active drug safety surveillance: a tool to improve public health

pharmacoepidemiology and drug safety 2008; 17: 1175–1182ce (www.interscience.wiley.com) DOI: 10.1002/pds.1668
Published online 30 September 2008 in Wiley InterScien
ORIGINAL ARTICLE

Active drug safety surveillance: a tool to improvepublic healthy,z

Richard Platt MD, MS1*,x, Leanne Madre JD, MHA2x,Robert Reynolds ScD3,4x and Hugh Tilson MD, DrPH5x

1Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, Boston,MA, USA2Duke Translational Medicine Institute, Durham, NC, USA3Department of Epidemiology, Safety and Risk Management, Pfizer Inc, New York, NY, USA4Department of Epidemiology, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA, USA5University of North Carolina School of Public Health, Chapel Hill, NC, USA

SUMMARY

Purpose Ensuring that drugs have an acceptable safety profile and are used safely is a major public health priority. TheCenters for Education and Research on Therapeutics (CERTs) convened experts from academia, government, and industry toassess strategies to increase the speed and predictive value of generating and evaluating safety signals, and to identify nextsteps to improve the US system for identifying and evaluating potential safety signals.Methods The CERTs convened a think tank comprising representatives of the groups noted above to address these goals.Results Participants observed that, with the increasing availability of electronic health data, opportunities have emerged tomore accurately characterize and confirm potential safety issues. The gain for public health from a highly coordinatednetwork of population-based databases for active surveillance is great and within reach, although operational questionsremain. A collaborative network must create a working definition of a safety signal, screening algorithms, and criteria andstrategies to confirm or refute a signal once identified through screening. Guidelines are needed for when and how tocommunicate a signal exists and is being evaluated, as well as the outcome of that evaluation.Conclusion A public–private partnership to create a network of government and private databases to routinely evaluate andprioritize safety questions is in the public interest. Better methods are needed, and a knowledgeable workforce is required toconduct the surveillance and understand how to interpret the results. The international community will benefit from theavailability of better methods and more experts. Copyright # 2008 John Wiley & Sons, Ltd.

key words—drug safety; active surveillance; safety signals

Received 17 July 2008; Accepted 31 August 2008

*Correspondence to: Dr R. Platt, Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim HealthCare, Boston, MA, USA. E-mail: [email protected] Platt has received funding since 2000 fromAmerica’s Health Insurance Plans, Agency for Healthcare Research and Quality, Centersfor Disease Control, Crohn’s and Colitis Foundation of America (via the CDC Foundation), the US Food and Drug Administration,GlaxoSmithKline, Massachusetts Department of Public Health, National Institutes of Health, National Committee for Quality Assurance,Parke Davis, Pfizer, Robert Wood Johnson Foundation, Sanofi-Aventis, Sanofi-Pasteur, SmithKlineBeecham, TAP Pharmaceuticals, Wyeth.LeanneMadre: none. Robert Reynolds is an employee and shareholder of Pfizer Inc. Hugh Tilson consults widely on safety surveillance withindustry, academia, healthcare organizations, and government agencies (complete list available on request).zThe views expressed in this paper do not necessarily represent the views of any individual participant or organization (government,academic, or private).xOn behalf of participants in the CERTs Pharmacovigilance Think Tank, Rockville, MD, 22–23 May 2007. For a complete list of CERTsPharmacovigilance Think Tank participants see the Appendix.

Copyright # 2008 John Wiley & Sons, Ltd.

1176 r. platt ET AL.

