that they were paraphrased for use in the presentreport.
Benefit: Benefit usually refers to a gain (positive result) for an individualor a population. ‘‘Expected'' benefit can be expressed quantitatively, andthis would ordinarily incorporate an estimate of the probability ofachieving the gain.
These uses of the term benefit are those employed in this report.
Some current definitions of benefit include reference not only to clinical improvement but also to quality of life and economic consequences, as inthe following example:1 ‘‘The improvement attributable to the drug, in terms of human health, health-related quality of life, and/or economic benefit to the individual orgroup.'' This definition is beyond the scope of the concepts developed in this CIOMS report.
Dominant risk: The risk that is considered to be the major contributor tothe overall risk profile.
Note: Other terms used to describe the dominant risk are, e.g., primary risk or riskdriver. Dominant risk is the one adverse reaction that outweighs the others in theoverall risk profile and risk management of the product.
Efficacy: Efficacy is the ability of a medicine or medical technology tobring about the intended beneficial effect on individuals in a definedpopulation with a given medical problem, under ideal conditions of use.
Note: Efficacy generally refers to how well a particular medicine will bring aboutthe intended effect under ‘‘ideal'' or near ideal conditions, as in a clinical-trialsetting, for example.
Benefit, Risk and Cost Management of Drugs. Report of the CPHA National Advisory Panel on Risk/Benefit Management of Drugs. Canadian Public Health Association, January 1993.
Effectiveness: Effectiveness is a measure of the effect a medicine (ormedical technology) is purported, or is represented, to have underconditions for the use prescribed, recommended or labelled.
Note: Effectiveness refers to how well a drug achieves its intended effect in theusual clinical setting (‘‘real world'') and reflects its impact in the community(benefits observed at the population level).2 Hazard: A situation that under particular circumstances could lead toharm. A source of danger.
Relative risk: The ratio of the incidence rate of an outcome (event) in anexposed group to the incidence rate of the outcome (event) in anunexposed group.3 Risk: The simple, standard, epidemiological definition of risk is theprobability that something will happen.
Note: In the context of medical interventions (drugs, e.g.), the ‘‘something'' isalmost always associated with a negative event. In defining or describing a specificrisk, it is always important to include information on intensity (severity, e.g.),time of the event (onset or duration), and time period over which the probabilityapplies. Some definitions attempt to include concepts of rate, intensity and time: The probability of the occurrence of an adverse or untoward outcome and theseverity of the resultant harm to the health of individuals in a defined population,associated with the use of a medical technology for a specified medical problemunder specified conditions of use.
Risk evaluation: Risk evaluation is the complex process of determiningthe significance or value of the identified hazards and estimated risks tothose concerned with or affected by the process.
Risk management: The making of decisions concerning risks, or action toreduce the consequences or probability of occurrence.4 Serious: Usually the word ‘‘serious'' has two connotations. One is thecommon use of the term ‘‘medically serious,'' implying a diagnosis orcondition that is dangerous, critical or alarming. The other is aregulatory-administrative definition created for purposes of definingregulatory reporting obligations for adverse reaction reports. Althoughdifferent regulators use several similar definitions, the following definitionencompasses Abramson, J.H., Survey Methods in Community Medicine, 4th Edition, p. 49. Churchill Livingstone,New York (1990); and Cochrane, A.L. Effectiveness and Efficiency, Random Refections on HealthServices. Nuffield Provincial Hospital Trust, London, 1972.
B.L. Strom, ed., Pharmacoepidemiology. John Wiley and Sons, New York, 1994.
Risk: Analysis, Perception and Management. Report of a Royal Society Study Group. The RoyalSociety, London, 1992.
all of them and is the official definition given in the 1995 Guideline onexpedited reporting of adverse drug reactions, of the InternationalConference on Harmonization of Technical Requirements for Registra-tion of Pharmaceuticals for Human Use (ICH)5: A serious adverse event (experience) or reaction is any untoward medical occurrence that at any dose: is life-threatening requires inpatient hospitalization or prolongation of hospitalization results in persistent or significant disability/incapacity is a congenital anomaly/birth defect.
Medical and scientific judgement should be exercised in deciding whether expedited reporting is appropriate in other situations, such as importantmedical events that may not be immediately life-threatening or result in deathor hospitalization but may jeopardize the patient or may require interventionto prevent one of the other outcomes listed in the definition above. Theseshould also usually be considered serious.
Severe/Severity: The term severe is not synonymous with serious in thiscontext. Severe is used to describe the intensity (severity) of a specificevent (as in mild, moderate or severe myocardial infarction).
Signal: A report (or reports) of an event that may have a causalrelationship to one or more drugs; it alerts health professionals andshould be explored further.6 Note: In addition to information on a new (unexpected), potentially importantevent, a signal can refer to an unexpected finding, or a finding exceeding adetermined threshold, for an already known event — for example, data involvingthe nature (specificity), intensity or rate of occurrence.
Gordon, A.J. Implementation and Impact of ICH Guideline E2A: Definitions and Standards forExpedited Reporting. Proceedings of the Third International Conference on Harmonization, QueensUniversity, Belfast, pp. 461-469, 1996.
Hartzema, A.G., Porta, M.S. and Tilson, H.H. Pharmacoepidemiology: An Introduction. Harvey,Whitney Books. Cincinnati, Ohio, 1988.
MEMBERSHIP AND ACTIVITIES OF CIOMS WORKING GROUP IV The Working Group was composed of representatives of five European and four United States pharmaceutical companies; of the European Agencyfor the Evaluation of Medicinal Products (EMEA) and of regulatoryauthorities in Canada, Denmark, France, Germany, Italy, Japan, Sweden,the United Kingdom, and the United States of America; of the WorldHealth Organization (WHO) (Geneva) and the WHO Collaborating Centrefor International Drug Monitoring (Uppsala, Sweden), and, as observers, ofthe EMEA and the Bundesverband der Pharmazeutischen Industrie (BPI,Germany). There were also three independent consultants and one invitedexpert. The Group met six times between January 1995 and October 1996;sub-groups met at various times through the end of 1997.
The Group outlined the scope of the task and explored concepts.
Concepts were refined and the report chapters were defined.
Small sub-groups were allocated individualchapters.
Twenty-four members independently scored20 different adverse-reaction terms selectedarbitrarily by Win Castle.
The first drafts of each chapter were discussed.
mainly related to gathering and summarizingdata on the drug case-histories chosen asillustrative examples. Fourteen members re-peated the earlier survey exercise withoutaccess to their earlier assessments to determinewhether the results were reproducible.
The case histories were presented and agreed.
Hugh Tilson and Sue Roden presented theresults of their survey on how regulators andmanufacturers made decisions for action onbenefit-risk reviews.
Each chapter was reviewed, and Win Castle and Arnold Gordon undertook to prepare thefirst draft of the report.
The first draft was reviewed and amendments, corrections and additional modifications weresuggested.
Review of draft 2 led to more changes. Arnold Gordon prepared another draft, which Win Castle and Mac Lumpkin reviewed; after final changes and corrections the manuscript wassubmitted to CIOMS in February 1998 forpublication.
The full-time members of the Working Group and their affiliations at the Zbigniew Bankowski – CIOMS (Geneva) Christian Be´nichou – Synthe´labo (Paris) Annekarin Bertelsmann – EMEA Rudolf Bruppacher – Ciba-Geigy and as independent consultant – Independent consultant (Raleigh, North Carolina) and SmithKline Beecham (US)(Co-Chair) – Independent consultant (Gloucester, Mas- – Roussel Uclaf (Paris) – WHO Collaborating Centre for Interna- tional Drug Monitoring (Uppsala, Sweden) – Pfizer (New York) (Editor in Chief) – Bundesinstitut fu¨r Arzneimittel und Medi- zinprodukte (Berlin) – Food and Drug Administration (Rockville, Maryland, USA) (Co-Chair) – HMR (Frankfurt am Main) – Eli Lilly (Saint Cloud, France) – Glaxo Wellcome (Greenford) (Secretary) – HPB (Ontario) – University of Copenhagen and CPMP Barbara Sickmu¨ller – BPI (Frankfurt am Main) – Merck Research Laboratories (West Point, Ryuichi Takahashi – Ministry of Health and Welfare (Tokyo) – World Health Organization (Geneva) – Glaxo Wellcome (Raleigh, North Carolina) Bengt-Erik Wiholm – National Board of Health (Uppsala, Sweden) – Medicines Control Agency (London) The following attended one or two meetings only: – Ministry of Health (Rome) – International Federation of Pharmaceutical Manufacturers Associations (Geneva) – Takeda Chemical Industries (Osaka) – Ministry of Health and Welfare (Tokyo) – Pharmacologie Hoˆpital Fernand Widal Carmela Santuccio – Ministry of Health (Rome) – Ministry of Health and Welfare (Tokyo) – HMR (Frankfurt am Main) Jean-Michel Weiss – Hoffman-La Roche (Basel) Dr William W. Lowrance (Geneva) provided consultation during the ILLUSTRATIVE CASE HISTORIES OF BENEFIT-RISK ASSESSMENT The seven case histories differ in format and content, largely because of the differing amounts and types of information available on the clinicaldata, the regulatory review process, options for decisions considered and thedecision-making process. The information was obtained from publicregulatory records, publications, and non-proprietary company sources.
1. Quinine and allergic haematological events 2. Felbamate and blood dyscrasias 3. Dipyrone and agranulocytosis 4. Temafloxacin and renal impairment, and hypoglycaemia in elderly 5. Remoxipride and blood dyscrasias 6. Clozapine and agranulocytosis 7. Sparfloxacin and phototoxicity Quinine had over-the-counter (OTC) status in the United States and was indicated for the treatment of nocturnal leg cramps. It was manufactured bya number of companies and regulated by standards set out in a tentativemonograph. As it was marketed prior to 1962 it remained on the marketuntil it could be reviewed for safety and efficacy. In the 1970s and '80s publicpanels were set up to examine OTC products and decide whether theyshould be discontinued, given full final monograph status (i.e., full datawere available), or reviewed later when new data became available. Quininewas placed in this last category.
On 22 August 1994 there was a general review of the safety of quinine for OTC status because of allergic haematological events.
Quinine was available as 200 mg and 300 mg tablets.
Data were available from only six clinical trials, sponsored by different manufacturers. The data did not support the efficacy of the product, eitherbecause the analyses were flawed, or several confounding factors were present,or there was no demonstrable difference between quinine and placebo.
Sensory disturbances and visual and auditory symptoms (symptoms of quinine toxicity) were seen in special studies in which therapeutic doses wereused.
Serious hypersensitivity reactions had occurred, including thrombocyto- penia. Patients had been hospitalized and some had died. There was no methodof predicting and, therefore, warning about thrombocytopenia, whichoccurred at an incidence of between 1 in 1000 and 1 in 3500 treated patients.
The options available were: remove from the market for the indication transfer to prescription status restrict to compassionate use (named-patient use) However, a drug cannot be transferred from one status to another if it is not effective. If a drug can be OTC it should be. To be a prescriptionmedicine it would be necessary to show that the indication cannot be self-diagnosed or the drug used safely.
After February 1995, quinine was no longer recognized in the United States as safe and effective for the treatment of leg cramps, and theindication was withdrawn.
The process was transparent and the public and the profession were informed by press releases and articles in nursing and medical journals.* * See Federal Register, Volume 59, No. 161, Monday, August 22, 1994. 21 CFR Part 310 Drug Products for the Treatment and/or Prevention of Nocturnal Leg Muscle Cramps for Over-the-Counter Human Use;Final Rule. Also see Chase, S.L. FDA Yanks Leg Cramp Product off the OTC Market, R.N., 58(5): 71,1995 and Leclerc, K.M. and Landry, F.J. Benign Nocturnal Leg Cramps. Current Controversies OverUse of Quinine, Postgraduate Medicine, 99(2): 181-184, 1996.
Points illustrated A mild, self-limiting disease treated with a drug that caused a rare, serious (sometimes fatal), and unpredictable reaction Problems associated with old drugs: a different process is probably required as there may be no proper efficacy data Effect of OTC status as distinct from prescription status Consequences of decision Risk estimation in the absence of efficacy.
Felbamate is an antiepileptic drug, marketed in the United States since September 1993 and in Europe since March 1994.
