Loading
A trial in which a coin toss (done by computer) decides which treatment each patient gets, so the groups are alike in every way except the treatment. It is the only reliable way to prove a treatment causes a benefit.
Without randomisation, patients who receive a new drug tend to differ from those who do not (fitter, younger, treated at better centres), and those differences, not the drug, may explain a better outcome; randomising removes this bias, and analysing every patient in the group they were assigned to (intention-to-treat) preserves it. The control group receives the current standard of care, sometimes with a placebo added so that the two arms look identical, and the trial is powered so that a real difference of a given size is unlikely to be missed. Large randomised trials are slow and expensive, which is why surrogate endpoints, single-arm accelerated approvals and real-world evidence are all attempts to shortcut them, each with known pitfalls.
Shares Stepped-wedge design, N-of-1 trial, SMART design (sequential multiple assignment randomised trial), Cluster-randomised trial.
Shares Blinded trial, Intention-to-treat (ITT) and per-protocol analysis, N-of-1 trial, Stratified randomisation, allocation concealment and minimisation.
Shares P-value, Confidence interval, Endpoint, Kaplan-Meier curve, censoring and proportional hazards.
Shares Intention-to-treat (ITT) and per-protocol analysis, Clinical equipoise, Pragmatic trial, Non-inferiority trial.
Shares N-of-1 trial, Estimands and intercurrent events (ICH E9(R1)), Clinical equipoise, Kaplan-Meier curve, censoring and proportional hazards.
Shares Confidence interval, Kaplan-Meier curve, censoring and proportional hazards, CHALLENGE (CCTG CO.21), ADAURA.
Shares P-value, Confidence interval, Non-inferiority trial, Why trials fail: underpowered, wrong endpoint, control arm drift, subgroup fishing, crossover.
Shares Cluster-randomised trial, Registry-based randomised trial, Clinical equipoise, Placebo.