Hepatitis B Vaccination and the Risk of Multiple Sclerosis

The study population for this investigation comprised participants in the large cohorts involved in two ongoing studies — the Nurses' Health Study and the Nurses' Health Study II. The Nurses' Health Study began in 1976, when 121,700 female registered nurses from 11 states, between 30 and 55 years old, responded to a mailed questionnaire about their history of diseases and their lifestyles. The Nurses' Health Study II began in 1989, when 116,671 female registered nurses from 14 states, between 25 and 42 years old, responded to a similar questionnaire. Follow-up questionnaires are mailed to the participants in both studies every two years. A specific question on the lifetime occurrence of multiple sclerosis was first included in the 1992 questionnaire for the Nurses' Health Study and in the 1991 questionnaire for the Nurses' Health Study II. It asked, “Have you ever had a diagnosis of multiple sclerosis made by a physician?” A question about the diagnosis of multiple sclerosis within the previous two years was included in subsequent questionnaires. In the Nurses' Health Study, most cases of multiple sclerosis diagnosed before 1992 had already been reported through an open-ended question about “other major illnesses.” Women who had received a diagnosis of multiple sclerosis before enrollment in the cohort (i.e., base line) were excluded from our study.

Details of the follow-up of patients with multiple sclerosis in these cohorts have been reported previously.⁹ Briefly, we wrote to the women who reported a new diagnosis of multiple sclerosis to obtain permission to contact their neurologists and review their medical records. After obtaining permission, we sent the neurologists a questionnaire regarding the certainty of the diagnosis (definite, probable, or possible multiple sclerosis or not multiple sclerosis), the date of onset of neurologic symptoms related to multiple sclerosis, other aspects of the clinical history, and results of relevant laboratory tests. If a neurologist had not been involved in the case or did not respond to the questionnaire, we mailed the questionnaire to the woman's internist. We confirmed the diagnoses made by these physicians by applying the criteria established by Poser et al. for the diagnosis of multiple sclerosis to the clinical and laboratory data provided in each questionnaire.¹⁰ Since 93 percent of all definite and probable diagnoses were confirmed by the application of these criteria, we classified women who had a diagnosis of definite or probable multiple sclerosis according to their physicians as having the disease. The date of onset of the disease was determined by asking both the women and their physicians for the date of the first neurologic symptoms. The earlier of the two dates was used in the analyses. Women with multiple sclerosis that began during the follow-up period but remained undiagnosed by April 1998 would not have been included in our analyses as women with multiple sclerosis. However, this exclusion is likely to be independent of the history of vaccination and is therefore unlikely to have biased the results of our study. The age-specific incidence of multiple sclerosis in these cohorts has been reported previously⁹; on the basis of those rates, we estimated the lifetime risk of multiple sclerosis as 4.9 per 1000 — a risk similar to that reported in other prospective investigations.⁹

A total of 318 newly diagnosed cases of definite or probable multiple sclerosis were documented between the base-line questionnaire and April 1998 and were included in the study. For each woman with multiple sclerosis, we randomly selected as controls five women with no history of multiple sclerosis or breast cancer (healthy controls) and one woman with breast cancer, who were matched to the woman with multiple sclerosis according to year of birth, study cohort, and (for the controls with breast cancer) date of diagnosis. Women with breast cancer were included as controls so as to address the potential bias that may derive from differential recall among women with a serious disease. Two of the 1590 randomly selected healthy controls and 1 of the 318 controls with breast cancer were deemed ineligible because of self-reported but unconfirmed multiple sclerosis and were excluded from further consideration.

