François Lamontagne and colleagues recently reported an association between receipt of pneumococcal vaccination and a lower risk of myocardial infarction with an adjusted odds ratio of 0.53.1 Paradoxically, they reported greater protection with more remote vaccination. If these results are to be considered an undistorted reflection of the true protection attributable to vaccination, the vaccine's effectiveness against myocardial infarction would be 47%.1,2 In other words, about half of incident myocardial infarctions would have to be attributed to Streptococcus pneumoniae infections, preventable simply through pneumococcal vaccination. Moreover, the authors' finding of an adjusted odds ratio of 0.33 for vaccination given 2 or more years previously means that up to two-thirds of myocardial infarctions should be preventable in just 2 years through universal pneumococcal immunization at middle age. Unfortunately, there is a much more plausible explanation for the authors' results.
Their results are strangely reminiscent of the approximately 50% reductions in all-cause mortality of elderly adults that were previously promulgated for influenza vaccination but more recently questioned under the lens of healthy-user bias and confounding by indication.3–5 In other words, immunization may be a marker for factors such as diet, lifestyle and exercise that are not documented in administrative databases of the kind used by Lamontagne and colleagues but that are known to be associated over the long term with risk of diseases including atherosclerosis and myocardial infarction.6 Moreover, the controls chosen by the authors (patients admitted to a surgical department for a reason other than myocardial infarction) are likely to be a healthier group generally than patients admitted to medical wards, as illustrated in their Table 1.1
Extreme caution is required when applying observational designs to administrative databases without full control of established but uncaptured covariates. Attributable risk (and its counterpart, vaccine effectiveness) provides a reality check that can help investigators place improbable findings into stark perspective. Consideration of attributable risk as an initial test of logic should be routine lest we draw phenomenal conclusions from controvertible evidence. Magic bullets are, after all, rare.
Footnotes
-
Competing interests: Danuta Skowronski has received 1 research grant from Sanofi Pasteur and 1 from GlaxoSmithKline. Gaston De Serres has received research grants from Sanofi Pasteur and GlaxoSmithKline. None declared for Naveed Janjua, Travis Hottes and David Patrick.