Another source of bias when comparing the toxicity of drugs is the non-coherence of denominators. It is therefore crucial to ensure that under the null hypothesis of a non-difference in toxicity, the risk of an adverse event is expected to be the same for the two drugs compared.
A typical example relies on different durations of exposure. Let us take the example of anaphy-lactic reactions which, by definition, are expected to occur at initiation of the treatment. Ninety-two reactions were reported for drug A and 242 for drug B. On the basis of a total exposure (estimated from sales statistics) of 2.8 million months and 1.9 million months, respectively, the risk appears to be 3.9 times greater for drug B. However, this conclusion could be reversed if the number of first users was much greater for this drug because of shorter duration of treatment. It could happen, even if the two drugs belong to the same pharmacological and/or therapeutic class. For instance, such differences are observed with analgesics or NSAIDs: for more or less obscure reasons, some are used chronically when others are preferred for the treatment of acute pain. Therefore, it is a safe practice, before any decision-making, to try to obtain relevant informations on the utilization patterns of the drugs compared in order to avoid gross misinterpretation errors.
Was this article helpful?