I am currently analysing the findings of a pilot RCT that is exploring the use of a psychological intervention. Obviously, there are dropouts and participants that dropped out from the trial before receiving the intervention. Apparently 'gold standard' analysis for RCTs is ITT. ITT does not take into account dropouts and people withdrawing after randomisation. What about for a pilot RCT? I am thinking to do PP analysis in addition to ITT. Any experiences and suggestions?
How about all of the above? All give different information, and can be useful.
I agree with the previous answer. In general a pilot study has lower statistical power and more selected patient population in comparison with confirmatory or pivotal trials. The major issue is how much a pilot study is predictive of efficacy and safety in a larger patient population. I would decide the primary analysis (e.g. a PP for a pilot study may be preferred as the actual treatment effect is quantified) and then the other two analysis (ITT and as treated) that can provide different useful information and possibly add robustness to the conclusions. But ITT could be choosen as primary analysis as well. The decision on the type of primary analysis could be based on the type, size and end point(s) of the pilot study.
If ITT is choosen the method of inputation of missing data iis also important to be defined and justified (e.g. LOCF, BOCF, MMRM, etc)
The papers at the link are a nice reading on ITT
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3159210/
http://meds.queensu.ca/medicine/obgyn/pdf/ebm/intention.to.treat.pdf
ITT analysis assumes importance when the dropouts/withdrawals happen after the subject has taken at least one dose of the study drug and there is at least one post-treatment observation to enable LOCF. In such a situation, ITT analysis would also be relevant from a pharmacovigilance perspective. If the situation is different from the above, perhaps per per protocol analysis can be done.
- Badges
- Science topic