Two common ways data is analyzed in randomized controlled trials include the intention-to-treat (ITT) and per protocol (PP) principles. ITT analysis uses data from every participant who was randomized, regardless of whether they dropped out early or did not stick to the study regimen. A catchphrase for ITT is “once randomized, always analyzed”. PP analysis excludes participants who dropped out early or did not stick to the study regimen, so only the data from the participants who fully adhered to the trial protocol are included.^[1]

🔎 Example

20 people were randomized to take ingredient A or a matched placebo daily for 4 months, and blood pressure was measured every month. Three people from the ingredient A group and 2 people from the placebo group withdrew from the study at the two-month mark. At the end of the trial, ITT analysis will include all 20 participants, typically using the 2-month mark values for the 5 people who left the study. PP analysis will only include the 15 people who completed the full 4-month trial.

Why is ITT better than PP?

ITT analysis has less confounding than PP because randomization is preserved. Excluding participants from the analysis after randomization introduces bias because the groups are no longer balanced in terms of known and unknown characteristics. The analysis becomes confounded.^[1]

ITT is considered more pragmatic. It better reflects what would occur in real life, as some participants quit the trial early, others might not regularly submit questionnaires or attend trial sessions, while some don’t even start the study regimen. PP analysis is used for trials that want to measure the effect of a treatment under ideal experimental conditions, where it is not affected by nonadherence.^[2]

The risk of bias is increased whenever treatment groups are not analyzed according to the group to which they were originally assigned. If an intervention is truly effective, an intention-to-treat analysis will provide an unbiased estimate of the efficacy of the intervention at the level of adherence in the study. See this paper for a more detailed review of ITT versus PP analysis and to see how how using the wrong method of analysis can lead to a significantly biased assessment of the effectiveness of an intervention.

What is a drawback of ITT analysis?

The risk with ITT analysis is that an effective intervention might be interpreted as less effective than it is. If the intervention is effective when taken exactly as recommended, including participants who did not take the intervention or took it incorrectly underestimates the effect size. So the ITT analysis would show that the intervention is no better than a placebo, when under perfect experimental conditions, it is indeed effective. Still, the ITT analysis does show an accurate result in the case of imperfect use of the intervention — which is usually more akin to a real-world clinical outcome.^[1]