would require comparison with LLDBs and with active surveillancesystems to see what proportion of cases are reported to each andhow that proportion changes with the type and severity of adverseevent reported. Some existing research on statistical modeling ofadverse event reporting systems shows that relative risks must bevery high (given the state of underreporting) or the adverse eventextraordinarily unusual for passive reporting systems to detect adverseevents.

Individual case reports, if well-reported, have a role in assessingcausality under specific circumstances. There was considerable disagreementabout the value of larger investment in passive surveillance systemsfor studying the safety of vaccines. There was call for further consideration of when case reports are useful, how they are useful, how to improvethem, and how to make sure that cases are reported and that reports are disseminated.

LARGE-LINKED DATABASES

Introduction

In 1990 CDC began plans for a large-linked database (LLDB) knownas the Vaccine Safety Datalink. The use of automated vaccinationreports and the ability to link them to health outcomes permit oneto screen for associations regardless of their latencies. Workingbackward from the outcome, it is possible to identify children whohave or have not experienced a particular outcome and then determinethe proportion of children who had a history of recent vaccine exposure.Alternatively, working forward from the vaccine exposure, it is possibleto identify children who have or have not been vaccinated and determinethe rates at which particular outcomes occur.

The Datalink presently involves more than 500,000 children from birththrough 6 years of age (with an annual birth cohort of about 70,000)who obtain health care from one of four large health maintenanceorganizations (HMOs) in Seattle, Portland, Oakland, and Pasadena.The Indian Health Service is considering a proposal to include NativeAmerican children in the study. Extension of the database to includeinformation on adolescents and adults is planned as well. In addition,a new database that would provide more timely, county-specific mortalitydata is envisioned.

The objectives of the study are (1) to evaluate apparent associationsbetween vaccines and adverse events (identified from the medicalliterature or from VAERS) and, if confirmed, to seek characteristicsthat predispose children to these outcomes and thus might contraindicatevaccination; (2) to search for currently undiscovered associations;and (3) to characterize and standardize the national passive surveillancesystem by determining reporting efficiency, discovering biases, anddescribing other operating characteristics of passive surveillancesystem. The system could also serve



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 10
Research Strategies for Assessing Adverse Events Associated withVaccines:: A Workshop Summary would require comparison with LLDBs and with active surveillancesystems to see what proportion of cases are reported to each andhow that proportion changes with the type and severity of adverseevent reported. Some existing research on statistical modeling ofadverse event reporting systems shows that relative risks must bevery high (given the state of underreporting) or the adverse eventextraordinarily unusual for passive reporting systems to detect adverseevents. Individual case reports, if well-reported, have a role in assessingcausality under specific circumstances. There was considerable disagreementabout the value of larger investment in passive surveillance systemsfor studying the safety of vaccines. There was call for further consideration of when case reports are useful, how they are useful, how to improvethem, and how to make sure that cases are reported and that reports are disseminated. LARGE-LINKED DATABASES Introduction In 1990 CDC began plans for a large-linked database (LLDB) knownas the Vaccine Safety Datalink. The use of automated vaccinationreports and the ability to link them to health outcomes permit oneto screen for associations regardless of their latencies. Workingbackward from the outcome, it is possible to identify children whohave or have not experienced a particular outcome and then determinethe proportion of children who had a history of recent vaccine exposure.Alternatively, working forward from the vaccine exposure, it is possibleto identify children who have or have not been vaccinated and determinethe rates at which particular outcomes occur. The Datalink presently involves more than 500,000 children from birththrough 6 years of age (with an annual birth cohort of about 70,000)who obtain health care from one of four large health maintenanceorganizations (HMOs) in Seattle, Portland, Oakland, and Pasadena.The Indian Health Service is considering a proposal to include NativeAmerican children in the study. Extension of the database to includeinformation on adolescents and adults is planned as well. In addition,a new database that would provide more timely, county-specific mortalitydata is envisioned. The objectives of the study are (1) to evaluate apparent associationsbetween vaccines and adverse events (identified from the medicalliterature or from VAERS) and, if confirmed, to seek characteristicsthat predispose children to these outcomes and thus might contraindicatevaccination; (2) to search for currently undiscovered associations;and (3) to characterize and standardize the national passive surveillancesystem by determining reporting efficiency, discovering biases, anddescribing other operating characteristics of passive surveillancesystem. The system could also serve

