Miguel AHernán, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115miguel_hernan@post.harvard.edu
Susan S. Jick
Submitted November 11, 2004
We welcome the reanalysis by DeStefano et al which not only
constitutes a methodologic improvement, but also allows for a more direct
comparison with our estimates. Unfortunately, the stricter criteria used
in the reanalysis resulted in period-specific estimates with confidence
intervals too wide to draw firm conclusions. This sample size problem
would have only been aggravated had the authors mimicked our eligibility
criteria more closely by restricting their cohort to individuals who were
members of the health maintenance organizations for at least, say, three
years before the start of follow-up (see below).
MacIntyre et al do not consider ours to be a prospective study. We
believe this is a semantic disagreement. In our study, the exposure
information was collected before first symptoms of MS. This key feature of
fully prospective investigations prevents recall bias. Whether the
investigators made the decision to conduct this study in 1990 or in 2000
is irrelevant: the information was collected in the same way and would not
change because of the timing of the decision. The hypothesis that
intravenous drug use and sexual practices are important risk factors for
MS needs to be tested given its potential consequences for MS research.
MacIntyre et al are concerned about our exclusion of cases who had MS
before the start of follow-up. To understand the rationale for this
exclusion, it is helpful to consider two cohorts with different
eligibility criteria: (a) all individuals without MS at start of follow-
up; (b) individuals without MS at start of follow-up who had at least
three years of computerized information. An individual’s follow-up starts
at her first computer record in cohort (a), and three years after her
first computer record in cohort (b). All individuals in (b) had the
opportunity to have their exposure recorded during the three-year period
prior to the start of follow-up, thus reducing bias from between-subject
differences in exposure misclassification. Of the 438 MS cases, 282 in (a)
and 163 in (b) occurred after the start of follow-up; the odds ratios
estimated from case-control studies nested in these cohorts were 2.7 for
(a) and 3.1 for (b).1 As a further clarification, our exposure information
came from the computerized records only. Paper medical records were used
to confirm the cases’ diagnosis and to determine their date of first
symptoms.
Finally, MacIntyre et al’s questions about the time between HB
vaccination and MS onset are crucial. Future research efforts should be
directed towards answering them.
1. Hernán MA, Jick SS, Olek MJ, Jick H. Recombinant hepatitis B
vaccine and the
risk of multiple sclerosis: a prospective study. Neurology 2004; 63:838-
42.
We welcome the reanalysis by DeStefano et al which not only constitutes a methodologic improvement, but also allows for a more direct comparison with our estimates. Unfortunately, the stricter criteria used in the reanalysis resulted in period-specific estimates with confidence intervals too wide to draw firm conclusions. This sample size problem would have only been aggravated had the authors mimicked our eligibility criteria more closely by restricting their cohort to individuals who were members of the health maintenance organizations for at least, say, three years before the start of follow-up (see below).
MacIntyre et al do not consider ours to be a prospective study. We believe this is a semantic disagreement. In our study, the exposure information was collected before first symptoms of MS. This key feature of fully prospective investigations prevents recall bias. Whether the investigators made the decision to conduct this study in 1990 or in 2000 is irrelevant: the information was collected in the same way and would not change because of the timing of the decision. The hypothesis that intravenous drug use and sexual practices are important risk factors for MS needs to be tested given its potential consequences for MS research.
MacIntyre et al are concerned about our exclusion of cases who had MS before the start of follow-up. To understand the rationale for this exclusion, it is helpful to consider two cohorts with different eligibility criteria: (a) all individuals without MS at start of follow- up; (b) individuals without MS at start of follow-up who had at least three years of computerized information. An individual’s follow-up starts at her first computer record in cohort (a), and three years after her first computer record in cohort (b). All individuals in (b) had the opportunity to have their exposure recorded during the three-year period prior to the start of follow-up, thus reducing bias from between-subject differences in exposure misclassification. Of the 438 MS cases, 282 in (a) and 163 in (b) occurred after the start of follow-up; the odds ratios estimated from case-control studies nested in these cohorts were 2.7 for (a) and 3.1 for (b).1 As a further clarification, our exposure information came from the computerized records only. Paper medical records were used to confirm the cases’ diagnosis and to determine their date of first symptoms.
Finally, MacIntyre et al’s questions about the time between HB vaccination and MS onset are crucial. Future research efforts should be directed towards answering them.
1. Hernán MA, Jick SS, Olek MJ, Jick H. Recombinant hepatitis B vaccine and the risk of multiple sclerosis: a prospective study. Neurology 2004; 63:838- 42.