How does an independent data monitoring committee protect a trial during interim analyses?

Why

An interim analysis compares arms before the trial ends. ICH E9 says interim analyses should be planned in the protocol, that stopping boundaries for efficacy should usually demand more evidence than those for safety, and that overall type I error must stay controlled. The danger is knowledge.

If investigators learn the drug looks good, they may enroll differently, treat differently or assess outcomes differently, and none of that can be fixed in the analysis. So an independent committee of clinicians and a statistician, working under a charter, reviews unblinded data, often prepared by an independent statistical group. Investigators are told only the decision.

FDA's draft guidance on data monitoring committees, issued in February 2024 to replace its 2006 guidance once final, describes when such a committee is useful, for example in trials with mortality or serious morbidity endpoints. Take a cardiovascular outcome trial where the committee sees a hazard ratio of 0.70 at the first look but the boundary is set at a very small p value. It recommends continuing, and nobody outside learns the number.