INTRODUCTION product safety thus contributed to the emerging

Ensuring that our nation’s drugs have an acceptablesafety profile and are used safely is a major Americanpolitical and public health priority. During the pastseveral years, there have been many efforts—notablyby the Food and Drug Administration (FDA), Instituteof Medicine (IOM), the private sector, and, not least,Congress itself—to evaluate and improve the way inwhich drug risks are assessed and managed. In 2005,the FDA requested that the IOM review the US drugsafety system and make recommendations forimprovement in light of the then recent withdrawalof Vioxx from the market and other high profile safetyconcerns.1 Based on the committee’s findings andrecommendations, the IOM urged the FDA to take agreater role in gathering and analyzing data onmedications’ risks and benefits after approval, and tosupport the science and infrastructure for doing so.2

Not long after the committee report, ‘‘The Future ofDrug Safety,’’ was released, the FDA held a publicmeeting to explore the idea of linking private sectorand public sector post-marketing safety efforts tocreate a virtual, integrated, electronic ‘‘sentinelinitiative’’ to collect, analyze, and disseminatemedical product safety information to clinicians andpatients.3

Investigators from the Centers for Education andResearch on Therapeutics (CERTs), a public–privatepartnership program of the Agency for HealthcareResearch and Quality (AHRQ), have actively con-tributed to FDA meetings discussing the sentinelinitiative and several other national discussions abouthow to improve the US system for drug safety. As partof the FDA Modernization Act of 1997, Congresscreated the CERTs to conduct research and provideeducation to advance the optimal use of drugs, medicaldevices, and biological products. In furtherance ofits mission, the CERTs have convened a number ofthink tank meetings to develop recommendations toimprove the assessment, communication, and appro-priate use of therapeutics.Subsequent to FDA’s sentinel initiative meetings,

the CERTs convened a group of experts fromacademia, government, and the private sector inMay 2007 to discuss strategies to improve how thenation identifies and evaluates potential safety signalsusing active surveillance. The meeting occurred soonafter the FDA’s first public meeting to solicit inputabout a new post-marketing safety program, andbefore Congress passed the FDA Amendment Act of20074 or the FDA’s announcement of the sentinelinitiative.5 This CERTs-convened deliberation of

Copyright # 2008 John Wiley & Sons, Ltd. Pha

policies and program development that occurredduring 2007–2008.

The meeting was conducted under the leadership ofthe CERTs Steering Committee, in collaboration withthe AHRQ, FDA, the Critical Path Institute, and thePharmaceutical Research and Manufacturers ofAmerica (PhRMA). The meeting objectives were toassess the potential for existing and new strategies toincrease efficiency of generating and evaluating safetysignals, and to identify next steps necessary toimprove the US system for identifying and evaluatingpotential safety signals. In order to ensure focus uponthese objectives, several issues were declared ‘‘offlimits.’’ Such issues included therapeutic benefitevaluation, passive surveillance systems (includingso-called ‘‘spontaneous’’ or voluntary adverse reac-tion reports), and international considerations. Thisreport summarizes the deliberations of this think tankon active surveillance.

BACKGROUND

During the development of a medical product, safetydata are systematically evaluated in rigorouslydesigned pre-clinical and clinical studies conductedby the sponsor for evidence of a signal of a potentialsafety problem. The World Health Organization hasdefined a signal as ‘‘reported information on a possiblecausal relationship between an adverse event and adrug, the relationship being unknown or incompletelydocumented previously.’’6 A signal may be clinicallyimportant, even when there is only a single event, as isoften the case with liver failure or Stevens JohnsonSyndrome. More often, a signal is recognized from anunexpected frequency of a more commonly occurringserious event in a patient with or without an apparentunderlying risk. The sentinel system discussed in thispaper could complement existing passive surveillancesystems in identifying more common serious adverseevents such as myocardial infarction. With improvedtechnology measurement, signals are also increasinglybeing identified from surrogate markers for clinicalevents; for example, an increase in liver function testsor worsening of an existing condition that may beassociated with medication use. In these cases, therelationship between the surrogate marker and theclinical event may, or may not, be well characterizedor have been shown to be predictive.

