Abstract:
|
RCTs, where patients are allocated to receive treatment for different durations, may be used to determine the optimal duration of treatment. In these trials, patients in different arms receive the same treatment for a period. For example, in a trial comparing 3 treatment durations,0,6,and 12 months, patients in the 6 and 12 month arms receive the same treatment during the first 6 months. The overlap allows for more powerful analyses than the conventional comparisons. Suppose the outcome is survival time. For the 0 vs 6-month comparison, the power can be increased by including patients in the 12 month arm as patients in the 6-month arm for the first 6 months and censoring at 6 months. Moreover, any difference observed between the 6 and 12 month arms during the first 6 months can only be noise. Hence, this comparison should be restricted to patients who stay on study at 6 months and include only those events that occur after 6 months. Note that with equal allocations, events during the first 6 months will reduce the number of patients for the 6 vs 12-month analysis, thus lowering the power of the analysis. A new sample size allocation procedure has been designed to balance the power.
|