The drug was first indicated as monotherapy or adjunctive therapy for partial epileptic seizures which may or may not become generalized in adultsand as adjunctive therapy for partial and generalized seizures associatedwith Lennox-Gastaut syndrome in children. Except for one country, theEuropean Commission through its Committee on Proprietary MedicinalProducts (CPMP) (March 1994) approved the drug for monotherapy oradjunctive therapy for adults and for adolescents of over 14 years withpartial onset seizures not controlled with other anti-epileptic agents. Francewas the rapporteur.
In June 1994 the company sent reports of five poorly documented spontaneous cases of aplastic anaemia to the different authorities. Furtherto other reports of blood dyscrasia (leucopenia and thrombocytopenia)occurring during a compassionate-use programme, and following threeother cases of aplastic anaemia, the European Commission's Committee forProprietary Medical Products in agreement with the French authoritiesdecided in July 1994 to suspend all registration procedures and marketing,and recommended that haematological and clinical follow-up studies beconducted.
In August 1994, after examination of new cases of aplastic anaemia, the French authorities discontinued compassionate use and stopped inclusionof new patients in clinical trials. The US Food and Drug Administrationinformed the prescribers of the occurrence of aplastic anaemia and modified the package insert to include warnings, restriction of indication, and arecommendation on haematological monitoring.
From entering the market in the United States up to the beginning of October 1994, 32 cases of aplastic anaemia were reported; 16 serious hepaticdisorders were also reported to the French regulatory authorities.
Based on an estimated treated population of 120,000 patients by October 1994, the incidence can be estimated as at least 1/4000 treated patients. Nocases were reported during clinical trials in the development programme, inwhich about 1600 patients took part; there were some reports of decreasedwhite-blood-cell counts, but none of aplastic anaemia or agranulocytosis.
The characteristics of the cases are as follows: All the patients developed pancytopenia with severe thrombocytopenia,and in all cases the bone-marrow biopsy was consistent with aplasticanaemia; most patients presented with signs of infection or bleeding. In37% of cases, pancytopenia was discovered during a systematic bloodcount. There were 21 females and 11 males of 12 to 70 years of age andone child of 21 months.
In 34% of cases the onset was sudden, with an immediate diagnosis.
The median time to onset was five months. In one case the diagnosis wasmade nearly six weeks after felbamate had been discontinued.
In 34% of cases felbamate was the only drug that could have beeninvolved.
In two cases rechallenge was positive.
In all cases the outcome was serious. Nine died (28%) and four othersreceived transplants.
The clinical-trial pattern is similar for all cases — aplastic anaemia seems to be a severe complication of treatment with felbamate and occurs after amean duration of treatment of about 5.6 months. The occurrence is sudden,with serious thrombocytopenia that regresses very slowly. A causalrelationship seems probable. All cases have the same profile. In a few casesthere was no other aetiology or drug that could account for aplasticanaemia.
Lennox-Gastaut syndrome is a dramatic, life-threatening epileptic encephalopathy, which occurs in a well-defined population, especially children. Most patients are often refractory to all antiepileptic drugs;felbamate has demonstrated its efficacy in atonic seizures. The onlytherapeutic alternative is surgery, and its associated risk must also be takeninto account.
Aplastic anaemia is a very rare but severe disorder, with a background incidence of 1-20/1,000,000 person-years; the risk associated with felbamateis considered to be about 200 times greater than the background incidence.
The prognosis is poor, with a death rate of about 25%.
It is difficult to know whether the drug-induced cases could have been prevented by blood monitoring and early discontinuation of the drug;however, regular monitoring may prevent irreversible aplastic anaemia.
Given the available data, it is impossible to determine the mechanism; the mean interval of onset is not characteristic of an immuno-allergic reaction.
Thus, an idiosyncratic reaction is probable but no clear risk factors havebeen found.
Options considered Complete withdrawal: Felbamate is the only drug that leads to a majorimprovement in the quality of life, with a dramatic decrease in thenumber of fits, in Lennox-Gastaut syndrome. It appears to be the lastalternative treatment before surgery.
Maintenance of the marketing authorization with: very restrictive indications; felbamate could be kept only assecond-line treatment for patients with Lennox-Gastaut syndromeor partial-onset seizures that are refractory to all availableantiepileptic drugs, including the most recent informed consent from the patient before felbamate therapy isbegun strict conditions for the prescription and restriction of supply(hospital use, neurologists) clear information on the risk of aplastic anaemia and hepatitis, andrecommendations on monitoring of blood and liver (Summary ofProduct Characteristics (SmPC)) modifications, boxed warning,Dear Doctor letter).
Withdrawal of the marketing authorization but making the drug availableto treat a small number of patients under compassionate use.
The Committee for Proprietary Medical Products (CPMP) The CPMP at its December 1994 meeting decided to keep the drug on the market because a very limited epileptic population needed it, with strictprescribing conditions and requirements for treatment surveillance.
Because of the seriousness and the frequency of aplastic anaemia reported with felbamate, but in view of the expected benefit of the drug tochildren resistant to other antiepileptic drugs, the indication was restrictedto Lennox-Gastaut syndrome in patients aged 4 years or more andrefractory to all available antiepileptic medicinal products.
The company was obliged to undertake a post-authorization surveillance study in all patients receiving felbamate. To do so, the company proposed tosend a letter to all potential felbamate prescribers, and to establish a registerof all treated patients (pre-paid card, to be returned to the company at theinitiation or discontinuation of treatment). All potential prescribers weresent a letter setting out the risks and recommended monitoring.
The Summary of Product Characteristics (SmPC) was subsequently The Food and Drug Administration (FDA) In the United States felbamate is restricted to patients with such severe epilepsy that the benefits of felbamate outweigh its risks (refractoryepileptics who respond inadequately to other anticonvulsant drugs).
As of 1996, of 20,000 US patients originally on felbamate, 10,000 were still taking the drug after inclusion of the black box warning, restrictedindications, and 26,000 Dear Doctor letters. There had been 10 cases ofhepatic failure, of which four had been fatal and one had had a livertransplant.
Points illustrated This example highlights the difficulty of a benefit-risk assessment when the disease is so severe that a very serious drug risk can be acceptable. Thedecision-making took into account the need for each physician to make abenefit-risk assessment for each patient.
Dipyrone is one of the non-narcotic analgesics (NNA) that have no anti- inflammatory action. It is an effective medicine for the treatment of feverand pain and is used in more than 100 countries, including Germany,France, Spain, Italy and Finland.
Dipyrone was introduced into clinical therapy in 1922 and registered in Sweden in 1934 for the treatment of various states of acute pain. Because ofthe risk of agranulocytosis it was withdrawn from the Swedish and USmarkets in 1974 and 1977, respectively. At that time, the number ofspontaneous reports suggested a causal relationship between dipyrone andagranulocytosis.
To provide population-based evidence for the adverse public-health impact of dipyrone, the International Agranulocytosis and AplasticAnaemia Study (IAAAS) was carried out in 1980-1986. It provided a moreprecise assessment of the occurrence of agranulocytosis and its associatedmortality than earlier studies had. The re-evaluation of the clinical safetydata indicated that the risk of agranulocytosis had been overestimated in the1970s. Consequently, in 1994 Hoechst resubmitted an application formarketing approval of dipyrone in Sweden.
In addition to its antipyretic activities dipyrone is an effective oral and parenteral treatment for moderate to severe post-operative pain. Oraldipyrone is superior to acetylsalicylic acid and comparable or superior toparacetamol. In its parenteral form it compares in analgesic efficacy to non-steroidal anti-inflammatory drug (NSAID) therapy and weak opioids.
Materials and methods All potentially fatal adverse events (AEs) reported in connection with non- narcotic analgesics (NNAs) were identified: anaphylaxis, Stevens-Johnsonsyndrome and toxic epidermal necrolysis (SJS/TEN), agranulocytosis, aplasticanaemia, upper gastrointestinal complications (haemorrhage or perforation),toxic liver disease and end-stage renal failure. The Medline database wassearched for all English-language reports of epidemiological studies publishedbetween January 1970 and November 1995 on the association of those adverseevents with NNAs, including dipyrone and NSAIDs.
All studies with a case-control or a cohort-study design that provided data to assess the excess mortality in the population studied were selected. Forcase-control studies the required data included estimates of the relative riskassociated with drug exposure compared with non-exposure, the percentageof cases exposed to the drug, the overall incidence rate of the disease in thesource population, and the case-fatality rate. Where the original article didnot include all the necessary data, a Medline literature search was performedto locate auxiliary reports by the study investigators, containing estimates ofthe overall risk of disease in the study population or of the case fatality rates.
For each study the excess mortality attributed to short-term use (a one-weekperiod) was calculated by multiplying the estimate of the one-week excess riskby the case-fatality rate of the adverse event. One-week risks of disease in thegeneral population were estimated from risk estimates for longer timeperiods, on the assumption of a constant risk of disease.
The following procedure was used to calculate from the case-control studies the one-week risk difference: the aetiological fraction (EF), i.e., theproportion in the study population of studied adverse events that could beattributed to specific NNA or NSAID use, was estimated by [(RR-1)/RR]Pe, where RR is the relative risk of the specific adverse event in theexposed (users of a specific drug) compared with the non-exposed (non-users of the drug) and Pe is the percentage of exposed cases (number of casesexposed to the drug/total number of cases).
Multivariate RR estimates, when provided, were used in the calculations.
The risk of the adverse event in non-users (Ru) was calculated as (1-EF)RT,where RT is its overall weekly incidence in the study population. Its risk inusers (Re) was calculated by multiplying the risk in the non-exposed by theestimated relative risk in the exposed, (Ru)(RR). The excess mortality wassummed across adverse events to determine the overall excess mortalityattributed to short-term use of each NNA. When more than one of thestudies assessed the association between a drug and a given adverse event,the median excess mortality estimate for the drug was used.
Ten case-control studies contained the necessary information to calculate four adverse-event-specific excess mortality rates attributed to NNAs: oneon anaphylaxis(1), one on agranulocytosis (2), one on aplastic anaemia (2),and seven on serious gastrointestinal complications, including haemorrhageand perforation (3-9). Only for dipyrone, propyphenazone, aspirin and Reference to the studies cited and the literature are found at the end of this case history.
diclofenac were relative risk estimates for all four adverse events given.
Excess mortality rates were calculated for all NNAs (dipyrone,propyphenazone and paracetamol) and NSAIDs (aspirin, diclofenac,indomethacin and naproxen), which were investigated in at least onestudy of gastrointestinal complications.
Of the seven studies examining gastrointestinal complications, three (4, 7, 9) restricted the analyses to hospitalizations or deaths due to bleedingor perforated peptic ulcer, and another (8) provided data to evaluatebleeding due exclusively to peptic ulcer. For five studies (4-8) the estimatesof the overall incidence or the case fatality of the gastrointestinalcomplications were obtained from auxiliary reports.
Anaphylaxis is the maximum variant of an immediate type reaction that occurs in previously sensitized persons after re-exposure to the sensitizingantigen. A clinically indistinguishable syndrome which is not antibody-mediated and does not require previous exposure to the antigen is called ananaphylactoid reaction. The anaphylactic and anaphylactoid responseappears usually within minutes of administration of the specific antigen.
It is characterized by cutaneous, gastrointestinal, respiratory, cardiovas-cular or central-nervous symptoms that can occur either singly or incombination. Life-threatening conditions involve respiratory obstructionleading to respiratory failure or cardiovascular collapse or shock.
Only one study, in the Netherlands, provided the necessary data for the calculation of the excess mortality. It attributed to drugs 107 of 336 casesclassified as probable or possible anaphylaxis, indicating an annual incidenceof drug-induced anaphylaxis of 3.7 per million (107 cases in a two-year periodin a population of 14.5 million). The multivariate RR estimates foranaphylaxis leading to hospital admission were given separately for 1987and 1988. Therefore, a pooled estimate was calculated by the methodoutlined by Greenland (10). In the category ‘‘other analgesics,'' the authorsincluded salicylic-acid derivatives, pyrazolones and dipyrone. The relativerisk of these agents was assumed to be identical, therefore. The excessmortality from anaphylactic shock ranged from 0.001 per million forpropyphenazone and paracetamol to 0.004 per million for diclofenac.
Agranulocytosis, defined as a granulocyte reduction below 0.5x109/L, is associated with clinical symptoms of acute bacterial infections. Its initialmanifestations are due to the severity, site (e.g., lungs, oropharynx, urinary tract or blood) and duration of the infection. Typical symptoms are feverand sore throat. Two mechanisms of drug-induced agranulocytosis havebeen proposed: immune and toxic. The immunological mechanism iscorrelated with rapid peripheral destruction of granulocytes in a previouslysensitized patient. Drug-dependent antibodies adhere to circulatinggranulocytes, causing their abrupt destruction. The toxic type of the diseasealso leads to sudden destruction of granulocytes. However, this may occurafter a latent period during which the patient receives a substantial amountof the causative drug resulting in damage to granulocyte precursors andmature peripheral granulocytes. There is considerable evidence thatdipyrone-induced agranulocytosis is immune-mediated.