OCR for page 10
Research Strategies for Assessing Adverse Events Associated withVaccines:: A Workshop Summary to compare products made by different manufacturers and to studyvaccine efficacy, lot-specific differences in vaccines, reactionsthat occur following the simultaneous administration of more thanone vaccine, and the incidence of age-appropriate administrationof vaccines. The exposures of interest are all childhood vaccines. The outcomesof particular and current interest are a number of adverse eventsthat have been reported to occur after vaccination. The literaturewas carefully reviewed in 1990, and 34 adverse events were thoughtto be biologically plausible. The determination of which outcomesto study, however, is envisioned as a long-term dynamic process,and the IOM reports published since work began on designing the LLDB(Institute of Medicine, 1991, 1994a,b) are being used to reevaluatethe list to see whether other events can and should be included.The study is evolving, and data collection plans can be changed asthe study proceeds. Hospital discharge records are not sufficient as a basis for gatheringinformation about vaccine-related adverse events, partly becausesome of the serious reactions of interest are treated or evaluatedin clinics and urgent care facilities. Thus, case ascertainment isaccomplished directly from diagnoses upon hospital discharge, treatmentrecords at clinics and urgent care centers, and death certificatesand autopsy reports and indirectly from records of laboratory culturesand tests, specialist referrals, diagnostic procedures, and drugprescriptions; all of these data exist in the computerized data systemsof the HMOs. Even the use of a system that includes records of alltreatments for putatively vaccine-related adverse reactions doesnot guarantee that all relevant information will be obtained, becausediagnoses are not always made, medical attention is not always sought,and the disposition to resolve medical problems and to be vaccinatedmay be related. Associations that emerge consistently from site tosite and year to year by either path will be subjected to furtherepidemiologic studies to confirm the association based on a thoroughreview of conventional medical reports. Study statisticians have calculated the statistical power of thetechnique, that is, the probability of being able to detect particularvaccine-adverse event associations during the time allotted to thestudy, taking into account the timing of vaccine exposures, misclassificationof adverse events, compliance with recommended vaccination schedules,and the like. Investigators have recommended that the study continuebeyond the allotted 4 years because some rarely occurring adverseevent-vaccine relations will not be detectable statistically overthat period of time, if they do in fact occur. However, it is notclear how long of a follow-up would be required. Most of the first 2 years of the project were spent on creating casedefinitions and standardization among the HMOs and their databases.Some existing databases, for example, serve only administrative purposes,and attempts are being made to increase the quality of the recordsfrom those databases for scientific study. Much heterogeneity remains.

OCR for page 10
Research Strategies for Assessing Adverse Events Associated withVaccines:: A Workshop Summary The United Kingdom has successfully used LLDBs in a three-step processof signal detection, causality assessment, and risk calculation.In the incident related to the Urabe-strain mumps vaccine mentionedearlier, for example, 5 public health laboratories participated initially;the number has since been expanded to 20. Each of the laboratoriessearched its database for mention of any child who had undergonea lumbar puncture (and thus might have been treated for suspectedmeningitis). This information was then cross-checked with districthealth databases to determine the vaccination history of each suchchild. The success of this project was attributed in part to thesmall, manageable size of the individual component (a public healthlaboratory). Discussion Strengths and Weaknesses of LLDBs The LLDB was thought to be a powerful tool for vaccine safety studies.Two valuable contributions of LLDBs noted by participants were theability to compare children receiving different formulations of thesame vaccine and the ability to compare new vaccines with standardvaccines. It is possible, for example, to compare vaccinated andunvaccinated children at age 3 months or to compare the same childbefore and after vaccination. Comparisons between children who receivedvaccines according to different schedules could be made as well.Because the cohort is being followed for several years, it mightbe possible to study long-term outcomes and chronic conditions aswell. The LLDB was characterized by one participant as a framework in whichthe data for most of the key elements of a good epidemiologic studyare already in place. It then should be possible not only to generatehypotheses but to begin to test those hypotheses as well. Passivesurveillance reporting systems are not amenable to answering questionsin a timely manner. Well-designed LLDBs should avoid that pitfall. It was suggested that, in addition to safety concerns, issues aboutthe persistence of vaccine efficacy over the time frame of the studycould also be addressed by LLDBs. It might be possible to look atthe occurrence of vaccine-preventable diseases in vaccinated children,for example. A PHS representative indicated that LLDBs have indeedbeen used to look at one particular efficacy issue. Recently, resultsfrom a clinical research study led a manufacturer to question thepotency of a lot of a particular vaccine. An LLDB was used to identifywhich children had been given that vaccine. It was then possibleto test a sample of those children for antibody levels and to identifywhich vaccine lots produced lower antibody levels than others. Several participants stressed that some questions will not be answerableby the LLDB method and that this should be acknowledged from thestart. The

OCR for page 10
Research Strategies for Assessing Adverse Events Associated withVaccines:: A Workshop Summary successful use of LLDBs in the United Kingdom such as described abovewas due in part to the adverse event investigated. Viral meningitisis a fairly clearcut diagnosis. The difficulty of addressing issuesabout conditions such as epilepsy, sensorineural deafness, or diabetesfor which the time of onset may be different from the time of diagnosiswas raised. Any condition that is chronic, longterm, or of insidiousonset may be difficult to detect by this method. Unfortunately, theseare the same conditions that are not easily identified in case reports.Fundamental laboratory research, such as the identification of biomarkersfor vaccine-induced pathology, may be required to study conditionsof this sort. Case Definitions The importance of standardized case definitions was stressed by severalparticipants; to plan ways of finding outcomes reliably, it is necessaryto know with some precision what is being sought. The indicationby a representative of PHS that PHS's case definitions could be madewidely available once they have been determined was well receivedby the participants. As part of developing the case definitions, attention has been paidto planning search strategies for particular conditions within thedatabase. Given an interest in a particular outcome, it is necessaryto delineate the appropriate combination of possible discharge diagnoses,laboratory data, radiology procedures, pharmaceutical information,and so forth that will be both sensitive and specific in helpingto detect cases of the outcome in question. Evaluation of LLDBs A formal system for evaluating LLDBs on a regular basis was suggested.Perhaps this could be done through an oversight panel of three tofive people who could provide an external critique of the study andits methodology. This could be similar to the oversight done forefficacy trials, in which knowledgeable people who are separatedfrom the immediacy of the study review the procedures of a study. The LLDBs show great, but not unlimited, promise. Unfortunately,they will be unable to detect adverse events of insidious onset orwith a long latency from exposure. Participants urged careful consideration of the adverse events to be studied.