On sample size calculation in drug interaction trials

Abstract Drug–drug interaction (DDI) trials are an important part of drug development as they provide evidence on the benefits and risks when two or more drugs are taken concomitantly. Sample size calculation is typically recommended to be based on the existence of clinically justified no‐effect boundaries but these are challenging to define in practice, while the default no‐effect boundaries of 0.8–1.25 are known to be overly conservative requiring a large sample size. In addition, no‐effect boundaries are of little use when there is prior pharmacological evidence that a mild or moderate interaction between two drugs may be present, in which case effect boundaries would be more useful. We introduce precision‐based sample size calculation that accounts for both the stochastic nature of the pharmacokinetic parameters and the anticipated width of (no‐)effect boundaries, should these exist. The methodology is straightforward, requires considerably less sample size and has favorable operating characteristics. A case study on statins is presented to illustrate the ideas.