continue to monitor it carefully; this is the action most often taken. Less commonly, more information is obtained, an on-site facility inspection is requested, or lot samples are retested. The review process may lead to changes in the required labeling.
IMPROVING DETECTION OF AND RESPONSE TO ADVERSE EVENTS9
VAERS, LLDB, vaccine manufacturer surveillance measures, and FDA response procedures have several limitations. Speakers and participants voiced several suggestions and discussed measures being undertaken to improve the detection of and response to adverse events associated with vaccination.
Increasing Reporting to VAERS
A CDC study indicated that, as with all passive surveillance systems, underreporting occurs with VAERS; however, reporting efficiency is higher for severe events than for mild ones (Rosenthal and Chen, 1995). An FDA study found that the reporting rate to VAERS varies substantially by state, but those states that reported the most serious events also reported the most events of lesser severity (Braun, 1997).
The concern of some physicians about potential legal liability for an adverse event following vaccination was cited by some participants as a reason for underreporting in VAERS. Theoretically, this should no longer be the case, because the existence of the National Vaccine Injury Compensation Program (established by federal statute in 1986) has resulted in very few suits filed against physicians for vaccine-related injuries since that time. To the extent that physicians are unaware of these circumstances, however, it was suggested that increased education of physicians about the compensation program would help to allay their concerns about reporting adverse events as required.
As another way to increase the number of reports filed with VAERS, representatives from the VAERS contractor, CDC, and FDA will be presenting results obtained from VAERS at professional meetings and in journal articles.
Such efforts are intended to increase the reporting rate to VAERS by conveying to physicians the value and usefulness of the reports that they submit.
To help raise adverse event reporting rates by physicians further, a consumer representative suggested that a course on vaccine risks and benefits, adverse event reporting, and educating parents be introduced into medical and nursing school curricula. The speaker also suggested using funds from the U.S. Department of Health and Human Services to create a public education campaign designed to educate physicians and the public about the importance of monitoring children after vaccination, reporting adverse events following vaccination to the government, and screening children for factors that would place them at higher risk of experiencing adverse events following vaccination. (However, specific risk factors for most adverse events are not known.) CDC is working with several universities to develop materials for both physicians and parents that the explain adverse events that may occur after vaccination and how to report them. It was also suggested that more efforts be made to educate adult vaccines about the possible adverse effects of vaccines. A vaccine researcher noted that physicians have vaccine information statements that they can give to the parents of children being vaccinated, but that such statements for adults about to be vaccinated are less common and not as readily available.
More easily accessible means for reporting adverse events to VAERS were also suggested. A consumer representative suggested that, in addition to existing toll-free telephone numbers,10 an Internet address or World Wide Web site could be instituted to allow physicians, parents, and patients to report adverse events easily and quickly; issues of confidentiality, however, would have to be addressed before such a procedure could be implemented. The Internet contains a vaccine safety information sharing system for people who are actively involved in vaccine research. It is not available to the general public, however, but functions as an alerting mechanism containing, among other things, press reports of vaccine safety issues, newly published research, and study results presented at conferences or meetings, so that researchers are informed and able to respond to the inquiries that they receive.
Noting that most pediatricians lack the time to fill out long forms, a physician suggested that the number of reports to VAERS could be increased by placing the burden of such reporting on the parents of vaccinated children. He suggested that parents be given two cards at the time of vaccination, one for reporting acute reactions to the vaccine and the other for reporting chronic reactions.
A consumer representative suggested that, as part of the legal requirement for physicians to report serious adverse events following vaccination, failing to report such events should result in penalties. Parallel to mandatory vaccination laws that keep children out of school if they are not vaccinated, such laws could prevent physicians from administering vaccines if they neglect to report hospitalizations, deaths, and injuries following vaccination.
Improving the Quality of Information in VAERS
A frequent problem encountered with VAERS is misclassification. An FDA representative pointed out that serious events can be misclassified as nonserious if the reporter does not recognize the seriousness of the event or if a serious event is treated on an outpatient basis. (The current definition of a ''serious" event requires that the individual be hospitalized.) He indicated that he has seen reports of paralyzing disorders such as Guillain-Barré syndrome and Bell's palsy incorrectly labeled as nonserious.