The process of evaluating emerging safetyinformation continues after a new drug, biologicagent, or medical device is marketed. In addition to

rmacoepidemiology and Drug Safety, 2008; 17: 1175–1182DOI: 10.1002/pds

active drug safety surveillance 1177

post-marketing clinical trials and other evaluationsconducted for new indications, observational epide-miology studies may be conducted to quantify/identifyknown or suspected potential risks. These studiescomplement clinical trials because they reflect real-world use and provide data on at-risk subpopulationsand patients who have comorbid conditions and useother concomitant treatments. Such individuals areoften underrepresented in pre-marketing clinicaltrials. When observational epidemiology studies areconducted using large health insurance claims andelectronic medical record (EMR) databases, it may bepossible to examine a larger number of exposedpatients for longer durations than generally is feasibleor achieved during drug development, and provide anopportunity to assess the effects of chronic exposure.The larger the population whose experience iscaptured in the database, the larger the populationexperiencing treatment with (exposure to) an agentunder study, and thus the more rapidly assembled thecohort and the more rapidly such studies may be ableto achieve the detection of important adverse events, ifthey are occurring.

While the more frequent use of enhancedpharmacoepidemiologic approaches holds promise,observational studies have many known limitations—especially confounding by indication and potentialchanneling bias—which make it challenging toselect an appropriate comparison group. Variable orunmeasurable adherence to therapy additionallyconstrains the ability to assess treatment effectiveness.

DATA RESOURCES

For many years, pharmacoepidemiologists have usedadministrative data, such as Medicaid and privateinsurance claims data, to evaluate therapeutic safetyquestions. With the increasing availability ofelectronic clinical data that have been collected indefined populations, great opportunities have emergedto more accurately characterize and confirm potentialsafety issues. Additional information available viaEMRs can include laboratory results, medications notnecessarily reflected in billing systems, vital signs,and demographics. The addition of clinical data withinautomated databases can enhance researchers’ abilityto harness the size and power of these data to help toidentify and evaluate safety signals.

As an example, the Centers for Disease Control andPrevention (CDC), through its Vaccine SafetyDatalink (VSD) project, has detected early signs ofadverse events associated with vaccines based ona study population of about 8 million covered lives of


8 health plans.7 Notably, the VSD functions as adistributed data network, with primary data remainingin the possession of the individual health plans. Avariety of methods are used to allow the data to beanalyzed without creating a pooled dataset. When it isnecessary to create a pooled dataset to address aspecific question, the size of the population and theamount of data per person that is included areminimized because, for example, it will only includethe relevant patient population—that is, patients whohave received the drug or vaccine of interest or whoare part of the comparator group. Standard format filesand programs for analysis and quality checks can beused across dataholders so that the amount of pooleddata is reduced. In addition to the distributed datamodel, a potential future model is one which usestechnology to allow actual linkage of data systems insuch a way as to address concerns of confidentialityand privacy.If the publicly funded programs of Medicaid,

Medicare, and the Veterans Administration healthsystem made their data available for surveillancepurposes, almost 90 million lives, including a numberof vulnerable populations, could be studied foridentifying potential public health safety concerns.While public health programs present an opportunity,the private sector is already organizing initiatives toexpand drug safety monitoring capabilities. i3Drug-Safety and HealthCore use health plan data coveringtens of millions of individuals for these purposes. Adistributed data network that includes both public andprivate sector data sources would serve a huge publichealth need. The five largest health plans alonerepresent almost 90 million lives, and, in combinationwith the publicly funded programs, would cover alarge majority of the population.The think tank participants noted that a fully

effective surveillance program requires access to fulltext medical records in order to confirm outcomes formany of the questions of interest and to ascertain someinformation that is not available in automated form.Full medical records would be necessary to confirmrelatively rare outcomes and, in some cases, to collectcertain test results.Automated population-based data do not comprise

the sole opportunity for more active surveillance.More traditional hands-on observational study app-roaches such as registries provide another resource foractive safety surveillance. A registry is an organizedsystem that uses observational methods to collectuniform data (clinical and other) generally prospec-tively from patients and/or physicians to evaluatespecified outcomes for a population defined by a



