A multi-arm platform trial testing four COVID-19 treatments starts with equal randomization (25% per arm). After an interim analysis, Arm C shows no benefit and is dropped, with its allocation redistributed to the remaining arms. Why is this more efficient than running four separate two-arm trials?
AIt uses a smaller total sample size — the shared control arm serves all comparisons, and patients are never allocated to a treatment known to be futile
BIt produces larger treatment effects
CIt eliminates the need for a control group
DIt is faster only because it uses a single site
Platform trials gain efficiency in two ways. First, the shared control arm serves all experimental arms simultaneously, requiring fewer total control patients than four separate trials each with their own control group. Second, dropping ineffective arms early prevents further enrollment to futile treatments, redirecting patients to more promising arms or new candidates. The RECOVERY trial (COVID-19) demonstrated this approach, rapidly identifying dexamethasone as effective and dropping hydroxychloroquine as ineffective within a single adaptive infrastructure.
Question 2 Multiple Choice
An adaptive trial re-estimates the sample size at an interim analysis based on the observed treatment effect. If the observed effect is smaller than originally assumed, the trial enrolls more patients. Why does this require careful statistical handling?
AIncreasing the sample size always inflates the Type I error
BThe interim data used for re-estimation are also used in the final analysis, creating a dependency that can inflate Type I error if not properly accounted for in the test statistic
CIt is unethical to extend a trial beyond the original sample size
DThe sample size increase makes the trial less powerful
The statistical challenge is that the decision to increase the sample size is based on the interim treatment effect estimate, which also contributes to the final test statistic. If handled naively, this creates a positive bias — the trial is more likely to continue (and recruit more patients) when the interim trend is in the right direction, inflating the overall Type I error. Methods like the Chen-DeMets approach or combination test methods (combining p-values from stages) properly account for this dependence and maintain valid inference.
Question 3 True / False
All adaptive trial modifications must be pre-specified in the protocol to maintain inferential validity. Unplanned modifications, even if scientifically reasonable, can compromise the trial's statistical properties.
TTrue
FFalse
Answer: True
Pre-specification is the dividing line between adaptive design (valid) and data-driven modification (potentially invalid). If the adaptations and their decision rules are specified before data collection, the statistical properties (Type I error, power) can be computed and controlled through simulation. Unplanned modifications introduce researcher degrees of freedom — the temptation to change the design in response to disappointing results — which inflates Type I error in unmeasurable ways. Regulatory agencies (FDA, EMA) require that all adaptations be pre-specified in the statistical analysis plan.
Question 4 Short Answer
Explain the ethical advantage of response-adaptive randomization over fixed randomization in a clinical trial.
Think about your answer, then reveal below.
Model answer: Response-adaptive randomization allocates a larger proportion of patients to the arm that is performing better as data accumulate. This means fewer trial participants are assigned to the inferior treatment compared to fixed equal randomization. The ethical advantage is that each individual patient has a higher probability of receiving the better treatment, reducing the total number of patients exposed to ineffective or harmful therapy. The tradeoff is reduced statistical efficiency (the groups become unequal, reducing the power of the comparison), which must be weighed against the ethical benefit.
The tension between individual ethics (each patient should get the best available treatment) and collective ethics (society needs reliable evidence from well-powered trials) is central to adaptive randomization. Fixed equal randomization maximizes statistical power; fully adaptive allocation minimizes patient harm. Practical adaptive designs use moderate adaptation rates that balance these competing goals.