The International Agranulocytosis and Aplastic Anaemia Study (IAAAS) carried out a population-based case-control study in Germany,Italy, Hungary, Spain, Israel, Sweden and Bulgaria; it gave an overallannual incidence rate of community-acquired agranulocytosis of 3.4 permillion (2). Estimates of annual incidence rates differed by geographicregion. They were lowest in Milan (1.5 per million) and highest in Budapest(5.5 per million). In Uppsala, Sweden - where dipyrone was not availableduring the study period - the annual rate was 5.1 per million. Incidence ratewas age-dependent and rose sharply with age. The overall case fatality ratewas 10% (2). The excess mortality ranged from 0.0013 per million users ofparacetamol and propyphenazone to 0.074 per million dipyrone users.
Aplastic anaemia refers to a diverse group of potentially severe bone- marrow disorders characterized by pancytopenia and a marrow largelydevoid of haematopoietic cells. Its clinical features, such as bleedingmanifestations, infections, fatigue and pallor, are the effects of too fewfunctional peripheral blood cells. Its annual incidence rate associated withhospital admission is 2.0 per million (2). Fatality rates increase with age; theoverall rate is 46% within two years (2). Marrow transplantation hasdramatically improved survival, however. The excess mortality from thedisease ranged from 0. 0032 per million aspirin users to 0.12 per millionusers of indomethacin.
Serious upper gastrointestinal complications The mechanisms by which NSAIDs cause gastrointestinal injury are those dependent on cyclooxygenase inhibition and those that exert directtoxic effects on the local mucosa. A major protective mechanism againstmucosal injury is the concentration of prostaglandin in the gastrointestinal mucosa; it is suppressed by inhibition of cyclooxygenase, the rate-limitingenzyme in the synthesis of prostaglandin. Aspirin inhibits it irreversibly, andmost other NSAIDs reversibly.
Overall excess mortality The overall excess mortality from anaphylaxis, agranulocytosis, aplastic anaemia and serious upper gastrointestinal complications was higher forNSAIDs than for other NNAs. The excess mortality associated withgastrointestinal complications was the principal influence on the overallestimate, contributing 99% of the excess mortality in users of all NSAIDs(aspirin, diclofenac, indomethacin and naproxen). The main influence onthe overall mortality estimate among users of NNAs was gastrointestinalcomplications of treatment with paracetamol (96%) and dipyrone (69%).
Propyphenazone contributed the least, showing no excess risk of gastro-intestinal complications.
Mortality per million users of NSAIDs was 2.0 for aspirin, 5.9 for diclofenac, 6.5 for naproxen and 11.7 for indomethacin. The rates for NNAsranged from 0.002 for propyphenazone to 0.25 and 0.2 for paracetamol anddipyrone users, respectively (Table 1 and Figure 1).
Table 1: Excess mortality per million short-term users, by adverse event Discussion on risk assessment The absolute risk of mortality associated with dipyrone appears to be substantially lower than that associated with NSAIDs for short-term reliefof pain. For dipyrone the estimated excess mortality due to agranulocytosis,aplastic anaemia, anaphylaxis and serious upper gastrointestinal complica-tions was 1.7 per million.
Figure 1. Overall mortality from short-term use of NNAs and NSAIDs (total)
Expected deaths per million users Prop Non-narcotic
The overall excess mortality estimates were influenced largely by the excess mortality associated with upper gastrointestinal complications, an adverseevent more commonly associated with NSAIDs, such as aspirin, diclofenac ornaproxen, than with dipyrone, paracetamol or propyphenazone.
A limitation of this evaluation was the use of relative risk estimates that did not account for duration of therapy, drug dose, or previous therapy withthe drugs under study. In this case it was assumed that the risk of adverseeffects was constant throughout therapy, which is unlikely to be the case.
Most anaphylactic reactions occur within one hour of exposure to a drug.
The risk of gastrointestinal haemorrhage is highest early in NSAID therapy(within one month of initiation) and decreases over time. Prophylactic useof aspirin has been shown to increase the risk of peptic ulcer bleeding (1, 12).
Patients who began treatment with aspirin one week before admission tohospital had an increased relative risk of 4.8, compared with a constant RRof 2.9 for aspirin intake begun between one week and three months beforeadmission. Another limitation is that the published studies contained noinformation on indication for drug use. However, no hypotheses suggestany increased frequency of the studied adverse events in the treatment offever, pain or inflammation.
Conclusions on risk assessment To compare the adverse drug-attributed public-health impact of non- narcotic analgesics, an epidemiological perspective was chosen and thedrug-related mortality quantified. Gastrointestinal complications ac-counted for most of the excess mortality in the overall estimate, for 99%of it among users of NSAIDs (aspirin, diclofenac, indomethacin andnaproxen). They accounted for most also among those treated withpropyphenazone and paracetamol. For dipyrone, gastrointestinal compli-cations accounted for 69%, and agranulocytosis for 30%, of the overallexcess mortality.
Benefit-Risk Assessment The benefits of dipyrone are equivalent or superior to those of other non- narcotic analgesics. The relative risk of dipyrone-associated agranulocytosisis substantial but the excess risk is very small. Dipyrone carries a lower riskof fatal complications from potentially fatal adverse events than aspirin orother NSAIDs.
New methods developed from a body of epidemiological studies showed that the risk of agranulocytosis had been overestimated in the 1970s and thatthe overall risk of dipyrone compared favourably with that of alternativeanalgesics. Although dipyrone was not protected by patent, Hoechst in 1994decided to submit Novalgin (dipyrone) for re-approval in Sweden.
In 1995, Sweden re-approved Novalgin 500 mg tablets and Novalgin solution for injection, 500 ml, for short-term treatment of acute moderate tosevere pain after tissue injury (e.g., from surgical procedures) and acutemoderate to severe colicky pain, in the urinary or biliary tract, for example.
Points Illustrated Drug safety is often evaluated by determining the frequency of occurrence of one particular drug-associated adverse event. This assumesthat all other adverse events have an identical profile and frequency. Whenthis is not the case safety can be assessed only by a three-dimensionalconstruct of the qualitative profile of adverse events, their frequency ofoccurrence, and a common measure of health outcome.
The purpose of the present evaluation was to estimate the adverse drug- attributed public-health effects of dipyrone and to compare them with thoseof other non-narcotic analgesics and non-steroidal anti-inflammatorydrugs. Ideally, such an evaluation would take account of all known drug-attributed adverse events, their duration and frequency of occurrence, andtheir varying degrees of seriousness. To be meaningful, a comparative safetyevaluation must employ an objective measure to incorporate and quantifythe impact of all adverse events. Such a common outcome measure coulddepend on the perspective chosen, however.
The relative benefit-risk profile was evaluated from an epidemiological perspective. Death was chosen as the common outcome of the differentadverse events, thereby restricting the evaluation to potentially life-threatening events. The drug-related mortality for dipyrone and othernon-narcotic analgesics, as a measure of the adverse drug-attributed public-health impact, was estimated and expressed by the excess mortality.
(1) van der Klauw MM, Stricker BHC, Herings RMC, Cost WS, Valkenburg HA: A population based case-cohort study of drug-induced anaphylaxis. Brit J ClinPharmacol 1993; 35: 400-408.
(2) Kaufman DW, Kelly JP, Levy M, Shapiro S: The drug aetiology of agranulocytosis and aplastic anaemia. Monographs in Epidemiology andBiostatistics. Vol.18, Oxford University Press, 1991.
(3) Laporte JR, Carne´ X, Vidal X, Moreno V, Juan J: Upper gastrointestinal bleeding in relation to previous use of analgesics and non-steroidal anti-inflammatory drugs. Lancet 1991, 337; 85-89.
(4) Langman MJS, Weil J, Wainwright P, et al.: Risks of bleeding peptic ulcer associated with individual non-steroid anti-inflammatory drugs. Lancet 1994;343: 1075-1078.
(5) Henry D, Dobson A, Turner C: Variability in the risk of major gastrointestinal complications from non-aspirin anti-inflammatory drugs. Gastroenter 1993;105: 1078-1088.
(6) Langman MJS, Coggon D, Spiegelhalter D: Analgesic intake and the risk of acute upper gastrointestinal bleeding. American Journal of Medicine 1983; 74:79-82.
(7) Faulkner G, Pritchard P, Somerville K, Langman MJS: Aspirin and bleeding peptic ulcers in the elderly. BMJ 1988; 297: 1311-1313.
(8) Needham CD, Kyle J, Jones PF, Johnston SJ, Kerridge DF: Aspirin and alcohol in gastrointestinal haemorrhage. Gut 1971; 12: 819-821.
(9) Armstrong CP, Blower AL: Non-steroidal anti-inflammatory drugs and life- threatening complications of peptic ulceration. Gut 1987; 28: 527-532.
(10) Grenland S: Quantitative methods in the review of epidemiologic literature.
Epidemiologic reviews 1987; 9: 1-30.
(11) Weil J, Colin-Jones D, Langman M, Lawson D, Logan R, Murphy M, Rawlins M, Vessey M, Wainwright P: Prophylactic aspirin and peptic ulcer bleeding.
BMJ 1995; 310: 827-830.
(12) Slattery J, Warlow CP, Shorrock CJ, MJS Langman: Risks of gastrointestinal bleeding during secondary prevention of vascular events with aspirin —Analysis of gastrointestinal bleeding during the UK-TIA Trial. Gut 1995; 37:509-511.
Temafloxacin, a fluoroquinolone used most widely in the United States, where it was first approved in January 1992, was also available in the UnitedKingdom, Germany and Sweden. It was no more efficacious than otherfluoroquinolones.
In the United States it appeared to be associated with a higher rate of spontaneous reporting than other fluoroquinolones. Reported eventsincluded disseminated intravascular coagulation (DIC), haemolytic anae-mia, thrombocytopenia, renal impairment and hypoglycaemia. Apart fromhypoglycaemia, which occurred predominantly in the elderly, it was notpossible to predict a high-risk group for the other reactions.
In Sweden, where temafloxacin was launched in September 1991, an estimated 30,000 patients had received it through May 1992. During thattime it accounted for 36 spontaneous reports of medically importantadverse reactions possibly or probably connected with it. None referred tohypoglycaemia but three were of anaphylactoid reactions; no drug-relateddeath was reported. Also, several cases of acute haemolysis/haemolyticanaemia were reported, and one 70-year-old male developed haemolyticuraemic syndrome with DIC. The overall adverse-reaction reporting rate(>1 case per 1000 patients) was unprecedented for an antibiotic in Sweden.
Temafloxacin and three other fluoroquinolones were the subjects of a comparative safety analysis undertaken of spontaneous reportsreceived in the United States during the first 120 days post-approval.
IMS data were used for denominators. The reporting rate for temofloxacin was 108/100,000 prescriptions, compared with 13, 25 and 20per 100,000 for the other three drugs. Intensive review of the comparatorfluoroquinolones revealed none of the serious reactions associated withtemofloxacin. The differences could not be attributed to reporting bias orpublicity; they were possibly related directly to the chemical structure oftemofloxacin.
The company together with the Food and Drug Administration agreed to withdraw temafloxacin in the United States in June 1992. It waswithdrawn worldwide at the same time.
Points illustrated A carefully controlled analysis showed that important information and consequences can be ascertained from spontaneous-report data that showespecially large differences between products. A transcript of the FDAAdvisory Committee proceedings is available. The case is also documentedin papers in the literature.* Remoxipride, (Roxiam), an antipsychosis agent of the benzamide group, was approved in Sweden in January 1991 for the treatment of schizophreniaand other types of psychosis in which the classical neuroleptics were foundto cause intolerable side-effects. Remoxipride was approved in mostEuropean countries shortly afterwards. In the clinical trials it was foundto be as effective as haloperidol, chlorpromazine and thioridazine. Its mainadvantages were that it caused significantly fewer extrapyramidal symptoms(EPS) than haloperidol and less sedation and anticholinergic adverse effectsthan chlorpromazine and thioridazine. In general, patients tolerated it welland it penetrated the market quickly.