particular disease, condition, or exposure, and thatserves a predetermined scientific, clinical, or policypurpose.8 Registries can be important tools for drugsafety because they can collect safety information thatcannot be ascertained from other sources, such asadministrative claims or EMRs data. Based upon thisapproach, several drug exposures in pregnancyregistries have been created to monitor for possiblesignals of birth defects.9 Registries also exist to trackvital statistics, various types of cancer, and birthdefects.10 The participants noted potential synergiesthat could result by augmenting registries’ prospec-tively collected data with information obtained fromEMRs and claims data.Like other data resources, observational studies

established for other purposes can potentially be usedto evaluate drug benefits and risks, assumingappropriate parameters are taken into account. Asan example, the Can Rapid Risk Stratification ofUnstable Angina Patients Suppress Adverse Out-comes with Early Implementation of the ACC/AHAguidelines (CRUSADE) registry was a nationalquality improvement initiative designed to increasethe practice of evidence-based medicine for patientsdiagnosed with non–ST-segment elevation acutecoronary syndromes.11 Related to its purpose, severalstudies were undertaken to evaluate care of thispopulation of patients, including a study of the use ofintravenous morphine to treat patients with symp-toms of chest pain or heart attack which found thatthe use of this drug may actually increase death ratesfor these patients. 12

EMRs are also likely to become an importantresource for post-marketing safety surveillance,particularly as these become more widely dissemi-nated and are used in defined populations for whichcomplete medical care data are available. WhileEMRs provide a rich source of clinical data on drugand vaccine exposures and outcomes, it is importantthat EMR databases provide both enrollment anddisenrollment dates in order to be maximally usefulfor safety monitoring.AHRQ has released a user guide for registries that

includes information about the design and imple-mentation of patient registries, the analysis andinterpretation of data from patient registries, and theevaluation of the quality of a registry or one of itscomponents.8 An area of great promise is combiningregistry data with clinical and claims data available inother data sources. A registry that includes detailsabout the devices (e.g., make/model) could becombined with clinical data in order to better trackpositive as well as potentially negative outcomes.


AN ACTIVE SURVEILLANCE APPROACH

Think tank participants agreed the potential gain forpublic health from a highly coordinated system ofpopulation-based databases for active surveillancewill be great and is within our reach. Public healthefforts will be strengthened by earlier identificationand assessment of potential therapeutics risks, whichwill facilitate better communication about andmanagement of actual risks within the context ofbenefits for individual patients. Once established, thesystem can be leveraged to answer other importantpublic health questions, including those of therapeuticeffectiveness and use. The FDA has recentlyannounced its sentinel initiative that would use thismodel.5 The think tank provides some recommendedsteps to ensure success.

To begin, the system will need to be constructedfrom the candidate data resources. It was anticipatedthat the FDAwill have a leadership role in identifyingpost-marketing safety topics of priority. In the best ofcircumstances, the system will be able to address post-marketing safety plus other topics. In that case, amultidisciplinary board with public and privaterepresentation would ideally develop a strategy forensuring the system can support multiple purposes.

Many operational questions must be addressed. Acollaborative effort must create a working case orevent definition of a safety signal, screening algor-ithms, and criteria and strategies to confirm or refute apotential signal once identified through screening. Apublicly available library of codes for case definitionswould also be useful to standardize approaches toevaluation of potential safety signal events. Guidelineson when and how to communicate to stakeholders—inparticular physicians and patients—that a signal existsand is being evaluated, as well as the outcome of thatevaluation, must also be developed, thereby ensuringtransparency (and trust) on the one hand withoutcreating premature or inappropriate concerns on theother.

