*Geek Box: Per Protocol Analysis
You may recall from a previous Deepdive that we covered something known as ‘intention to treat’, which is a form of analysis which includes all participants that were randomised in the study, irrespective of whether they dropped out or not. To do this, the investigators take the last value recorded for that participant forward into the analysis, as if that value represented the data point the participant finished the trial with. Intention-to-treat is recommended in any trials comparing the effects of a treatment, because if the intervention and control arm are not balanced, it may lead to bias in the results. In contrast, a ‘per-protocol analysis’ is an analysis which only includes the data of participants who completed the study according to which intervention they were originally allocated to. This has the potential to introduce bias, if the arms of the trial are not balanced. In the present study, 24 were excluded from analysis in the LSFA arm, while 15 were excluded in the HSFA arm; the HFSA arm thus had 11 more participants in the final analysis. Whether this influenced the results, we don’t know, because the investigators did not do an intention-to-treat analysis alongside the per protocol analysis.