Data monitoring committee (DSMB, IDMC)
A data monitoring committee is a small group of independent doctors and statisticians who see the unblinded results while a trial is running and tell the sponsor to stop, change or continue, so that neither patients nor the sponsor are exposed to what the data are showing before the trial is done.
Overview
Once a randomised trial starts, someone has to watch the accumulating results for harm and for overwhelming benefit, yet the sponsor and the investigators must not see them, or they would change how they treat, recruit and report. The data monitoring committee, called a data and safety monitoring board in the United States, resolves this. Typically three to seven members, clinicians and at least one statistician, none with a stake in the result, meet at planned intervals, review unblinded safety and efficacy data prepared by an independent statistician in a closed session, and issue a recommendation: continue unchanged, modify (for example, stop enrolling a subgroup), stop for futility, or stop or unblind for efficacy or harm. The committee's charter, written before the first patient, sets the schedule, the stopping boundaries it will use and how it communicates. Everyone else, including the trial steering committee, sees only pooled or blinded data until the trial ends.
The corpus shows the committee at work in every direction. ADAURA was unblinded early on its independent data monitoring committee's recommendation because the disease-free survival benefit of adjuvant osimertinib was far beyond the pre-specified boundary. JAVELIN Head and Neck 100 and TrilynX were stopped for futility when interim analyses showed the experimental arm could not win, TrilynX with worse event-free survival on the experimental drug. MAGNITUDE's committee closed the marker-negative cohort for futility while the BRCA cohort continued to a positive result. DREAM3R was stopped early without meeting its endpoint. In each case the committee acted on boundaries fixed in the statistical analysis plan, not on a judgement made in the moment.
The committee is only as independent as its structure. Members should have no financial ties to the sponsor, the unblinded statistician should sit outside the sponsor's team, and recommendations, not data, should flow to the sponsor. Adaptive and platform trials put more weight on the committee because more decisions are taken during the trial. Single-arm and early-phase trials often use a lighter safety review committee instead. When a paper says a trial was stopped early, the questions to ask are who recommended it, against what pre-specified rule, and what happened to the endpoints the trial no longer had time to measure.
Similar pages
not linked directly; found by shared links- TermStatistical power, sample size and re-estimation
Shares DREAM3R, Futility analysis (stopped for futility), Group sequential design, stopping rules and alpha spending, Interim analysis, readout and data cut-off.
- TermPrimary, secondary and co-primary endpoints
Shares Futility analysis (stopped for futility), Trial protocol and statistical analysis plan, Group sequential design, stopping rules and alpha spending, Interim analysis, readout and data cut-off.
- TermAccrual and enrolment
Shares Futility analysis (stopped for futility), Ethics review (IRB, research ethics committee), Trial protocol and statistical analysis plan, Trial lifecycle: from protocol to label.
- TermTrial registration and results reporting (ClinicalTrials.gov, EU CTR)
Shares Ethics review (IRB, research ethics committee), Trial protocol and statistical analysis plan, Trial lifecycle: from protocol to label, Phase 1, 2 and 3 trials.
- TermInformed consent
Shares Ethics review (IRB, research ethics committee), Clinical equipoise, Trial lifecycle: from protocol to label, Phase 1, 2 and 3 trials.
- TermFirst-in-human (FIH) trial
Shares Ethics review (IRB, research ethics committee), Trial lifecycle: from protocol to label, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials.
- TermStratified randomisation, allocation concealment and minimisation
Shares Response-adaptive randomisation, Clinical equipoise, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials.
- TrialNAVIGATE
Shares Clinical equipoise, Trial lifecycle: from protocol to label, Phase 1, 2 and 3 trials.