In August 1993 the company informed regulatory agencies of four cases of aplastic anaemia and during the next four months four more cases wereidentified. Two of the eight patients died of the blood dyscrasia. Theestimated incidence of aplastic anaemia varied between countries from 0 in15,000 patients in Germany to 2 in 16,000 in Sweden and 5 in 10,000-12,000 * Norrby, S.R. and Leitman, P.S. Fluoroquinolone toxicities: update. Drugs, 45 (Supplement 3):59-64, 1993. Also, Finch, R.G. Withdrawal of temafloxacin: are there implications for other quinolones? DrugSafety, 8: 9-11, 1993.
in the United Kingdom. In all, 50,000 patients were estimated to havebeen exposed to the drug, yielding a rate of 1 fatal reaction in 25,000patients. The Swedish Medical Product Agency (MPA) evaluated therisk with aplastic anaemia as acceptable but the Committee forProprietary Medicinal Products (CPMP) gave a negative opinion andthe company withdrew the product from the world market in February1994. This document is in itself a summation and update of the benefit-risk evaluation done by the MPA (Sweden) in December 1993.
Schizophrenia: occurrence, and related morbidity and mortality Schizophrenia is a widespread disease, occurring all over the world and in all kinds of societies. Its annual incidence in the age-groups 15 years andover is around 1 per 1000 population. Life-time prevalence is about 1%, inmen and women alike. Around 2 million new cases are diagnosed annuallyworldwide. The peak onset occurs between 15 and 25 years of age in menand somewhat later in women. Onset is rarely before 10 or after 50 years ofage. Early onset is considered particularly ominous. Irrespective of theattitude adopted by society, whether these patients are treated as hospitalinpatients or mainly at outpatient clinics, schizophrenia is responsible for asubstantial part of the costs of health care.
Schizophrenic symptoms may be divided into positive (including thought disorder, loose associations, hallucinations, bizarre behaviour) and negative(affective blunting, poverty of speech, anhedonia and lack of motivation,social withdrawal). These two categories of symptoms differ with regard totheir amenability to medical treatment.
Patients with schizophrenia run a considerable risk of premature death, mainly due to suicide. About 50% attempt suicide and10% succeed. On ayearly basis three to five of every 1000 schizophrenic patients commitsuicide. Because of their poor rapport with both friends and therapists thesesuicides are often unpredictable. There is some evidence that patients whodevelop akathisia also may exhibit aggressive behaviour, including suicide.
During a phase of marked exacerbation of schizophrenic symptoms there is sometimes a rise in body temperature with elevated levels of plasma ureaand increasing muscular rigidity, and the patient may die of lethal catatonia.
These cases are difficult to distinguish from the neuroleptic malignantsyndrome, which has contributed to making such crises more common sincethe introduction of neuroleptics.
Whether patients with schizophrenia are at increased risk of somatic diseases (including pulmonary tuberculosis) remains uncertain. Increased general morbidity could probably be ascribed to environmental factors inthe old mental hospitals, but even recent linkage-studies have continued toshow an excess of both natural and unnatural deaths.
General treatment The introduction of neuroleptic therapy in the mid-1950s brought a major change in schizophrenic symptomatology as well as in the quality oflife of patients. Today, 25% seem to recover completely and another 25%partially, being able to support themselves at a lower level than before theybecame ill. Around 50% remain significantly impaired for the rest of theirlives.
Although neuroleptic treatment has made it possible to discharge patients, there are several reasons for non-compliance with this medicaltreatment: patients deny their symptoms, and they suffer from suchunpleasant side-effects as rigidity, akathisia, sedation, orthostatism andsexual disturbances. Neuroleptics have been particularly effective in themodification of positive symptoms, but they have less effect on negativesymptoms and sometimes even aggravate them.
Improvements claimed for new antipsychotic drugs have been linked mainly with allegedly improved effects on negative symptoms and tobeneficial effects in a therapy-resistant group of patients with schizophrenia.
Evidence has suggested also an improved safety profile with regard to theinduction of syndromes related to the extrapyramidal system.
Shortly after remoxipride (Roxiam) was approved in Sweden in January 1991, it was approved in several other European countries (Denmark,Germany, Luxembourg and the United Kingdom). Remoxipride differsfrom the classical neuroleptic drugs mainly in having a shorter half-life,normally 4-7 h, and less affinity for muscarinic, serotonergic, alpha-adrenergic and histaminergic receptors. Like most other neuroleptic drugs(and anti-depressants), remoxipride is metabolized by cytochrome P450CYP2D6, which exerts genetic polymorphism. About 7% of Caucasians areof the slow-metabolizer phenotype. The plasma half-life is about doubled insuch individuals, as it is also among elderly patients and in patients withseverely decreased renal function. The plasma concentration may also beincreased by concomitant treatment with neuroleptics that are metabolizedby the same enzyme and which have higher affinity to the enzyme thanremoxipride, while remoxipride will probably not increase the plasmaconcentrations of such substances.
Comparisons in clinical trials In the clinical trials remoxipride was compared mostly with haloperidol but also in a few studies with thioridazine and chlorpromazine. Nodifference was found in efficacy between remoxipride and haloperidol,thioridazine or chlorpromazine. In general, remoxipride had been welltolerated and shown to have a significantly lower incidence of sedativeeffects than haloperidol, and a two- to three-fold lower incidence ofextrapyramidal side-effects, including akathisia. The drop-out rate becauseof adverse effects (3.3%) was significantly lower than for haloperidol(14%), as was the need for anticholinergic treatment. Remoxipride wasalso significantly less sedative than chlorpromazine and thioridazine, andhad fewer autonomic side-effects such as hypotension, obstipation andblurred vision.
In the clinical trials programme, which covered about 2500 patients exposed to remoxipride, and in clinical practice, with an estimated 50,000exposed patients, the adverse-reaction profile of remoxipride was found tobe favourable until the occurrence of aplastic anaemia.
The risk with remoxipride In August 1993 the company informed the MPA of three cases of aplastic anaemia, and on 27 September 1993 reported a total of four cases in theUnited Kingdom. Until then there were no such reports in Sweden. TheMPA asked the old network of the International Aplastic Anaemia andAgranulocytosis Study to search the patient registers and all haematology-pathology registers; a ‘‘dear doctor'' letter was sent to all haematologists,asking for information on all cases of confirmed or suspected aplasticanaemia occurring between 1 January 1991 and 30 August 1993. Nursemonitors were sent to check all case records for (a) a psychiatric history, and(b) a history of neuroleptic drug intake. An experienced pathologistreviewed all suspected cases. By 18 October, all cases had been analysed. Innone of the 26 cases of confirmed aplastic anaemia (and in none of the 15suspected cases) was there any indication of exposure to remoxipride. Twopatients had been exposed to other neuroleptics.
However, two cases were reported to the Northern Regional Pharmacovigilance Centre during the survey: a woman hospitalized on1 September and a man on 16 September. Bone-marrow samples wereobtained quickly from both patients and a serum sample from the first.
During the actual period an estimated 16,000 patients had been treatedwith remoxipride.
From September through December four more cases of aplastic anaemia were identified: two in Sweden, one in Luxembourg and one more in theUnited Kingdom. These, in addition to the four reported in the UnitedKingdom in August and September (see above), made eight cases, of whomtwo died from the blood dyscrasia. The estimated incidence of aplasticanaemia varied between countries from 0 in 15,000 patients in Germany, 2in 16,000 patients in Sweden to 5 in 10,000-12,000 in the UK. In all, 50,000patients were estimated to have been exposed to the drug, yielding a rate ofone fatal reaction in 25,000 patients.
It was possible to compare remoxipride with haloperidol and the older neuroleptic drugs chlorpromazine, thioridazine and perphenazine.
The risk profiles The following table shows the profile of adverse reactions reported in Sweden in relation to sales.
The incidence is calculated by dividing the number of reports with a possible relation between drug and reaction by person-years underexposure, estimated by dividing by 365 the total annual sales figures, usinga defined daily dose (sales figures were available only up to 30 June 1993).
Comparative benefit-risk evaluation Schizophrenia is a relatively common and very serious chronic disease with substantial disease-related morbidity and mortality. On a lifetimebasis it has been reported that 50% of patients attempt suicide and thatabout 10% die of suicide. On a yearly basis 3-5/1000 patients commitsuicide. Such suicides often occur during psychotic phases and are difficultto predict. Therefore, control of the psychosis and prevention of relapse areof major importance to reduce disease-related mortality. Even moderntreatment is only partially successful. About 25% of patients are estimatedto recover completely and another 25% partially, while 50% remainsignificantly impaired for the rest of their lives. A most important aim inthe maintenance of treatment and avoidance of relapse is to secure thatpatients continue treatment as long as it is necessary. With the currenttrend in the management of mental illness to minimize long-termhospitalization, much of the success depends on how well the patientaccepts the treatment.
Reports of adverse reactions to neuroleptics in Sweden through 30 June 1993(per 10,000 patient-years exposure) Total personyears 6 (5 agranulocytosis) 5 (1 agranulocytosis, 1 pancytopenia) 7 (4 agranulocytosis, 3 sudden death) 7 (4 of 39 cases of agranulocytosis, 1 myocarditis, 1 sudden death, and 1 pulmonary oedema) 1 (aplastic anaemia, probably only thrombocytopenia peripherally) The antipsychosis drugs available in Sweden belong to several different chemical groups. Perphenazine and haloperidol dominate but thioridazineis also common. Remoxipride has been shown to be as effective ashaloperidol and thioridazine. Remoxipride is less sedative than both ofthese drugs and has significantly fewer extrapyramidal symptoms, includingakathisia. than haloperidol. It has fewer autonomic as well as cardiac side-effects than thioridazine. In general, remoxipride has high patientacceptability, which is important for the possibility of maintenance oftreatment.
From available data it seems clear that the different antipsychosis drugs have individual profiles of adverse effects but, for a thorough evaluation oftheir comparative risks, additional data were needed, especially on the riskof serious reactions from the older drugs. In Sweden, remoxipride was theonly available drug in the new group of substituted benzamides. Sulpiride,which is chemically close, was not approved because of problems in thetoxicological documentation, and risperidone had not yet been approved.
The benefits of clozapine are well-documented in refractory schizo- phrenia, as are the high risks of agranulocytosis, sedation and seizures, andthere is no doubt that it is a valuable drug in selected cases. With anincidence of agranulocytosis of 0.5% and a case fatality rate of 9%, the riskof fatal agranulocytosis from clozapine is of the order of 1 per 2000-3000patients.
At that time, the risk of aplastic anaemia from remoxipride seemed high, but the risk of fatal reactions did not seem higher than that from alternativedrugs or from pharmacotherapeutic strategies in general, for moderate toserious disease.
Indeed, if the drop-out rates from clinical trials are considered, more than four times as many patients taking haloperidol as those takingremoxipride would stop their treatment prematurely and then run anincreased risk of suicide.
Consequently, it did not seem warranted to withdraw remoxipride from the Swedish market; rather, its indications were restricted. For patients whodid not tolerate the side-effects of the other available antipsychosis drugs therisk of the disease was judged to outweigh that of aplastic anaemia fromremoxipride. A strict programme of blood monitoring was recommendedfor such patients, including measurements of haemoglobin, thrombocytesand white blood cells (including a differential count). Such a system hasworked well for patients treated with clozapine and it should be possible touse it also for patients on remoxipride. It was judged possible to limit the damage to the bone marrow by early withdrawal of the drug but it could notbe guaranteed that the aplastic anaemia would not progress.
Clozapine, a tricyclic dibenzodiazepine, was developed by Sandoz- Wander in the late 1950s for the treatment of schizophrenia. Between 1972and 1975 it was approved and marketed in many European countries and inSouth America. During this time isolated cases of granulocytopenia werereported to the company.
In February 1975 the drug was launched in Finland. Within six months 20 cases of suspected blood dyscrasias were reported among the estimated3000 treated patients. Nine died, eight of agranulocytosis and one of acuteleukaemia.
From multiple open studies covering more than 1000 patients, investigators tended to conclude that clozapine was as efficacious as, ormore efficacious than, standard neuroleptics such as perphenazine orchlorpromazine. Clozapine produced extrapyramidal effects less frequentlythan chlorpromazine.
Angst et al., in a controlled clinical trial with a randomized double-blind parallel group of 64 patients receiving either clozapine 140 mg/day (averagedaily dose) or methotrimeprazine 180 mg/day (average daily dose) over fourweeks, showed clozapine to be as efficacious as methotrimeprazine.