In order to obtain consensus on these commonprocess elements, a multi-sector advisory committeewith public and private sector representation should beconvened. Some of the tasks this advisory committeewill be charged with include: (1) prioritize signals forevaluation (the board may wish to consider adesignated events list and criteria for ensuring thatsignals are evaluated with the proper context of allavailable data); (2) incorporate lessons from past drugsafety evaluation experiences and system ‘‘failures’’to refine operational definitions and improve thesignal evaluation process; (3) determine types of



signals that can and can not be evaluated; (4)determine best analytic methods and approaches;and (5) provide guidance on the factors to consider insignal interpretation and communication. A coordi-nating mechanism must engage the expertise of thefield to integrate findings from this system with theother vital complementary data regarding drug safetyand safe drug use emerging from randomized clinicaltrials, mechanism-of-action studies, spontaneousreports, and observational studies. Likewise, timeshould be spent considering strategies to addressprivacy, liability, and safe harbor concerns.

Among the most pressing of the needs articulatedabove is development of better methods to identifysignals of excess adverse outcomes. If not properlytested and validated, quantitative activities mightgenerate many potential signals that are of limitedpredictive value but still must be evaluated. Such ascenario would overwhelm current pharmacovigilancesystems, fail to provide meaningful information forregulatory and public health decision-making, anddivert resources away from the evaluation of signalsarising from more robust data sources or study designs(e.g., randomized clinical trials). In order to avoid thisscenario, it is critical that a number of key operational,methodological, and validation activities are initiatedand completed before the effective and reproducibleapplication of quantitative data mining activities to theenvisioned system of automated epidemiology data-bases is realized. Further, development of thesemethods must occur in conjunction with workoutlining the process for systematic evaluation ofsafety signals. Likewise, there are methodologicalchallenges for using population-based datasets andregistries to evaluate effectiveness and benefit.

In order to make progress on these issues, theestablishment of technical working groups is recom-mended to compare and contrast current methods,including exploring the application of public healthsurveillance methods and other data analytic methodsnot currently applied within the field of study of drug,device, and vaccine safety. Separately, it is alsoimportant to establish principles for selecting drugsand signals for regular surveillance. Questions toconsider include: what is the likelihood of detecting asignal with the available data? Assuming limitedresources, should the priority be to study drugs withhigh use or ones used in vulnerable populations, forexample?

Another key methodologic challenge emerged fromthe think tank conversations about statisticalexpressions of certainty/uncertainty, particularlywhen the usual probabilistic statistics that describe


likelihood of error due to sampling become lessrelevant when examining such a large segment of thepopulation. For these and other reasons, it is importantthat there be an ongoing commitment to methodsdevelopment and evaluation.While think tank participants believe much of the

important work of addressing the methodologic needsof an active surveillance system must be addressed bydedicated workgroups, they did begin discussing someof the high priority topics. That discussion is brieflysummarized here.

Signal identification

New strategies for signal identification might include adesignated list of high priority medical events thatshould be considered for certain drug classes, and theevaluation of signals of potential concern, based onwhat is known about the class and any issues that aroseduring pre-licensure studies. Methods that could beuseful in monitoring drug safety are currently beingdeveloped for vaccines by the VSD. The challenge isto detect real safety problems as soon as possiblewhileminimizing false alarms. The VSD does this bymaximized sequential probability ratio testing(maxSPRT), an extension of Wald’s well-establishedmethod.13 While the maxSPRT method has beenapplied to monitoring vaccine safety, the methodrequires further evaluation to fully understand itsapplication to drug safety monitoring; in particular,how it will address channeling bias and other biasesthat occur frequently in drug safety evaluation, whichmay not be as prominent in vaccine safety evaluation.Another emerging approach is the application of

quantitative data mining activities to spontaneousreports databases to enhance the qualitative screeningcapabilities of expert medical reviews at pharmaceu-tical companies and regulatory agencies. The potentialapplication of these quantitative signal detectionmethods to electronic observational databases (e.g.,insurance claims or EMRs databases) is an additionalsource from which signals may arise.These various analytic methods will need to be

compared and contrasted to fully understand thebenefit and the risks of each method in findingpotential signals. Monitoring the safety of a drug orvaccine beyond its initial use can continue for as longas it is worthwhile, as in the case of concomitant usewith a new complementary product. The trade-offbetween false alarm risk and timely signal detectioncan be reassessed periodically and managed inaccordance with the FDA’s and other stakeholders’concerns and priorities. The answer to ‘‘how certain is


KEY POINTS

� Ensuring that our nation’s drugs have anacceptable safety profile and are used safely isa major public health priority.

