Loading
A SMART trial randomises patients once at the start and again at a decision point, such as after the first scan or a residual disease test, to compare whole treatment strategies rather than single drugs.
Most trials compare drugs. Doctors, however, follow strategies: start with A, and if the scan at three months is poor switch to B, otherwise continue. A sequential multiple assignment randomised trial is built to compare strategies. Patients are randomised at the start; at a pre-defined decision point (response, residual disease, toxicity) those in a given state are randomised again between the next options. The trial can then estimate the outcome of each embedded strategy (for example, A then B on poor response) and compare them, which a series of separate trials cannot do without assuming the pieces add up.
The corpus contains the ingredients of the design even where the label is not used. CAPTIVATE randomised patients who reached undetectable measurable residual disease after fifteen months of ibrutinib and venetoclax to stop treatment or continue it, a second randomisation triggered by response, and found no progression-free survival difference at three years. PERSEUS embedded a stopping rule for daratumumab after sustained residual-disease negativity in the experimental arm. IMvigor011 randomised only patients whose blood test for leftover cancer turned positive, and DYNAMIC and CIRCULATE-Japan use the same test to decide who is randomised to what after colon cancer surgery. These are response-adapted strategy questions, and a full SMART is the design that answers them without confounding by who happened to respond.
SMARTs need larger numbers than a single comparison because patients are split at each stage, the decision rule must be operational at every site (the same test, the same timing, the same threshold), and the analysis is of strategies, so the result does not say which single drug is best. They are common in behavioural and supportive care research and rare in oncology drug development, where each stage tends to be run as a separate trial. That is a gap the design could fill, especially for de-escalation questions where the answer depends on early response.
Shares Response-adaptive randomisation, Enrichment and biomarker-stratified designs, IMvigor011, Randomised trial.
Shares Randomised trial, De-escalation, escalation and response-adapted therapy, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials.
Shares Randomised trial, De-escalation, escalation and response-adapted therapy, DYNAMIC, Seamless, adaptive and Bayesian trial designs.
Shares Enrichment and biomarker-stratified designs, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials, Minimal / molecular residual disease (MRD).
Shares CAPTIVATE, PERSEUS, IMvigor011, Phase 1, 2 and 3 trials.
Shares IMvigor011, DYNAMIC, Minimal / molecular residual disease (MRD), Trial design, endpoints and cost.
Shares Response-adaptive randomisation, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials, Trial design, endpoints and cost.
Shares Randomised trial, Seamless, adaptive and Bayesian trial designs, Phase 1, 2 and 3 trials.