Ekblom and Haggstrom in a controlled clinical trial with a randomized double-blind parallel group of 41 schizophrenia patients receiving clozapine310 mg/day (average daily dose) or chlorpromazine 365 mg/day (averagedaily dose) for six weeks showed clozapine to be as efficacious aschlorpromazine.
Fischer-Cornelssen et al., in a random double-blind parallel-group, controlled multi-centre clinical trial of 217 schizophrenia patients receivingclozapine 300 mg/day (average daily dose) or chlorpromazine 360 mg(average daily dose) for six weeks, showed that clozapine had significantlyhigher efficacy than chlorpromazine.
Gerlach et al., in a single-blind cross-over trial of 20 schizophrenia patients each receiving clozapine 200 mg (average daily dose) for 12 weeksfollowed by haloperidol 10 mg (average daily dose) for 12 weeks, showedclozapine to be significantly better than haloperidol.
Singer and Law, in a randomized double-blind parallel-group trial of 40 acute patients with acute schizophrenia receiving clozapine 155 mg (averagedaily dose) or chlorpromazine 200 mg (average daily dose), showedclozapine to be similar to chlorpromazine.
Vencovsky et al., in a randomized double-blind parallel group trial with 52 schizophrenia patients receiving clozapine 300 mg (average daily dose) orchlorpromazine 600 mg (average daily dose), showed clozapine to besignificantly better than chlorpromazine.
Chiu et al., in a randomized double-blind parallel-group trial with 64 schizophrenia patients receiving clozapine 300 mg (average daily dose) orchlorpromazine 300 mg for six weeks, showed clozapine efficacy to beequivalent to that of chlorpromazine.
Fisher-Cornelsson and Ferner, in a randomized double-blind parallel- group trial with 723 schizophrenic patients receiving clozapine 300 mg oreither chlorpromazine 350 mg, haloperidol 8 mg, trifluoperazine 30 mg orclopenthixol 100 mg for six weeks found clozapine to be significantlysuperior to haloperidol and clopenthixol, equivalent to chlorpromazine andsignificantly inferior to trifluoperazine.
In all the trials extrapyramidal effects were rare with clozapine.
Claghorn et al. found that clozapine had significantly higher efficacy, but significantly fewer extrapyramidal side-effects, than chlorpromazine in amulti-centre randomized double-blind parallel-group trial with 124 schizo-phrenia patients with histories of tardive dyskinesia or extrapyramidalsymptoms on previous therapy. Patients received either clozapine 370 mg(average daily dose) or chlorpromazine 740 mg (average daily dose) overfour weeks.
Kane et al., in a randomized double-blind parallel-group trial with 265 schizophrenia patients refractory to previous therapy, found clozapine to besignificantly better, and to cause significantly fewer extrapyramidalsymptoms, than chlorpromazine. The test drug was given for six weeks.
Frequency of clozapine agranulocytosis estimated by Sandoz-Wander * In clinical trials (1962-1972) Four cases of agranulocytosis occurred in 2900 patients — an estimatedfrequency of 1.4 per 1000. Agranulocytosis was defined as the presence ofvery few or no neutrophil granulocytes in the peripheral blood(granulocytopenia <1600 neutrophils/mm3).
Spontaneous reports from 1972 to August 1976 Fed. Rep. Germany ** The absolute numbers of cases of agranulocytosis and granulocytopenia reported. Average amount of clozapine used per patient=24.2 g. This estimate was derived from sampling inpatient and outpatientprescribing practices. Amount of clozapine sold was derived from kilograms sold through May 31,1976. Patient estimate=amount of clozapine sold in kg divided by 24.2.
Agranulocytosis as an adverse reaction to clozapine was found to occur most frequently during the first 18 weeks of treatment, similar in this respectto chlorpromazine. The following table shows the history of reports from1954 to 1971 on various phenothiazines other than clozapine.
* From: Anderman and Griffith, Clozapine-Induced Agranulocytosis: A Situation Report up to August 1976, Europ.J.Clin.Pharmacol. 11, 199-201 (1977).
Frequency of cases of agranulocytosis related to various phenothiazines * Estimated frequency/ * From: Anderman and Griffith, Clozapine-Induced Agranulocytosis: A Situation Report up to August 1976, Europ.J.Clin.Pharmacol. 11: 199-201 (1977) Frequency of clozapine agranulocytosis estimated by the Finnish authorities A total of 20 cases of suspected blood dyscrasias had been reported among an estimated 3000 patients; nine died, eight of agranulocytosis and one ofacute leukaemia. Of the surviving cases, two had asymptomatic thrombocy-topenia and 9 had neutropenia/agranulocytosis. The overall estimatedincidence was 0.5% for agranulocytosis and 0.6% for neutropenia. Sixpatients had experienced neutropenia in connection with previous neuroleptictreatment. Only 25% of the cases occurred within 6 weeks; 75% occurredwithin 12 weeks and 25% after treatment for 12 weeks or more. Other thanprevious episodes of neutropenia, there were no special risk factors (J.
Ida¨npa¨a¨n-Heikkila¨, E. Alhava, M. Olkinuora and I.P. Palva. Agranulocy-tosis during treatment with clozapine. Lancet, 1975, No. II, p. 611).
Benefit-risk assessment By 1977 the efficacy of clozapine was judged to be about equal to conventional therapy. The frequency of extrapyramidal reactions was less thanfor conventional treatment, which was regarded as a clear benefit. The risk ofagranulocytosis was judged to be clearly higher than that of conventionaltherapy and the case fatality rate at that time was considered high. Thediscrepancy between the incidence estimates of agranulocytosis from clinicaltrials, 1.4 per 1000, and from the Finnish post-marketing experience, 5 per 1000, can be explained by the short duration of most trials (4-6 weeks). Only25% of the cases reported in the post-marketing situation occurred within thefirst six weeks. The benefit-risk balance was judged differently in differentcountries and the drug was withdrawn from most, though not all, countries.
Around 1985, clozapine was reintroduced in many countries since (a) it hadbeen proven in further clinical trials that it was efficacious in about 30% ofotherwise treatment-resistant patients, and (b) it was shown that regular bloodmonitoring could prevent the neutropenias from developing to agranulocy-tosis in most cases and virtually precluded a fatal outcome.
Action taken by the manufacturer In response to the Finnish cases: All cases were investigated for the possibility of special susceptibility ofthe Finnish population.
Information was sent to all regulatory authorities: – Information was sent to the medical community – Clozapine was withdrawn from the market in some countries (including France) Implementation of world-wide white blood cell monitoring mademandatory in each country where the drug continued to be marketed The ‘‘Ten Commandments for clozapine use'' were developed Further investigations of the pathogenesis of clozapine-induced granu-locytopenia were sponsored Promotion was stopped Clinical trials ceased temporarily Data gathered from the discontinued clinical trials were analysed Periodic updates were sent to regulatory authorities.
Points illustrated The clinical trials were too short to detect the true frequency of this adverse reaction.
The initial trials had not been able to demonstrate the unique benefits of the drug before the Finnish epidemic of agranulocytosis but continued useand trials showed this later.
Continuous blood monitoring made it possible to decrease the risk of this type of agranulocytosis and especially to reduce the fatality rate.
Sparfloxacin is an antibiotic of the quinolone family. It has been marketed in France (September 1994), Luxembourg (November 1994),United Kingdom (December 1994), Denmark (January 1995), Belgium andFinland (March 1995). Introduction into the market was pending for someother Member States of the European Union, although the authorizationhad been granted (Austria, Germany, Greece, Ireland, Italy, Netherlands,Portugal, Sweden, United Kingdom). Sparfloxacin has been marketed alsoin Japan (August 1993), Morocco (February 1995), and the Philippines,Peru and Uruguay (April 1995).
Sparfloxacin is available as 100 mg and 200 mg film-coated tablets. There is no harmonized Summary of Product Characteristics (SPC) within theEuropean Union or worldwide, particularly with regard to the approvedindications and the labelling of undesirable effects, precautions andwarnings.
In Japan, products containing sparfloxacin are indicated for the treatment of cutaneous, upper and lower respiratory, urogenital, hepato-biliary, digestive, ophthalmological and stomatological infections. In othercountries the indications are more restricted. As an example, in two memberstates of the European Union (France and Luxembourg) medicinalproducts containing sparfloxacin have been approved and marketed forthe treatment in adults of: presumed community-acquired acute bacterial pneumonia, whetherpneumococcal or non-pneumococcal exacerbation of chronic obstructive pulmonary disease (COPD) purulent acute sinusitis.
The medicine is given in a single-dose regimen of 400 mg as a loading dose on the first day, followed by 200 mg/ day as maintenance in a 10-daytreatment of infections of the lower respiratory tract and for about fourdays for sinusitis. The drug was used mainly for the treatment of sinusitis(*60%), bronchitis (COPD *20-22%) and pneumonia (*3-5%).
The issue prompting the risk-benefit evaluation After eight months of marketing sparfloxacin-containing medicinal products in France, many more reports of serious phototoxic reactions werebeing received about them than about other fluoroquinolones. Therefore, in May 1995 France issued a pharmacovigilance rapid alert related to 371 casereports received, of which 80% concerned cutaneous reactions.
In June 1995, Denmark referred its concern about the safety of the drug to the European Agency for the Evaluation of Medicinal Products (EMEA)and asked the Committee on Proprietary Medicinal Products (CPMP) foran opinion on the risks and benefits of products containing sparfloxacin, inaccordance with Article 12 of Directive 75/319/EEC.
Sparfloxacin is an antibiotic for the treatment of infections caused by Gram-positive pathogens. It has very few alternatives for respiratory-tractinfections caused by multi-resistant pneumococci. It has been the subject ofa series of prospective, randomized and double-blind comparative trials.
The antibacterial spectrum of sparfloxacin combines the usual activity of a fluoroquinolone with an activity against S. pneumoniae at least two to fourtimes higher than that of other antibiotics of its class. It is active also againstrespiratory infections caused by such pathogens as H. influenzae, Myco-plasma pneumoniae, and Chlamydia. A half-life of approximately 20 hoursallows oral administration as a single daily dose, after a double loadingdose.
Its efficacy has been demonstrated in patients with community-acquired pneumonia. It has been shown to be as efficacious as Augmentin anderythromycin in community-acquired pneumonia caused mainly byS. pneumoniae and H. influenzae.
Clinical equivalence has been demonstrated between sparfloxacin 200/ 100 mg (loading/maintenance dose) and Augmentin in acute exacerbationof chronic obstructive pulmonary disease (COPD).
For acute sinusitis, sparfloxacin at a dose of 400/200 mg four times a day for five days was as efficacious as cefuroxime axetil. No benefit was shownby increasing the daily dose.
Although it has not been claimed that urinary infections are indications for treatment with sparfloxacin, studies in some European countries haveshown 200 mg sparfloxacin in a single dose to be as efficacious asciprofloxacin for the treatment of acute gonococcal urethritis, and 200/100 mg for seven days to be as efficacious as seven-day treatment withdoxycycline for non-gonococcal urethritis. No data are available forevaluating the effectiveness of sparfloxacin in general practice.
In general, the spectrum of spontaneously reported adverse reactions to sparfloxacin has been similar to that for other fluoroquinolones.They includephototoxicity, allergic reactions, muscle and joint pain, tendonitis, rupturedtendon, gastrointestinal disorders, headache, sleep disorders, hallucinationsand dizziness. However, sparfloxin has been associated also with rareinstances of the cardiac rhythm disorder known as torsade de pointes, andwith a higher frequency of unexpectedly serious phototoxic reactions.
The increasing number of reports of serious cutaneous reactions became a matter of concern; 47% of the cutaneous reactions were phototoxic, suchas sunburn-like eruptions on unprotected parts of the skin. They occurredeven after treatment was discontinued. Of the reported cases, 25% hadsecond-degree burns; about 6% were hospitalized and they recovered onlyslowly and sometimes with sequelae. Some of the differences in frequency ofreporting of reactions to antibiotics of the quinolone family may beexplained partly by reporting bias (due, e.g., to publicity or to time on themarket). However, within this class the association of sparfloxacin withphototoxity is of major concern, because: it occurs sometimes with minimal exposure to ultra-violet light there is a high frequency of severe cases with serious outcomes recovery is slow and in some cases problems persist or recur long after thewithdrawal of the drug.
As phase I studies have shown, phototoxic reactions are dose-dependent.
The other particular concern is cardiotoxicity. However, cardiotoxicity is unlikely to have major clinical consequences except in the case of inpatientswith known risk factors such as underlying cardiac disease or treatment withdrugs that prolong the Q-T interval. Q-T prolongation is a rare dose-relatedreaction.