� A distributed data network that includes bothpublic and private sector data sources derivedfrom the routine delivery of care will serve animportant public health need, particularly incombination with registries.

� A coordinated network of population-baseddatabases for active surveillance must create aworking definition of a signal of excess risk ofadverse events, screening algorithms, andcriteria and strategies to confirm or refute asignal once identified through screening.

� Guidelines on when and how to communicate tostakeholders—particularly physicians andpatients—that a signal exists and is beingevaluated, as well as the outcome of thatevaluation, must be developed.

� Better methods are needed to identify andevaluate potential safety signals, and a knowl-edgeable workforce is required to conduct thestudies and to understand how to interpret theresults.


sure enough?’’ is situation-dependent and may dependon the seriousness of the signal, use of the product,public health implications, and benefit for thepopulation using the product.

Signal evaluation

There are many different data sources to which onecan look for potential signals. An important questionto address is how much weight should a data source begiven in a particular situation, and whether any filterscan be applied to better understand whether a signal isan artifact of surveillance itself (‘‘noise’’) or a realpotential risk. Ideally, the system needs to have highreliability to filter out false-positive drug-event pairsand to find true drug-event pairs. As noted earlier,there is a problematic reality of generating too manysignals to evaluate, the number of which wouldpotentially be false and unhelpful, if not harmful, tothe decision-making process. As with signal identi-fication methods, signal evaluation methods alsorequire additional analytic evaluation on best methodsand procedures.

Workforce

Another key factor to improve the process ofevaluating signals is the challenge of developing aworkforce capable, competent, and of sufficient size toundertake this work, and of deploying it effectively.This will require identifying trained professionals toaddress current signal evaluation challenges at theFDA, its Advisory Committees and in industry, andassessing workforce needs and developing trainingefforts now for the future. Some participants suggestedthat the process could be improved if epidemiologistsare involved in the new drug application (NDA)process in order to better develop post-approvalactivities. In addition, following the IOM recommen-dations, epidemiologists should be included on theFDA advisory committees.Regarding future needs, a study of immediate and

long-term drug safety workforce needs is urgentlyneeded. Agreement must be reached regarding thecompetencies for pharmacoepidemiology, particularlyrelating to the rapidly evolving automated sentinelsystem approaches, and major work is needed tomatch curriculum to competency. Funding for trainingprograms in pharmacoepidemiology should be devel-oped by the National Institutes of health, industry, andothers, perhaps modeling after the CDC epidemiologyintelligence service model.14 In addition, academia,government, and industry should develop opportu-


nities for new epidemiologists to learn from theexperts in drug safety evaluation through fellowshipand mentoring programs.

CONCLUSION

The public will benefit from better information aboutthe benefits and risks of medications. There areongoing and emerging efforts to better harnessexisting data to more effectively and efficientlyidentify and evaluate potential medication risks. Apublic–private partnership to create a network ofgovernment and private data to enable evaluation ofroutine and priority safety questions is in the publicinterest. However, better methods are needed toaddress the limitations of pharmacoepidemiology, anda knowledgeable workforce is needed to conduct thestudies and to understand how to interpret the results.We need researchers who know the limitations andquirks of the data and practice settings, and we needadvisors who can help the network be appropriatelytransparent and who can communicate well withdecision-makers and the public. To ensure that theemerging coordinated sentinel system is truly a publichealth resource, the federal government should



provide funds and other support to the public share of apublic–private infrastructure.