Another criticism of VAERS is that the description of the adverse event is often inaccurate—that is, the reporter inaccurately labels the adverse event. An FDA study found that for nearly one-quarter of the fatalities reported to VAERS, the final cause of death was significantly different from the cause initially reported (Wise, 1994). A CDC study, however, found that nearly 90 percent of the reports of Guillain-Barré syndrome in VAERS were accurate (Haber et al., 1994).
To improve the accuracy of the information reported to VAERS, several speakers or participants suggested developing more standardized case definitions for evaluating the adverse events reported. Under the direction of CDC and FDA, the VAERS staff is currently undertaking several efforts to improve the quality of the data. Among these are improving the automated screening procedures designed to detect reporting or transcribing errors in VAERS data; improving-the consistency in entering, tracking, and analyzing VAERS data; and improving the detection of duplicate VAERS reports entered into the database. CDC, FDA, and Health Resources and Services Administration Division of Vaccine Injury Compensation are establishing a panel of medical experts to provide advice on analyses of difficult cases and are continuing to conduct studies to evaluate the biases present in the system, examining such issues as reporting delays and how they affect the analysis and validity of the data.
To facilitate the VAERS reporting process and to increase the likelihood that physicians will complete VAERS forms, a vaccine manufacturer's representative suggested that the form might be made more specific. Rather than
asking doctors to report all adverse events following vaccination, he suggested that it be limited to reporting only those conditions that are of most concern. A CDC representative noted, however, that if VAERS forms were made too specific, the system would lose the ability to detect previously unknown adverse events. Specificity is best achieved, he said, with more active surveillance measures, such as the standard follow-up protocol of certain VAERS reports or through LLDB.
To improve further the quality of information in VAERS reports, a consumer representative suggested that the person assigned to follow up death reports speak to both the parent of the child and the child's physician to obtain more details about the adverse event. Death reports are currently followed up by seeking additional information from the individual who submitted the report. Therefore, if the report was submitted by someone other than the parents the parents might not be aware that an investigation was undertaken. For example, a parent noted that no government representative had ever contacted him personally seeking more information about his son's death following vaccination with the hepatitis B vaccine. This death was misclassified as sudden infant death syndrome (SIDS) in the VAERS report, despite subsequent autopsy findings that strongly suggested cardiac causes. A consumer representative further suggested that deaths labeled as being caused by SIDS, especially when they occur in infants under 1 month of age, in whom SIDS is rare, should be thoroughly investigated so that they can be correctly reclassified if necessary. The criteria developed by the National Institute of Child Health and Human Development to classify a death as being caused by SIDS could be used to verify SIDS deaths reported to VAERS.
A CDC representative suggested that for those reported adverse events for which there are established diagnostic criteria (e.g., anaphylaxis or multiple sclerosis), a standard follow-up protocol should be sent to the reporter and the physician. FDA and CDC are working on improving follow-up to the reports made to VAERS. On a trial basis, letters will routinely be sent to request more follow-up information for certain adverse events, and more telephone follow-up has been implemented. FDA and CDC also plan to assess how to obtain more information from VAERS forms as well as how to make the forms easier to complete.
A public health policy analyst suggested linking infant mortality reviews, which are conducted extensively in the United States, with VAERS data to help detect and provide more information on vaccine-related deaths among infants. Other participants suggested reviewing other existing databases (e.g., the Medicare database) to seek significant information on immunization status and adverse events.
A consumer representative suggested monitoring the outcomes of serious events, such as seizures, with follow-up at 6 months, 1 year, and 2 years to gather data on persistent sequelae. VAERS currently requests information on recovery status at 60 days and 1 year after selected serious adverse events.
One criticism of VAERS is that its data are not easily accessible either to the medical and research communities or to the public. A physician noted that to obtain VAERS data, he had to employ the Freedom of Information Act, and then the information that he received was not readable by his computer. He suggested that VAERS data be put into a form that can be easily read on a personal computer.11
To help improve accessibility, the publicly releasable, nonconfidential component of VAERS data is being converted to a more user-friendly form. Both the hardware and the software used by VAERS are being updated to make it easier and more efficient to conduct analyses of the data. To improve the public's use of the data in VAERS, FDA also plans to update and clarify the explanatory information that accompanies a VAERS data set. A physician suggested that the accessibility of VAERS data be further improved by regularly publishing summaries of the data in CDC's Morbidity and Mortality Weekly Report.
