In the previous 2 articles, I discussed multiplicity in the context of subgroup analyses, multiple treatments, and multiple outcomes. In this article, I will discuss multiplicity in the context of repeated measurements.
It is common in clinical trials for an outcome to be recorded at multiple time points per treatment group. For example, Fleming et al evaluated the differences in the pain experienced during removal and insertion of orthodontic archwires between SmartClip self-ligating brackets and conventional Victory brackets. After appliance placement and engagement of a 0.016-in nickel-titanium archwire, pain experience was recorded after 4, 24, and 72 hours, and 7 days with a visual analog scale questionnaire. At a subsequent visit, the participants documented pain experiences during removal and insertion of 0.019 × 0.025-in stainless steel archwires on an additional 100-mm visual analog scale questionnaire ( Table I ).
|4 hours||24 hours||72 hours||7 days||Archwire removal||Archwire insertion|
Such designs invoke the problem of multiplicity because multiple comparisons might be performed from data collected at various time points between treatment groups. Data from repeated measures can be analyzed with different approaches; however, some methods could have important shortcomings. Here I will introduce and explain briefly several common methods for analyzing such data and state their main advantages and disadvantages ( Table II ).