Do measles vaccines have non-specific effects on mortality?

Abstract

Editor--Cooper et al. (1) reviewed the non-specific effects on mortality of childhood vaccines. Although there are numerous studies detailing mortality following measles vaccination (MV) (2), only two studies cited in their article (3, 4) satisfied the authors' methodological criteria. The Zaire (3) and Bangladesh (4) studies compared mortality in areas with MV to adjacent areas without MV. The MV-associated relative mortality reductions were 31% and 46% and the absolute reductions were 2.1% and 1.8%. Since the case-fatality rate was assumed to be 2-4%, Cooper et al. concluded that there was insufficient evidence to suggest a mortality benefit above that caused by the prevention of measles infection. Hence, they questioned our non-specific effects hypothesis (2) after having excluded almost all MV studies. However, even the two studies retained (3, 4) in the Cooper ct al. article support the existence of nonspecific effects. First, Cooper et al. have not analysed the two studies in the same manner. In the Bangladesh study (4), measles-vaccinated children were compared to measles-unvaccinated children, whereas in the Zaire study (3), only 83% of children in the vaccinated area received MV. If vaccinated children from the vaccinated area are compared with unvaccinated children from the adjacent area, the relative mortality reduction was 48% (3, Table 1) giving an absolute reduction of 3.0%. Children in Bangladesh were vaccinated between 9 and 60 months of age; the difference in the proportion of children who died was 1.8% and it was this proportion which was used to indicate the absolute reduction. However, if accumulated mortality is used to estimate the absolute reduction as in Zaire (3), the absolute reduction seen in Bangladesh would be around 3.9% (4, Fig. 3). The relative (48%, 46%) and absolute 3.0%, 3.9%) reductions in the Zaire and Bangladesh studies were similar to the estimates seen in the less methodologically rigorous studies (1, 2). Interestingly, all of the study designs have yielded similar estimates (2-6), including: studies of mortality before and after the introduction of MV (2), blind studies with ineffective vaccine (2), and randomized studies (2, 6). Second, the assertion that the absolute reduction corresponds to measles-associated mortality is not supported by any study (2, 5, 6). Cooper et al. claim that MV-associated mortality differences were not examined in areas with concurrent morbidity and mortality surveillance (1). In fact, we reanalysed the Bangladesh study to determine the MV-associated mortality reduction that could be explained by the prevention of measles infection (5); surprisingly, prevention of measles infection accounted for very little of the reduction. When measles cases were censored in the survival analysis, the relative reduction changed merely from 49% to 43%. Mortality was lower after measles infection than among measles-uninfected children (2, 5). (There were no similar data from Zaire.) In Zaire, the difference in accumulated measles incidence for vaccinated and unvaccinated children was 25%, with a case-fatality rate of 7%, indicating that the measles-associated mortality difference would be less than 2% before 5 years of age. …