A participant suggested that VAERS be made more effective by committing more U.S. Department of Health and Human Service funds to increase the staffing and other resources needed to detect and respond to adverse events more expeditiously. A balance could be struck between the resources spent on VAERS and other mechanisms for gathering data on vaccine safety (e.g., LLDB), each of which has its own inherent strengths and weaknesses.
Improving Vaccine Lot Information
Frequent problems encountered by those monitoring VAERS data include the lack of identification of vaccines and the lot numbers of the vaccines associated with the adverse events reported, inaccurate reporting of the lot numbers of vaccines associated with adverse events, and imprecise estimates of the usage of particular lots over time. Lot sizes are needed as denominators for calculating adverse event rates.
A parent of an injured child reported that she had obtained information about the vaccine lot from which her child's dose came and discovered that a number of other children had died or had been injured following receipt of
VAERS data are available electronically in standard ASCII format by contacting NTIS at (703) 487-4650 or http://www.ntis.gov.
vaccine from that lot. She wondered whether there is a threshold of such events that triggers recall of a vaccine lot. FDA officials indicated that because lots are highly variable in size, it is not possible to set a specific number as a trigger for recall. FDA officials reported that they monitor carefully the number of serious events reported relative to the size of each lot. However, they are precluded by federal statute from making public the size of any lot; this information is considered proprietary by vaccine manufacturers.
A physician suggested that FDA establish standard criteria for the recall of vaccine lots as well as effective procedures for ensuring that all recalled vaccine materials are collected. He recommended that whenever FDA recalls a vaccine lot, it notify the media as well as pharmacies, pediatricians, and family practice and internal medicine physicians. FDA officials indicated that such notification does occur.
A vaccine manufacture's representative reported on two pilot programs being planned or conducted in Europe to improve the recording of vaccine lot and identification numbers. One program records on a computerized card an individual's immunization information, including the identification and lot numbers of each vaccine received. The information is entered and updated by computer with software in physicians' offices. Another program plans to add bar codes on each vaccine vial to identify not only the product but also the batch number. An optical scanner at immunization clinics can then easily record and enter into a computer database the information each time that a vaccine is given.
To measure the rate of occurrence of rare adverse events for specific vaccine lots, researchers must know the number of vaccine doses administered. A researcher at the CDC has developed a mathematical model to estimate the number of doses of specific lots of vaccine administered and rates of adverse events (Glasser et al., unpublished). He illustrated two applications of this methodology: (1) identification of lots for which rates of reporting of adverse events exceed the average, and which should be scrutinized if many doses remain to be administered and (2) validation of prelicensure testing standards. This approach was developed in collaboration with scientists in the biological laboratories of the Massachusetts and Michigan public health departments, and the underlying model was validated with information about diphtheria and tetanus toxoids and pertussis vaccine administration inferred from vaccine inventories in county and regional public health departments throughout Michigan, which in turn distribute vaccine to the private sector. Although the observed and expected numbers of administrations of specific vaccine lots were concordant, an epidemiologist suggested that further work with vaccine manufacturers was needed to assess the accuracy of the predicted numbers of doses a particular lot administered. Manufacturers could test the model by
directly monitoring lots to see how well the predicted and actual number of doses in particular lots matched.
To better assess vaccine lot administration, a CDC representative suggested that vaccine manufacturers include postcards in some of the boxes of vaccines distributed. The cards would request that the provider, on the day that the box is opened and vaccine from that box is first given to a patient, write the date on the card and return it by mail. However, a manufacturer who had attempted such an approach believed that studies done in this manner do not generate reliable data because vaccine administration is extremely variable, making vaccine administration impossible to track through indirect methods such as this.
Improving the Usefulness of LLDB Data
The main problem with LLDB is that it is limited by how medical events are captured in their computer data systems during routine visits. It is difficult to discriminate between new events and chronic or recurrent conditions. For example, some databases do not distinguish among a visit for a first seizure, a visit for a recurrent seizure, or a follow-up or check-up visit. By using certain data screening techniques, however, it is possible to subcategorize visits. By limiting seizure reports only to those that occurred in the emergency room and that were separated by 30 days, for example, researchers were able to distinguish visits for actual seizure events from seizure follow-up visits or checkups (Davis, unpublished data). Such screening techniques might make the LLDB more useful, a vaccine researcher suggested.
An FDA representative suggested developing computerized screening methods to identify clusters of potentially related events. Researchers could also develop statistical methods that could more effectively evaluate adverse events following vaccination.