The pathophysiology of the phototoxicity has not been elicited.
Moreover, there is so far no evidence to clarify whether sparfloxacin hasphotomutagenic or photocarcinogenic properties.
Benefit-risk evaluation The overall benefit-risk evaluation takes into account the safety profile, the efficacy profile, and the use of the drug in everyday medical practice inthe general population. Thus the evaluation balances two factors: theassociation of the drug with a reaction that is not a common class-effect offluoroquinolones and with a higher frequency of expected adverse reactions than other fluoroquinolones; and the use of the drug for the indications forwhich physicians in France primarily prescribe it.
The analysis of prescribing data has shown that sparfloxacin wasprescribed in ‘‘all types of sinusitis'' without distinction, and thatabout 30% of cases for which it was prescribed had been wronglydiagnosed. Because of the efficacy of sparfloxacin againstpneumococci resistant to penicillin, the risk of inducing resistanceto it by the use of suboptimal doses, and the higher risk of inducingserious cutaneous adverse reactions, the risk-benefit balance wasregarded as favourable only for cases of acute sinusitis shown tobe, or suspected of being, due to pneumococcal strains resistant topenicillin or other antibiotics.
Chronic obstructive pulmonary disease (COPD) (*20-22%) The relevance of antibiotic therapy in this disease has been widelydebated. Antibiotics are only beneficial in COPD with a certaindegree of obstruction. However, in everyday medical practice thedegree of obstruction is often unknown, because of the absence ofrespiratory-function test results. Besides, the treatment is oftenpresumptive since the bacterial aetiology is generally undefined.
There is no standard therapy and numerous therapeutic alter-natives (including non-antibiotic treatment) have demonstratedefficacy in this disease. Also, acute exacerbation of infections inchronic bronchitis is fairly common and there is a risk of selectionof resistant bacterial strains if treatment is often given insuboptimal doses. Thus, the choice of antibiotic must take safetyinto account, and therefore the risk-benefit balance in respect ofsparfloxacin-containing medicinal products may be unfavourable.
Pneumonia (*3-5%) Although this indication was the most thoroughly evaluated in thepre-marketing clinical trials, it accounts for a limited prescriptionvolume, at least in France. These patients are generally bedriddenand are therefore the least exposed to UV radiation, comparedwith all patients receiving the treatment for the other indications.
However, the risk-benefit evaluation may be favourable only incases of radiologically-confirmed community-acquired acutepneumonia that has failed to respond to other antibiotic therapy.
Dosing considerations From the point of view strictly of efficacy, the proposed dose of 200/ 100 mg in the treatment of COPD could be supported, but this lower dosageof sparfloxacin might increase the risk of selection of resistance due to sub-inhibitory concentrations and repeated courses of treatment in this patientgroup. The higher dosage used in pneumonia should therefore also beconsidered for acute exacerbation of chronic bronchitis.
According to Directive 75/319/ECCC Art.13 (4), a European Article 12 procedure has three possible outcomes, which are binding on all MemberStates in the same way: withdrawal or suspension of the marketingauthorization or variation of the Summary of Product Characteristics(SmPC). This last entails restrictions of indications, supply and use, oramendments to defined sections of the SmPC, such as contraindications,undesirable effects, precautions or warnings.
For sparfloxacin-containing medicinal products, despite the important risks associated with the drug the possible outcome is driven by the medicalneed of it, given that strains of S. pneumoniae are developing resistance at anincreasing rate in some regions of the European Union.
However, the benefit-risk balance of sparfloxacin-containing medicinal products should be considered for each indication and dosing separatelywhen deciding on the options for action.
Based on the evidence available, the CPMP was of the opinion that the use of sparfloxacin-containing medicinal products should be restricted toinfections where no other treatment is available. It was decided that themedicinal product should only remain on the market if the conditions ofmarketing authorization were modified, further investigations were under-taken, and the SmPC was modified.
Sparfloxacin-containing medicinal products are indicated in the treat- ment of radiologically confirmed community acquired pneumonia whichhas failed to respond to conventional therapy and which either is caused bypneumococci highly resistant to penicillin (MIC 4 72 mg/l) and other antimicrobials, or occurs in an epidemiological environment indicating ahigh risk of such multiresistant strains.
The other indications were disallowed on grounds of the unfavourable The selected dose regimen for the restricted indication is 400 mg as a single dose on the first day, followed by a 200 mg single dose daily for 10days on average, and a maximum of 14 days.
The CPMP was of the opinion that the marketing authorization of this formulation should be withdrawn, as this strength did not correspond todoses needed for the recommended indication. In addition, it increased therisk of misprescribing or non-compliance with the recommended pharma-ceutical form for the recommended indication.
Further investigation The CPMP was of the opinion that studies should be undertaken to investigate the photomutagenicity and photocarcinogenicity of sparflox-acin. It recommended also six-monthly updates for two years on safety andefficacy.
Additional remarks The European Commission accepted the recommendations of the CPMP and announced its decisions on the variation of the marketing authorizationof 200 mg sparfloxacin tablets and on the withdrawal of the 100 mgsparfloxacin tablets on 7 May 1996 and 14 June 1996, respectively, in theOfficial Journal of the European Union (No. C188/5).
MODEL FOR QUANTIFICATION OF RISKS* The objective of this model is to make comparison easier between products from the same therapeutic class but with possibly very differentadverse reactions. It is based on the usual international criteria ofseriousness: fatal outcome, hospitalization, incapacity, sequelae (life-threatening reactions are included in the first criterion since, if they arelife-threatening, they must sometimes be fatal).
Comprehensive information on the safety of a drug, including all adverse reactions and their outcomes (number of fatalities, number and duration ofhospitalizations or incapacity, number and severity of sequelae, etc.), wouldrequire a cohort of many thousands of patients, which is rarely available. Inpractice, when adverse reactions occur and are diagnosed, their incidence isestimated on the basis of spontaneous reporting, with its numerous biases,or of epidemiological studies.
To permit comparison between drugs with different adverse reactions, a classification of seriousness and incidence into a few categories is proposed,on the basis of experience with the drug or of medical knowledge. Thechosen ranges or boundaries are arbitrary, but considered reasonable.
Fatal (F) or life-threatening (LF) outcome: three categories of frequencyF (LF) 4 7F (LF) < 10% of cases F (LF) 51% of cases Hospitalization (H): two categories, according to durationH 5 Transient incapacity (I): two categories, according to durationI 5 Permanent sequelae (disability): evaluated according to standard scales5 * Proposed by Christian Be´nichou, MD (Synthe´labo, Paris, France) According to these data, a classification of seriousness into six categories is proposed, as follows:Category 6: Fatal outcome in more than 10% of cases (e.g., TEN) Fatal outcome between 1 and 10% (e.g., hepatocellularinjury, agranulocytosis, anaphylactic shock) Fatal outcome 51%, permanent sequelae 4 Hospitalization 42 days or incapacity 47 days Hospitalization 5 72 days or incapacity 5 All other outcomes (of less intensity/severity) The possible classifications can be shown as a grid.
Classification of incidenceCategory 5 (very common): 410 % (>10/100) Category 4 (common or frequent): between 10% and 1%(10/100 to 1/100) Category 3 (uncommon or infrequent): between 1% and 0.1% (1/100 to 1/1000) Category 2 (rare): between 0.1% and 0.01%(1/1000 to 1/10,000) Category 1 (very rare): 50.01% (51/10,000) Global quantification of risks Combining seriousness scores that can range from 6 to 1 and incidence scores from 5 to 1, a reaction score from 1 (1 x 1) to 30 (6 x 5) would resultfor each reaction; then an addition of reaction scores for all identifiedreactions could give a global drug risk score.
Clearly, such a scheme requires testing and validation, with special attention to the multifactorial aspects of the categorization scale forseriousness.
DETAILED EXAMPLE OF A QUANTITATIVE BENEFIT-RISK EVALUATION: DIPYRONE* Dipyrone belongs to the class of non-narcotic analgesics that have no anti-inflammatory action. It is effective for the treatment of fever and painand is used in more than 100 countries, including Germany, France, Spain,Italy and Finland.
Dipyrone was introduced into clinical therapy in 1922 and registered in Sweden in 1934 for the treatment of various states of acute pain. Because ofan apparently unacceptable risk of agranulocytosis, it was withdrawn fromthe Swedish market in 1974 and the US market in 1977. At that time thenumber of spontaneous reports suggested a causal relationship betweendipyrone and agranulocytosis, though no reliable estimate of the overallincidence rate of agranulocytosis was available.
To obtain population-based evidence of the adverse public health impact of dipyrone, the International Agranulocytosis and Aplastic AnaemiaStudy (IAAAS), the largest ever blood-dyscrasia study, was carried outbetween 1980 and 1986. It provided better estimates than any previouslyavailable from several countries of the overall incidence rate of agranulo-cytosis and its associated mortality.
The Study found that the overall incidence of agranulocytosis was extremely low and that the risk associated with dipyrone had beenoverestimated in the 1970s. Hoechst, therefore, in 1994, resubmitted anapplication for marketing approval of dipyrone in Sweden.
This appendix describes the evaluation used to estimate the adverse drug- attributed public-health effects of dipyrone and to compare them with thoseof other non-narcotic analgesics, including non-steroidal anti-inflammatorydrugs (NSAIDs).
In addition to its antipyretic activity, dipyrone is an effective oral and parenteral treatment for moderate to severe pain, including post-operative * Courtesy of Carlos Martinez and Ernst Weidmann (Hoechst Marion Roussel), and Stephen Rietbrock (University of Cologne).
pain. Oral dipyrone is superior to paracetamol and comparable to aspirin.
In its parenteral form it compares in analgesic efficacy to the NSAIDs andto weak opioids.
III. RISK EVALUATION A. Materials and Methods of the Benefit-Risk Evaluation The benefit-risk evaluation was performed for competing treatment strategies for the indication of acute mild and moderate pain. The followinganalgesics and doses were selected: paracetamol 5 71500 mg/d, dipyrone 71500 mg/d, aspirin <1000 mg/d, ibuprofen 5 71500 mg/d and diclofenac 7100 mg/d. Acute pain was defined as requiring analgesic treatment for less than one week. Two age-groups were analysed: below 60 years, and 60and above.
All potentially fatal adverse events reported with non-opioid analgesics were identified: anaphylaxis, Stevens-Johnson syndrome (SJS), toxicepidermal necrolysis (TEN), agranulocytosis, aplastic anaemia, uppergastrointestinal complications (haemorrhage or perforation), toxic liverdisease, and end-stage renal failure. The Medline database was searched forall English-language reports of epidemiological studies published betweenJanuary 1970 and February 1997 on the association of those adverse eventswith the use of non-narcotic analgesics.
All studies with a case-control or a cohort-study design that provided data to assess the excess mortality in the population studied were selected.
The required data from the case-control studies were expected to provideinformation specific to the drugs and to the adverse events.
Drug-specific data included estimates of the relative risk of an adverse event associated with drug exposure compared with those associated withnon-exposure (for gastrointestinal complications, dose-specific relative riskestimates were necessary), and the percentage of cases exposed to the drug.
Information specific to adverse events comprised the incidence rate of thedisease in the source population and the case-fatality rates of the adverseevents. Age-specific estimates were essential for inclusion in the age sub-group analyses.
Drug-specific multivariate relative risk estimates (RR) from each study were combined, assuming a random effect of all estimates, and calculatedaccording to the method of DerSimonian and Laird (1). Aetiological fractions (EF), non-aetiological fractions (NEF) and the excess incidencerates (EIR) were calculated from the following equations: = RR–1 Pexposed NEF = { 1 – RR–1 P EIR = IR { 1 – RR–1 P where the incidence rate (IR) is the arithmetic mean of adverse event-specificincidence rates, and Pexposed the arithmetic mean of the prevalence ofexposure among cases from different studies. The last equation is derivedfrom the common expression: Excess Incidence = Incidence among exposed(IRxNEFxRR) – Incidence among non-exposed (IRxNEF)(3).
To address the problem of assessing and comparing different types of adverse events, a unique health-outcome measure was required whichattached a weight (wj) to each adverse event to facilitate the comparison.
Therefore, case fatality rates specific to adverse events were selected as theunique health-outcome measure. The case fatality rate is the proportion offatal cases of those who develop an illness. The arithmetic mean was used toestimate adverse-event-specific weights from all suitable studies.
