Following the introduction of plasma-derived hepatitis B vaccine in 1982, a passive surveillance effort was initiated by the Centers for Disease Control (CDC) to monitor for all serious adverse events. A study of the neurologic adverse events reported in approximately 850,000 vaccinees during the first 3 years of surveillance was published in 1988 (Shaw et al., 1988). Nine cases of putative GBS occurring after administration of hepatitis B vaccine came to attention, all in adults. The clinical information for eight cases was reviewed independently by four academic neurologists. They expressed a wide range of opinions as to whether these cases represented GBS. Two of the nine cases were judged to be definite GBS by three of the four reviewers, two cases received two of four votes as definite GBS, one case was thought to be definite GBS by one of the four reviewers, and three cases were not thought to be definite GBS by any of the reviewers.
Although the neurologists did not all agree that each of the nine cases was GBS, the authors used all nine cases in the analysis. Because no concurrent control populations were available, two population-based studies were used to calculate expected numbers of GBS cases for comparison purposes. One set of background incidence data came from a CDC study designed to evaluate the relation between GBS and swine flu vaccination, which presumably used case definition methods similar to those that led to the nine GBS cases in the study of Shaw et al. (1988). The second set of background incidence rates came from a linked medical records system conducted by the Mayo Clinic in Rochester, Minnesota, for Olmsted County, Minnesota. Relative risks for GBS following hepatitis B vaccination were calculated under a variety of assumptions, specifically, a 6- or 8-week at-risk interval and risk evenly distributed among three doses versus all risk associated with the first dose. Statistically significant increases in risk were found under all assumptions when the CDC data were used for comparison purposes, but only with a 6-week at-risk interval after the first vaccine dose when the Olmsted County data were used. Adjustments for age in the CDC data and age and sex in the Olmsted County data did not substantially change the results. The authors stated that "no conclusive epidemiologic association could be made between any neurologic adverse event and the vaccine" (Shaw et al., 1988, p. 337), presumably because their data derived from spontaneous reporting, they had no concurrent control information, and the diagnosis of GBS was sometimes suspect.
A recent uncontrolled observational study of 43,618 Alaskan native vaccinees used a different strategy to investigate the relation between plasma-derived hepatitis B vaccine and GBS (McMahon et al., 1992). A computer search for all GBS cases in hospitals to which these individuals could be admitted disclosed 10 patients with GBS during the period in which hepati-