Since the meeting, federal legislation has beenenacted to establish a post-marketing risk identifi-cation and analysis system drawing on public andprivate data resources and expertise, and the FDA hasannounced a national strategy for monitoring medicalproduct safety.4,5 The health of the public depends oneffective actions in this area now.

ACKNOWLEDGEMENTS

We thank the members of the planning committee forthe development of the meeting agenda and frame-work. Planning committee members are noted by an �

in the list of participants (see Appendix). We alsothank Elizabeth Andrews, Bob Assenzo, SolomonIyasu, Judy Staffa and Anne Trontell for leading thebreak-out groups and for summarizing workgroupdiscussions about strategies to improve signal identi-fication and evaluation. We thank Amanda McMillanof Duke Clinical Research Institute for editorial assist-ance and manuscript preparation. The meeting wassponsored by the CERTs, a program of the Agency forHealthcare Research and Quality (AHRQ); TheCritical Path Institute; and the PharmaceuticalResearch and Manufacturers of America. Grant sup-port was provided by the CERTs cooperative grant(U18HS013474). Preparation of the manuscript wassupported in part by the Duke Clinical and Transla-tional Science Award (1 UL1 RR024128-01) from theNational Center for Research Resources (NCRR), acomponent of the National Institutes of Health (NIH),and NIH Roadmap for Medical Research.

REFERENCES

1. Food and Drug Administration. FDA acts to strengthen thesafety program for marketed drugs (press release). FDA, Rock-ville, November 2004. http://www.fda.gov/bbs/topics/news/2004/NEW01131.html [Accessed 12 June 2008].

2. Institute of Medicine Committee on the Assessment of the USDrug Safety System. The Future of Drug Safety: Promoting andProtecting the Health of the Public. National Academies Press:Washington DC, 2007.

3. Institute of Medicine. Forum on Drug Discovery, Development,and Translation: Symposium—The Future of Drug Safety:Challenges for the FDA, 12 March 2007.

4. Food and Drug Administration Amendments Act of 2007(FDAAA), HR 3580, 110th Congress, 1st Session, 2007.

5. Food and Drug Administration. The sentinel initiative: nationalstrategy for monitoring medical product safety. FDA, Rockville,May 2008. http://www.fda.gov/oc/initiatives/advance/reports/report0508.pdf [Accessed 12 June 2008].


6. World Health Organization (WHO). The safety of medicines.WHO,Geneva, September 2005. http://www.who.int/mediacentre/factsheets/fs293/en/ [Accessed 12 June 2008].

7. Lewis E, Greene SK, Yin R, et al. Update: recommendationsfrom the advisory committee on immunization practices (ACIP)regarding administration of combination MMRV vaccine.MMWR Morb Mortal Wkly Rep 2008; 57: 258–260.

8. Gliklich RE, Dreyer NA. (eds). Registries for EvaluatingPatient Outcomes: A User’s Guide. Prepared by OutcomeDEcIDE Center (Outcome Sciences, Inc. dba Outcome) underContract No. HHSA29020050035I TO1. AHRQ PublicationNo. 07-EHC001-1. Agency for Healthcare Research and Qual-ity: Rockville, MD, 2007.

9. The Antiretroviral Pregnancy Registry. Who we are. Wilming-ton, NC, 2007. http://www.apregistry.com/who.htm [Accessed12 June 2008].

10. National Cancer Institute. Risk factor monitoring and methods:introduction to pharmacoepidemiology and pharmacogenomicsresources. Bethesda, MD, June 2008. http://www92.imsweb.com/tools/pharmaco [Accessed 10 July 2008].

11. Bhatt DL, Roe MT, Peterson ED, et al. Utilization of earlyinvasive management strategies for high-risk patients with non–ST-segment elevation acute coronary syndromes: results fromthe CRUSADE quality improvement initiative. JAMA 2004;292: 2096–2104.

