*Geek Box: Intention to Treat Analysis
In a randomised controlled trial, you want to match both arms of the trial to ensure that one side doesn’t influence (i.e., bias) the results more than the other. This can be a problem if there is, for example, a high drop-out rate in one arm of the trial; the other arm will then have more statistical power, and it may over-inflate the effect of that arm vs. the comparative arm (or the control, if it is a control arm).
Intention to treat [ITT] is where the researchers will conduct analysis as if all subjects randomised in the trial completed it, irrespective of whether they dropped out, or didn’t comply with the protocol. Drop-out and noncompliance are two issues which face many trials, in particular nutrition and weight loss interventions. True intention to treat analysis requires complete data to be available for all subjects who didn’t complete the trial according to protocol.
However, that is not always available, and so often researchers will make assumptions based on, for example, a last data point or a baseline measurement (for example, in this study for dropouts in the maintenance phase the researchers assumed that subjects returned to their baseline weight for ITT analysis – not unreasonable given the weight regain in subjects who completed the trial). Intention to treat is a positive because it maintains the sample size, and it assumes a real-world practicality – because in the real world, not everyone is compliant with a protocol (as any practitioner knows!).