History of immunization with the hepatitis B vaccine was initially assessed by means of a questionnaire. Women who reported that they had never been vaccinated against hepatitis B were considered unvaccinated. Women who reported having received the vaccine at any time in the past were asked for permission to obtain their vaccination records from their employers. The initial questionnaire included questions regarding childhood infections and vaccinations for measles and tetanus. The cover letter that was sent with the questionnaire did not make specific reference to the hepatitis B vaccine, since it was only one of several types of exposure addressed by the study. The questionnaire was returned by 301 of the 318 women with multiple sclerosis (95 percent), 1393 of the 1588 healthy controls (88 percent), and 280 of the 317 controls with breast cancer (88 percent). Among the respondents, all 301 women with multiple sclerosis had at least 1 matched healthy control (a total of 1317 matched healthy controls), and 263 had a matched control with breast cancer.

According to their own reports, 51.8 percent of the women with multiple sclerosis, 66.5 percent of the healthy controls, and 64.6 percent of the controls with breast cancer had been vaccinated against hepatitis B at some time in the past. After obtaining permission, we contacted the employers of these women to obtain a copy of their vaccination records. We sent up to three mailings and telephoned the nonrespondents. Vaccination records were obtained for 62 percent of the women with multiple sclerosis, 64 percent of the healthy controls, and 64 percent of the controls with breast cancer. The records obtained confirmed that the hepatitis B vaccine had been administered in more than 94 percent of the women with multiple sclerosis and the controls. However, the self-reported dates of vaccination were often inaccurate. For this reason, we excluded from the primary analyses women who reported having been vaccinated but for whom we could not obtain vaccination certificates. Those excluded for this reason included 106 of the 301 women with multiple sclerosis (35.2 percent), 466 of the 1317 healthy controls (35.4 percent), and 96 of the 263 controls with breast cancer (36.5 percent). Among the women with multiple sclerosis excluded were 94 of the 263 with a matched control with breast cancer (35.7 percent). As a last step, we excluded those who remained unmatched (5 women with multiple sclerosis and 317 controls for analyses using healthy women as controls; and 58 women with multiple sclerosis and 56 controls for analyses using women with breast cancer as controls). Thus, the final study population included 192 women with multiple sclerosis; 190 of these women were matched to 534 healthy controls, and 111 were matched to an equal number of controls with breast cancer. The diagnosis of multiple sclerosis was made by a neurologist in 93 percent of the 192 women with multiple sclerosis included in the analyses, and the diagnosis was supported by a positive result on magnetic resonance imaging in 86 percent. These proportions increased to 96 percent and 97 percent, respectively, for women with an onset of multiple sclerosis after 1987 (when a recombinant hepatitis B vaccine was introduced).

Other data obtained from the women included age, latitude of birthplace, ancestry (Scandinavian, southern European, other white, or nonwhite), pack-years of smoking (none, <10, 10 to 24, or ≥25), and (for the Nurses' Health Study II only) type of nursing job (inpatient, outpatient, emergency room, nursing education, or other). The associations of birthplace latitude, ancestry, and smoking history with the risk of multiple sclerosis in these cohorts have been previously reported.^9,11 The type of job was included because of its potential relation to both hepatitis B vaccination and the risk of multiple sclerosis.

The women with multiple sclerosis were classified as having been exposed or not exposed to the hepatitis B vaccine according to their vaccination status at the time of the onset of multiple sclerosis. For the controls, we used the time of the onset of multiple sclerosis in the woman with whom the control was matched. For both women with multiple sclerosis and controls, we refer to this date as the index date. Two categories of exposure were used in the main analyses: receipt of at least one dose of hepatitis B vaccine at any time before the index date, and receipt of the first dose of the vaccine within two years before the index date. All the analyses were conducted through comparisons of the women who had multiple sclerosis with their matched controls, unless otherwise indicated. The relative risk of multiple sclerosis for women who had been vaccinated against hepatitis B as compared with those who had not been vaccinated was estimated with the use of conditional logistic regression. We used a period of two years in categorizing exposure because analyses based on shorter periods of exposure would have been more sensitive to errors in the estimation of the date of the onset of multiple sclerosis. To assess the presence and magnitude of recall bias, we conducted further analyses using the self-reported dates of vaccination.