12. Meine TJ, Roe MT, Chen AY, et al. Association of intravenousmorphine use and outcomes in acute coronary syndromes:results from the CRUSADE quality improvement initiative.Am Heart J 2005; 149: 1043–1049.

13. Lieu TA, Kulldorff M, Davis RL, et al. Real-time vaccine safetysurveillance for the early detection of adverse events.Med Care2007; 45(Suppl. 2): S89–S95.

14. Centers for Disease Control and Prevention (CDC). Epidemicintelligence service. CDC, Atlanta, December 2007. http://www.cdc.gov/eis/ [Accessed 12 June 2008].

APPENDIX: PARTICIPANTS IN THE CERTSPHARMACOVIGILANCE THINK TANK,ROCKVILLE, MD, 22–23 MAY 2007

Elizabeth Andrews, PhD,� RTI Health Solutions;Joseph (Bob) Assenzo, PhD,� The Critical PathInstitute; Miles Braun, MD, MPH, US FDA; JaniceBush, MD, Johnson & Johnson PRD; Robert Califf,MD,� Duke Translational Medicine Institute; Eliza-beth Chrischilles, PhD, The University of Iowa;Michael Cook, PhD, Wyeth Research; Trinka Coster,MD, MS, US Army Medical Department; StephenCrystal, PhD, Rutgers, the State University of NewJersey; Francesca Cunningham, PharmD, Departmentof Veterans Affairs; Wanju Dai, MD, Sanofi-Aventis;Gerald Dal Pan, MD,� US FDA; William DuMouchel,PhD, Lincoln Technologies; Bruce Fireman, KaiserPermanente; Victor Gogolak, MS, DrugLogic Inc.;Alan Goldhammer, PhD, PhRMA; David Graham,MD, US FDA; Thomas Gross, MD, US FDA; JoannaHaas, MD, MS, Genzyme Corporation; ManfredHauben, MD, MPH, Pfizer Inc.; Amy Helwig, MD,



Agency for Healthcare Research and Quality; SeanHennessy, PhD,� University of Pennsylvania; AlanHochberg, BS, ProSanos Corporation; Paul Hoerner,US Air Force; Solomon Iyasu, MD, US FDA; MicheleJara, DSc, MPH, Boehringer Ingelheim; StephenKlincewicz, MD,� Johnson & Johnson PRD; JamesKotsanos, MD, Eli Lilly and Company; Judith Kramer,MD, MS, Duke Clinical Research Institute; RichardPlatt, MD,� Harvard Medical School and HarvardPilgrim Health; Marcus Reidenberg, MD, WeillCornell Medical College; Robert Reynolds, ScD,�

Pfizer, Inc.; Nancy Santanello, MD, MS, MerckResearch Laboratories; David Schroeder, MS, DoDTRICAREManagement Agency; Jeffrey Shuren, MD,US FDA; Judith Sills, PharmD, Novartis Pharmaceu-ticals; MaryLou Skovron, PhD, Bristol Myers SquibbCompany; Walter Smalley, MD, Vanderbilt Univer-sity; Scott Smith, PhD,� Agency for HealthcareResearch and Quality; J. Michael Sprafka, PhD,


Procter & Gamble Pharmaceuticals; Judy Staffa,PhD,� US FDA; Brian Strom, MD,� University ofPennsylvania; Hugh Tilson, MD, DrPH,�University ofNorth Carolina at Chapel Hill; Anne Trontell, MD,MPH, Agency for Healthcare Research and Quality;Alexander Walker, MD, i3 Drug Safety; Sheila WeissSmith, PhD, University of Maryland; Marcus Wilson,PharmD, HealthCore/WellPoint; Raymond Woosley,MD, PhD, The Critical Path Institute; Sean Zhao, PhD,Amgen Inc.

�Denotes planning group member

STAFF

Leanne Madre, JD, MHA, Duke Clinical ResearchInstitute; Eva Hill, Duke Clinical Research Institute;Mara McAdams, US FDA.


active drug safety surveillance: a tool to improve public health

Documents