COVID-19 vaccines: effectiveness and number needed to treat.

In a Lancet Microbe Comment, Piero Olliaro and colleagues suggest that reporting relative risk reduction (RRR) for vaccination does not reflect entirely its therapeutic performance and consider the solw use of RRR a reporting bias. In addition, they propose that absolute risk reduction (ARR) should be reported as a measure of the vaccine's effectiveness. The authors end up comparing the numbers needed to vaccinate to prevent one case of COVID-19 among the vaccines, which derives from the absolute reductions.

However, this suggestion might have a paradoxical effect in misleading perception of treatment performance. This approach disregards three epidemiological facts.

First, number needed to treat (NNT) is not an intrinsic property of a treatment, it is rather a property of the population that receives a treatment: for a constant relative risk reduction, populations of different baseline risks will have different absolute reductions. Therefore, NNT comparison of different treatments across studies should be avoided, because sample populations will always have baseline risk variations. Indeed, this approach is the actual reporting bias.

Second, the authors raise a concern that different levels of background risk might change relative risk reduction of studies. This statement disregards the constant property of relative risk repeatedly demonstrated by subgroup analysis of clinical trials and meta-scientific evaluations of a treatment across studies of different baseline risks. For example, statins,3, 4 anti-hypertensive therapy, and aspirin have the same relative risk reduction across the baseline risks of primary or secondary prevention.

Finally, effectiveness—a real-world property—is about clinical decision making, and not to be derived from efficacy studies (randomised controlled studies). As a clinician or an epidemiologist, one should multiply the RRR (intrinsic property of a treatment) by the baseline risk of a given population or patient, individualising the ARR and NNT. They are not scientific concepts, they are circumstantial information.

We declare no competing interests.