To further identify adverse reactions to vaccines, an FDA representative suggested that larger clinical trials of vaccines be performed before the vaccines are licensed for general use. Those studies would need to be large enough to detect rare serious adverse effects such as death or seizures. The speaker suggested the development of procedures for conducting such trials in an effective and efficient manner, noting that studies of this size have been done to evaluate agents used to reduce mortality following heart attacks.
A speaker raised two related questions: (1) Can the effects of individual vaccines be discerned even though vaccines are usually given in combination; and (2) Does the simultaneous administration of vaccines pose more risk than the administration of vaccine separately? A representative of a manufacturer stated that it is possible to answer these questions by designing a randomized trial in which, for example, one group is given three vaccines simultaneously, another group receives each of those vaccines at separate times, and other groups receive two of the three vaccines simultaneously. One could then assess whether there are any statistically significant differences in the risk of adverse events when these groups are compared. Such studies would be large, cumbersome, and expensive to run, he noted, but theoretically, they can be done.
One participant expressed concern that large prelicensure trials that take several years to conduct and analyze would place too much of a financial burden on vaccine manufacturers and could delay the availability of new vaccines. An epidemiologist commented that large-scale clinical trials, although potentially valuable, would not obviate the continued need for systems such as VAERS to detect the adverse effects of the vaccines. Even if a vaccine is shown to be safe in a large clinical trial, the potential for lot and batch variability and for variability in the use of vaccines require continual surveillance for adverse events.
Several speakers suggested conducting more focused, powerful, larger, and better-designed postlicensure clinical studies to detect serious , rare adverse events of vaccines that are not possible to detect in a clinical trial. FDA continues to work with manufacturers in this regard but recognizes that such studies, without randomized control groups, will be far less valid than the randomized trials that can be done prelicensure.
A vaccine manufacturer's representative suggested that each manufacturer should contribute money to a pool that could be used to establish a state-of-the-art, large, linked database system rather than doing its own small-scale postmarketing studies to detect adverse effects from vaccines. This would create a standard system for the detection of rare adverse events, that might go undetected in the smaller studies conducted by individual companies. Manufacturers could pay a maintenance fee for receiving data from such a large linked database system. To improve further the detection of rare adverse events, a vaccine manufacturer's representative suggested establishing an international network of databases of adverse events following vaccination. Such a network would require strict standardization of the information entered into it, however.
Noting that VAERS primarily receives reports of adverse events that occur shortly after vaccination, several speakers suggested developing procedures to detect longer-term or more subtle adverse events that might be associated with vaccination, such as learning disabilities or chronic fatigue syndrome. It was suggested that retrospective studies, as well as prospective studies that use nonvaccinated individuals as controls over a period of 10 or 20 years, be conducted for this purpose. A consumer representative suggested that there be more government funding for studies designed to detect long-term adverse effects of vaccines or to identify individuals at high risk for adverse effects. It was also suggested that children given vaccines in clinical trials be followed for 5 to 10 years to help assess long-term adverse events that do not become apparent until children reach later stages of development. Other participants commented that such studies would be extremely difficult to design, especially if a vaccine comes into general use so that it is given to large numbers of children.
Several speakers suggested that more studies be done using LLDB or other methods to compare vaccinated children with unvaccinated children. There is concern, however, that children who are not vaccinated are different in important ways from those who are vaccinated and therefore are not a suitable control population. To avoid the problem of controls who differ from cases in unmeasured ways, a CDC representative described a method that uses cases as their own controls (Farrington et al., 1996). He illustrated this methodology with a hypothetical example, in which the adverse events and vaccinations that an infant experienced during each month are monitored, and then applied this methodology to seizures after vaccination with the measles-mumps-rubella vaccine reported in CDC's LLDB study. By comparing the rates of occurrence of adverse events during periods immediately following vaccination and at all other times, the risk of adverse events with short latencies can be determined.
To better separate the adverse effects of a vaccine from background adverse events that occur in the age group vaccinated, a vaccine researcher suggested that different vaccine schedules be used for different groups of children in a random manner. This would not be unethical, the speaker noted, if the schedules were devised such that some children received a vaccine only a month or two later than other children.
To improve assessment of the frequency of occurrence of rare adverse events, a vaccine manufacturer's representative suggested that more effort be made to improve general surveillance of specific disorders that could potentially be associated with vaccination, such as Guillain-Barré syndrome or transverse myelitis. The likelihood of these disorders occurring as a result of vaccination