In Reply

Many thanks to Dr. Traut for his contribution. As correctly noted, it is important to consider both absolute effects (absolute risk reduction [ARR], number needed to treat [NNT]) and relative effects (e.g. relative risk, relative risk reduction). For all outcomes, we reported relative estimates of effect; for all-cause pneumonia, invasive pneumococcal disease, and acute otitis media, for example, we calculated absolute effects in the form of NNT, from which one can easily deduce ARR (NNT = 1/ARR).

Whereas a relative measure (e.g. relative risk) generally remains stable across various risk groups (1), ARR underestimates the effect for patients with high baseline risk and overestimates the effect for patients with low baseline risk (2). The preferred strategy is therefore to perform a meta-analysis with a relative measure of effect size and then, by population and setting, to apply the relative treatment effect to a specific health risk within a population, as we did in our article for three outcomes, for example (1, 3). Some health risks vary greatly between different populations or settings. NNTs are generally smaller in high-risk populations and larger in low-risk populations. Vaccination is usually performed in low-risk populations that nevertheless typically include a large number of individuals. From a public health perspective, NNTs of 1500 or 2000 individuals to be vaccinated in order to prevent one case of disease therefore certainly do have great practical significance. Population-relevant issues such as herd effects (i.e. the fact that unvaccinated individuals also benefit from vaccination if the vaccination rate is high enough), which are often overlooked by opponents of vaccination, should also be a factor in primary care physicians’ decision-making processes.