The global excess risk (GER) due to different adverse events with weights wj (common health outcome) was defined as the sum of drug-attributedweighted excess incidence rates of all considered adverse events.
Because of different distributions, i.e., lognormal and normal distribu- tion functions, confidence intervals could not be calculated with standardmethods. Therefore, 95% confidence intervals for the drug-attributedexcess mortality and for the global excess risks were determined by MonteCarlo simulation (2). Arithmetic means of simulated values were defined aspoint estimates.
A total of 16 epidemiological studies provided information to calculate the global excess risk attributed to non-opioid analgesics: two on anaphylaxis (3, 4), one on agranulocytosis (5), one on aplastic anaemia (5),three on Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis(TEN) (6-8), and 10 on serious gastrointestinal complications, includingupper GI haemorrhage and perforation (9-18). Two studies whichaddressed liver toxicity were identified, but they did not fulfil the inclusioncriteria (19, 20). Excess mortality rates were calculated for dipyrone,paracetamol and three NSAIDs — aspirin, diclofenac and ibuprofen.
Of the 10 studies on gastrointestinal complications, five provided information on the overall incidence rate (9-13) and five on the case fatalityrate (9, 13, 14, 17, 18); three provided relative risk estimates for themedicines (14-16).
Anaphylaxis is the maximum variant of an immediate type reaction that occurs in previously sensitized persons after re-exposure to the sensitizingantigen. A clinically indistinguishable syndrome which is not antibody-mediated and does not require previous exposure to the antigen is called ananaphylactoid reaction. The anaphylactic or anaphylactoid responseappears usually within minutes of administration of the specific antigen.
It is characterized by cutaneous, gastrointestinal, respiratory, cardiovas-cular or central-nervous symptoms that can occur either singly or incombination. Life-threatening conditions involve respiratory obstructionleading to respiratory failure or cardiovascular collapse or shock.
Only one study provided the necessary data for the calculation of the excess mortality incidence rate, indicating an annual incidence of drug-induced anaphylaxis of 0.18 per million per week (252 cases in a two-yearperiod in a population of 14.5 million) (3). In the category called ‘‘otheranalgesics'' the authors included salicylic-acid derivatives, pyrazolones anddipyrone. Therefore, the relative risk of these agents was assumed to beidentical. One study estimated a case fatality rate of 1.6% (4). Because ofmissing data age-specific excess mortality rates could not be determined.
The weekly excess mortality from anaphylaxis ranged from 0.007 permillion for dipyrone and aspirin users to 0.016 per million for diclofenacusers.
Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS/TEN) SJS and TEN are severe skin reactions characterized by erythema and blisters or erosions of the skin or mucous membranes. In SJF, blisters coverless than 10% of the body surface, and in TEN more than 30%. Cases with 10 to 30% of skin detachment are classified as ‘‘SJS/TEN overlap''(6). 17.1percent of all SJS/TEN cases are fatal: 4% of SJS, 25% of SJS/TENoverlap, and 36% of TEN (7). Death results from sepsis and renalfailure (7).
The first results of an international case-control study of serious cutaneous adverse reactions (SCAR) performed in France, Germany, Italy,and Portugal have been published (6). This is the only study allowing thecalculation of SJS/TEN-specific excess mortality rates for non-narcoticanalgesics, including NSAIDs. The annual incidence rate of SJS/TEN in thecommunity is estimated to be 1.82 per million (8). Because of missing data,age-specific excess mortality rates could not be determined.
Agranulocytosis, defined as a granulocyte reduction below 0.5x109/L, is associated with clinical symptoms of acute bacterial infections. Its initialmanifestations are due to the severity, site (e.g., lungs, oropharynx, urinarytract or blood) and duration of the infection. Typical symptoms are feverand sore throat. Two mechanisms of drug-induced agranulocytosis havebeen proposed: immune and toxic. The immunological mechanism iscorrelated with rapid peripheral destruction of granulocytes in a previouslysensitized patient. Drug-dependent antibodies adhere to circulatinggranulocytes, causing their abrupt destruction. The toxic type of the diseasealso leads to sudden destruction of granulocytes. However, this may occurafter a latent period during which the patient receives a substantial amountof the causative drug and granulocyte precursors and mature peripheralgranulocytes are damaged. There is considerable evidence that dipyrone-induced agranulocytosis is immune-mediated.
The International Agranulocytosis and Aplastic Anaemia Study (IAAAS) carried out a population-based case-control study in Germany,Italy, Hungary, Spain, Israel, Sweden and Bulgaria; it gave an overallannual incidence rate of community-acquired agranulocytosis of 3.4 permillion (5). Estimates of annual incidence rates differed by geographicregion. They were lowest in Milan (1.5 per million) and highest in Budapest(5.5 per million). In Uppsala, Sweden - where dipyrone was not availableduring the study period - the annual rate was 5.1 per million. Incidence ratewas age-dependent and rose sharply with age. The overall case fatality ratewas 10% (5). The overall weekly excess mortality from agranulocytosisranged from 0.019 per million for aspirin users to 0.11 per million fordipyrone users. For dipyrone users below 60 years the overall weekly mortality rate was 0.054 per million, and for those over 60 years 0.32 permillion. See Tables 1, 2 and 3.
Aplastic anaemia refers to a diverse group of potentially severe bone- marrow disorders characterized by pancytopenia and a marrow largelydevoid of haematopoietic cells. Its clinical features, such as bleedingmanifestations, infections, fatigue and pallor, are the effects of too fewfunctional peripheral blood cells. Its annual incidence rate associated withhospital admission is 2.0 per million (5). The fatality rate increases with age;the overall rate is 46% within two years (5). Marrow transplantation hasdramatically improved survival, however. The overall weekly excessmortality from aplastic anaemia was 0.063 per million for diclofenac users:0.04 and 0.18 for under and over 60 years of age, respectively. The rate forthe other drugs was not significantly increased above background.
Serious upper gastrointestinal complications The mechanisms by which NSAIDs cause gastrointestinal injury are those dependent on cyclooxygenase (COX-1) inhibition and those that exertdirect toxic effects on the local mucosa. A major protective mechanismagainst mucosal injury is the concentration of prostaglandin in thegastrointestinal mucosa; it is suppressed by inhibition of cyclooxygenase,the rate-limiting enzyme in the synthesis of prostaglandin. Aspirin inhibits itirreversibly, and most other NSAIDs reversibly.
The reported weekly estimates of incidence rate ranged from 3 to 40 per million. The case fatality rates associated with gastrointestinal complica-tions were 0.023 to 0.098 per million per week and were highest in older agegroups (9, 13, 14, 17, 18). In subjects under 60 the weekly rate was 4.6 permillion, with a 2.6 per cent case fatality rate. For patients over 60 it wassubstantially higher (33.8 per million; 9.5 per cent case fatality rate). Therate for the over-60 groups ranged from 2.5 per million per week foribuprofen users to 9.5 for aspirin users.
Global excess mortality risk The global excess mortality risk per million short-term drug users per week for all ages was 1.7 for aspirin 5 71000 mg per day, 1.5 for diclofenac 7100 mg per day, 0.4 for ibuprofen 5 71500 mg per day, 0.05 for 71500 mg per day and 0.11 for dipyrone 5 (see Table 1 and Figure 1). The excess mortality associated with gastro- intestinal complications had the most important influence on the overallglobal excess mortality in users of all NSAIDs (aspirin, diclofenac andibuprofen). Gastrointestinal complications did not contribute to the globalexcess mortality attributed to paracetamol and dipyrone.
The global excess risk associated with NSAID use in the age group over 60 was an order of magnitude greater than in the under 60 group(Figures 1, 2). Although there was an age-dependent increase in the globalexcess risk associated with dipyrone use of about fivefold, this excess riskwas significantly smaller than the respective risk estimates for aspirin anddiclofenac. Paracetamol was significantly associated with SJS/TEN but age-specific estimates were not available.
The global risk of mortality associated with dipyrone appears to be substantially lower than that associated with equivalent doses of NSAIDsfor short-term relief of mild and moderate pain, particularly in the older agegroup. The estimated weekly excess mortality from agranulocytosis, aplasticanaemia, anaphylaxis and serious upper gastrointestinal complications permillion short-term users was 1.7 per million for aspirin, 1.5 per million fordiclofenac, 0.4 per million for ibuprofen, 0.05 per million for paracetamoland 0.11 per million for dipyrone.
The overall excess mortality after short-term NSAID use was largely influenced by the excess mortality associated with upper gastrointestinalcomplications. In the selected doses, the investigated NSAIDs haveantipyretic and anti-inflammatory as well as analgesic properties. As theidentified epidemiological studies do not provide information on theindication for drug use, it is reasonable to expect that the indication forNSAID use also included inflammatory diseases. The current benefit-riskanalysis can thus be considered valid because the risk of potentially life-threatening events is normally not associated with the indication for druguse.
A limitation of this evaluation was the use of relative estimates that did not account for the duration of therapy or for previous therapy with theanalgesics under study. It is known, e.g., that the risk of gastrointestinalhaemorrhage is higher early during nonsteroidal anti-inflammatory druguse (within one month after initiation), decreasing with duration oftreatment. However, for purposes of this evaluation it was assumed thatthe risk of adverse effects was constant throughout therapy.
Incidence rates and the case-fatality rates of gastrointestinal bleeding and agranulocytosis show a strong age dependency. There were no age-specificdata available for the other included adverse events. The global excess riskassociated with NSAID use was substantially higher in the age group over60 years than at younger ages.
In order to compare the adverse drug-attributed public health impact of non-narcotic analgesics, an epidemiological perspective was chosen and thedrug-related global mortality quantified. The excess mortality associatedwith gastrointestinal complications exerted the major influence on theoverall estimate, contributing to 99% of the excess mortality in NSAIDusers (aspirin, diclofenac and ibuprofen). In users of paracetamol, the majorinfluence on the global mortality estimate was Stevens-Johnson syndromeand toxic epidermal necrolysis, whereas global mortality attributed todipyrone was due mainly to agranulocytosis. There was a strong associationof the global excess risk with age in NSAID users.
IV. BENEFIT-RISK ASSESSMENT The benefits of dipyrone are equivalent or superior to those of other non- narcotic analgesics. The relative risk of dipyrone-associated agranulocytosisis substantial but the excess risk is very low. Dipyrone carries a lower risk offatal complications from potentially fatal adverse events than aspirin orother NSAIDs and the difference is even more pronounced in the over-60age group.
V. SUMMARY OF ACTION TAKEN A new methodological approach based on a body of epidemiological studies has shown that the global risk of dipyrone compares favourably withthat of competing non-narcotic analgesics and that the risk of agranulo-cytosis had been overestimated in the evaluation during the 1970s. Althoughdipyrone no longer has patent protection, Hoechst decided to apply forreapproval in Sweden (where dipyrone is sold as Novalgin).
In 1995, Novalgin 50 mg tablets and Novalgin solution for injection 500 mg/ml were reapproved in Sweden for short-term treatment in acute,moderate to severe pain after tissue injury (e.g., as a result of surgicalprocedures) and acute, moderate to severe colicky pain (e.g., in the urinaryor biliary tract).
VI. POINTS ILLUSTRATED Drug safety is often evaluated by determining the frequency of occurrence of one particular drug-associated adverse event. This assumesthat all other adverse events have an identical profile and frequency. Whenthis is not the case safety can be assessed only by a three-dimensionalconstruct of the qualitative profile of adverse events, their frequency ofoccurrence, and a common measure of health outcome.
The relative benefit-risk was evaluated from an epidemiological perspective. Death was chosen as the common outcome of the differentadverse events, thereby restricting the evaluation to potentially life-threatening events. The drug-related mortality for dipyrone and othernon-narcotic analgesics, as a measure of the adverse drug-attributed public-health impact, was estimated and expressed by the excess mortality.
(1) DerSimonian N, Laird N: Meta-analysis in clinical trials. Controlled Clinical Trials 1987; 7: 177-188.
(2) Buckland ST: Monte Carlo confidence intervals. Biometrics 1984; 40: 811-817.
(3) van der Klauw MM, Stricker BHC, Herings RMC, Cost WS, Valkenburg HA, Wilson HP: A population based case-cohort study of drug-induced anaphy-laxis. Brit J Clin Pharmacol 1993; 35: 400-408 (4) International Collaborative Study of Severe Anaphylaxis. Members of study group: Laporte JR, Latorre FJ, Gadgill DA, Chandrasekhar DV, Laszlo A,Retsagi G, Alfredson L, Martinez C, Kaufman DW, Anderson T, Kelly J,Shapiro S: An epidemiologic study of severe anaphylaxis among hospitalpatients: methods and incidence. To be published in Epidemiology 1998.
