in those who had been vaccinated. This might be particularly relevantin studying adverse events with long latencies from the time of vaccination.There was concern by some that unvaccinated individuals are so differentfrom vaccinated individuals that these differences cannot be controlledfor; that is, completely unvaccinated individuals might possess otherhealth or socioeconomic traits if they are from a population thatdoes not believe in traditional medical treatments, or they mightrepresent a limited gene pool if the population consists of a longstanding,religious sect. One participant argued that comparison with unvaccinatedcontrols might make vaccines look more beneficial than they are,because those who are not vaccinated are more likely to be underprivilegedor have more diseases or disorders for other reasons. The decreasedincidence of SIDS following immunization with DPT is an example ofthis (Institute of Medicine, 1991). Others noted that there are individualswho would be willing to participate, or have their children participate,as unvaccinated controls, and it was suggested that statistical methodscould be used to address the problem of the self-selection of thegroup. Still others believed that the number of unvaccinated subjectsis too low for detection of rarely occurring adverse events.
One participant noted that religious groups that decline vaccinationhave occasionally been studied during acute outbreaks of diseaseand speculated that it might be possible to work with them in attemptingto find the background rates of particular adverse events in unvaccinatedchildren. It was noted that the possible problem of a restrictedgene pool would have to be taken into consideration, particularlyfor certain adverse events.
The need for more complete information on the background incidencerates of the adverse events was discussed. It was suggested thatvital statistics data or other studies of infant mortality be exploitedfor ecologic correlations, that is to look for changes in mortalityin periods before and after new vaccines are introduced. Some participantsthought that this was a fairly crude approach to the subject andthat it should be looked at in great depth. It was noted that a vaccinewould have to account for a large proportion of deaths (and thereis no reason to believe that any vaccine does) to show a differenceusing crude mortality data. In addition, one participant thoughtthat there would be too much seasonal and other variation for theinfant mortality approach to be helpful unless a concurrent, controlledstudy were to be done.
It was noted that the question of background incidence rates wasalso of concern to the IOM vaccine-adverse event committees (Instituteof Medicine, 1991, 1994a). It may be possible to derive reasonablyaccurate rates for such conditions as GBS and multiple sclerosisin adults, but when considering such conditions as encephalopathyand encephalitis in children, the conditions are