9.3.6  Multiple treatment attempts

Similarly, multiple treatment attempts per participant can cause a unit-of-analysis error. Care must be taken to ensure that the number of participants randomized, and not the number of treatment attempts, is used to calculate confidence intervals. For example, in subfertility studies, women may undergo multiple cycles, and authors might erroneously use cycles as the denominator rather than women. This is similar to the situation in cluster-randomized trials, except that each participant is the ‘cluster’. See methods described in Chapter 16 (Section 16.3).