(5) Kaufman DW, Kelly JP, Levy M, Shapiro S: The drug etiology of agranulocytosis and aplastic anemia. Monographs in Epidemiology andBiostatistics. Volume 18, Oxford University Press 1991.
(6) Roujeau JC, Kelly JP, Naldi L, Rzany B, Stern RS, Anderson T, Auquier A, Bastuji-Garin S, Correia O, Locati F, Mockenhaupt M, Paoletti C, Shapiro S,Sheir N, Scho¨pf E, Kaufman D: Drug etiology of Stevens-Johnson syndromeand toxic epidermal necrolysis, first results from an international case-controlstudy. N Engl J Med 1995; 333: 1600-1609.
(7) Mockenhaupt M, Schlingmann J, Schroeder W, Schoepf E: Evaluation of non- steroidal antiinflammatory drugs (NSAIDs) and muscle relaxants as riskfactors for Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis(TEN). Pharmacoepidemiology and Drug Safety 1996; 5:S116.
(8) Scho¨pf E, Rzany B, Mockenhaupt M: Schwere arzneimittelinduzierte Hautreaktionen: Pemphigus vulgaris, bullo¨ses Pemphigoid, generalized bul-lous fixed drug eruption, Erythema exsudativum multiforme majus, Stevens- Johnson-Syndrom und toxisch-epidermale Nekrolyse. Fortschritte der prak-tischen Dermatologie und Venerologie 1994; 89 — 95 (9) Laporte JR, Carne´ X, Vidal X, Moreno V, Juan J: Upper gastrointestinal bleeding in relation to previous use of analgesics and non-steroidal anti-inflammatory drugs. Lancet 1991, 337: 85-89.
(10) Henry D, Robertson J: Nonsteroidal anti-inflammatory drugs and peptic ulcer hospitalization rates in New South Wales. Gastroenterology 1993; 104: 1083-1091.
(11) Carson JL, Strom BL, Soper KA, West SL, Morse ML: The association of nonsteroidal anti-inflammatory drugs with upper gastrointestinal tractbleeding. Arch Intern Med 1987; 147: 85-88.
(12) Beard K, Walker AM, Perera DR, Jick H: Nonsteroidal anti-inflammatory drugs and hospitalizatiom in gastroesophagal bleeding in the elderly. ArchIntern Med 1987, 147: 1621 -1623 (13) Rockall TA, Logan RFA, HB Devlin, Northfield TC on behalf of the steering committee and the members of the national audit of acute upper gastro-intestinal haemorrhage. Incidence of and mortality from acute uppergastrointestinal adverse event haemorrhage in the United Kingdom. BMJ1995; 311: 222-230.
(14) Perez Gutthann S, Garcia Rodriguez LA, Raiford DS: Individual nonsteroidal antiinflammatory drugs and other risk factors for upper gastrointestinalbleeding and perforation. Epidemiology 1997; 8: 20-24.
(15) Kaufman DW, Kelly JP, Sheehan JE, et al.: Nonsteroidal anti-inflammatory drug use in relation to major upper gastrointestinal bleeding. Clin PharmacolTher 1993; 53(4): 485-494.
(16) Garcia Rodriguez LA, Jick H: Risk of upper gastrointestinal bleeding and perforation associated with individual non-steroidal anti-inflammatory drugs.
Lancet 1994; 343: 769-772.
(17) Katschinski BD, Logan RFA, Davies J, Langman MJS: Audit of mortality in upper gastrointestinal bleeding. Postgrad Med J 1989; 65: 913-917.
(18) Wilcox CM, Clark WS: Association of nonsteroidal antiinflammatory drugs with outcome in upper and lower gastrointestinal bleeding. Dig Dis Sci 1997;42: 985-989 (19) Carson JL, Strom BL, Duff A, Gupta A, Das K: Safety of nonsteroidal anti- inflammatory drugs with respect to acute liver disease. Arch Intern Med 1993;153: 1331-1336.
(20) Garcia Rodriguez LA, Perez Gutthann S, Walker AM, Lueck L: The role of non-steroidal anti-inflammatory drugs in acute liver injury. BMJ 1992; 305:865-868.
Figure 1. Age above 60 : global excess risk of non-narcotic analgesics
tality/million/w 8.0 — xcess mor 4.0 — global e 0.0 —paracetamol dipyrone Figure 2. Age below 60 : global excess risk of non-narcotic analgesics
tality/million/w 8.0 — global e 0.0 —paracetamol THE ‘‘TURBO''MODEL FOR BENEFIT-RISK ANALYSIS The benefit-risk balance spectrum is portrayed in Figure 1 of Chapter II D. The challenge is to quantify appropriately the benefits and risks so thatthe drugs under comparison can be represented in their rightful positions onthe graph. As already described, both risks and benefits have two basicdeterminants, degree and probability, which are quantifiable. In its simplestform: R-factor = Ro+ Rc B-factor = Bo + Bc where Ro is the risk associated with the medically most serious adverse effect, Rc represents an additional risk (e.g., the next most serious adversereaction or the most frequent), Bo = primary benefit, and Bc = ancillarybenefits(s). As formulated by Amery, scores for Bo and Ro range from 1 to 5,and for Rc and Bc from 0 to 2. Figure 1 represents an R-score grid withpossible scores; measurements should reflect the risk in its most severeappearance (e.g., Torsade de pointe, not QTc prolongation; hepatitis, nottransaminase increase), as determined from the best available data (clinicaltrials, epidemiological data, etc.). Figure 2 presents suggested scores andassociated definitions for risk severity. Figure 3 illustrates how to applythem to yield a value for Rc. A similar approach can be used to determinethe B-factor (see Figures 4-6). Placement of the resultant B-factor and R-factor on the Turbo diagram (Figure 7) provides a composite for between-drug comparisons.
Courtesy of Dr Willem Amery (Jannsen research Foundation, Turnhoutsweg 30, B-2340 Beerse,Belgium) Figure 1. «R» score associated with the more severe adverse effect (= R)
severe very severe Estimated Severity Figure 2. Estimating severity of risk severity = impact on health status and socioprofessional capabilities E.g., five scores (definitions are tentative): 1 = some hindrance, but not really incapacitating 2 = temporarily/intermittently incapacitating 3 = incapacitating, but not life-threatening/-shortening 4 = life-shortening, but not life-threatening 5 = life-threatening Score should refer to risk if properly managed. For example: preventability through monitoring (bleeding due to anticoagulant) (full) recovery if appropriately managed (hepatotoxicity in most instances) timely detection (presence of warning signs) Figure 3. The adjusted ‘‘R''score = the ‘‘R''-factor Take the next severe adverse effect or, if there is no other severe adverseeffect, the most frequent one and estimate ‘‘R''score for this adverseeffect = R' ‘‘R''-factor = Ro + correction factor Rc Correction factor Rc 7 3 (tentative example) Figure 4. «B» score associated with the benefit in that indication (=B )
Probability of Benefit uncommon Degree of Benefit Figure 5. Estimating degree of benefit (Bo) Benefit = impact on indication as reflected by change(s) in health status and socioprofessional capabilities E.g., five scores; treated condition becomes (definitions are tentative): 1 = less hindering, but capabilities remain unchanged2 = less frequently incapacitating or incapability lasts shorter3 = less incapacitating, but no change in life expectancy4 = less life-shortening5 = less immediately life-threatening Score refers to benefit associated with correctly used medicine (and leaves out of consideration aspects such as non-compliance).
Figure 6. The adjusted ‘‘B'' score = the ‘‘B''-factor Consider whether the medicine has relevant ancillary properties and assign a value to the correction factor as indicated below: ‘‘B''-factor = Bo + correction factor (Bc) for ancillary property Correction factor Bc (tentative example) = + 2 if ancillary medical property relevant to the indication (e.g., cholesterol lowering effect forantidiabetic or for antihypertensive medicine) = + 1 if ancillary practical property (e.g., once-daily dosage schedule orfast onset of action, etc.) Figure 7. The intrinsic RB balance: the TURBO diagram
T-scores (in grid) to be further defined A SURVEY OF MANUFACTURER AND REGULATORY AUTHORITY DECISION-MAKING PRACTICES FOR POST-MARKETING SAFETY ISSUES To understand some of the current decision-making practices of companies and regulatory bodies, separate survey questionnaires for eachgroup were designed and distributed to members of the Working Groupduring the second half of 1995. One company not represented in the Groupalso completed the survey. The responding parties were as follows: Manufacturers (11): (Note: The survey was completed before the Glaxo-Wellcome merger] The questions and tally of replies are presented below.
Results of survey of manufacturers 1. For major post-marketing safety issues, does your company retain at least one standing committee/established board of outside experts? If yes, describe its membership, function and frequency of meetings.
Quarterly – 2No reply – 8 2. If there is no ‘‘board'' do you routinely use individual outside/ independent experts for major safety issues? Note: Replies to most of the items of question 3 are summarized in the attached two tables (tables 1 and 2).
3. Please consider a recent significant safety issue/crisis/possible-with- drawal in your company.
With which drug did it occur? What was the issue? iii) Were outside experts involved iv) If yes, how were they involved?v) Was an official report developed by the board? vi) If yes, was it submitted to at least one agency? vii) Were the criteria used by the board for balancing the benefit/risk explicit? viii) If yes, can you provide them? (No useful replies to 3 viii) ix) When the agency reviewed the situation:[If multiple countries/agencies were involved, describe one majorcountry's effort] (a) was the manufacturer: – invited to meetings for hearing/testimony? (b) Were outside experts involved? (c) If yes, was the debate or opinion made public? Yes * No * Which country have you described? x) If the decision was adverse, was there an appeal? xi) If yes, how was it done? Results of survey of regulators 1. For major post-marketing safety issues, is your agency served by at least one designated/standing expert committee or board? 2. If yes, is/are board(s) legally mandated? (Citation for mandate provided for all 7) 3. Please provide details of the board or expert group.
Number of members: Varied from 5 (2 regulators) to as many as 35 Affiliation/QualificationsMedicine (1), multidisciplinary (7), specialists (7), consumer (1) iii) Frequency of meetings: Twice weekly (1), approximately monthly (4), bimonthly (1),quarterly (3) iv) Duration of typical meeting: From 2 to 6 hours (4), one day (3), 1-3 days (2) Purpose of meeting:Exchange of information (3), PMS drug suspension/withdrawal (7),Drug approval/withdrawal (3). Signal identification/confirmation,regulatory action, data analysis, evaluation of ADR reports.
vi) Are the recommendations of the board/expert group advisory or Advisory (9), Binding (0) 4. If there is no board, do you use individual outside/independent experts for major safety issues? Not applicable. Three indicated use of additional experts ad hoc.
If yes, please give details.
Traditional remedies (1) What is the mechanism for communication between the expert(s)and regulators? Answer: Ad hoc individuals and groups, telephone,reports.
5. Please consider the most recent major safety issue/crisis/withdrawal. Was this board/expert group involved? (Including ad hoc) With which drug did it occur?Third-generation oral contraceptives (5), anorectic (1), felbamate(1), oxytocin + prostaglandin derivatives (1), not specified (1).
What was the issue?Thromboembolism (5), restricted use (1), aplastic anaemia (1), deathand irreversible cerebral damage in mother or baby (1), not specified (1) iii) If yes, how? Review data, advise on action.
iv) Were the board/expert groups' deliberations leading to the final decision made public? Were the criteria used made public? No response – 1 vi) Were the criteria transparent? No response – 2 6. When the agency reviewed the situation, was the manufacturer involved Whether or not present and involved at a hearing, how was themanufacturer involved? submission of written report No response – 2 consulted for hearing/testimony No response – 2 iii) participated in debate No response – 3 7. What is the mechanism for appeal against such major adverse safety decisions?There is a process (8): within the same agency (2); with a higher federalagency (1); at minister's office level (2) (one involving a 5-stage process);through administrative courts (3) Report of CIOMS Working Group IV CIOMS publications may be obtained directly from CIOMS, c/o WorldHealth Organization, Avenue Appia. 1211 Geneva 27, Switzerland. Theyare also distributed by the World Health Organization, Distribution and Sales Unit, Avenue Appia, 1211 Geneva 27, Switzerland and are availablefrom booksellers through the network of WHO sales agents. A list ofthese agents may be obtained by writing to the above address.

Source: http://www.dec.gov.ua/site/files/farmakonaglyad/novunu/brb.pdf