SIMULATION OF AN ADDITIONAL GO/NO-GO EFFICACY INTERIM ANALYSIS IN A HEAD-TO-HEAD RCT
Author(s)
Van Montfort K
Nyenrode Business University, Breukelen, The Netherlands
OBJECTIVES: A head-to-head randomized clinical trial (RCT) to evaluate a new drug is financially risky, because a positive outcome is uncertain. We simulate and evaluate a head-to-head RCT design incorporating an additional go/no-go efficacy interim analysis to show the consequences of this additional interim analysis. METHODS: We simulate the endpoint event-free-survival (EFS) of patients in a head-to-head RCT. The decision rule of the additional interim efficacy analysis (i.e. stop or continue) depends on the number of patients (i.e. 300, 400 or 500 patients) and the significance level (i.e. α=0.05, α=0.10 or α=0.20) of the interim analysis. The RCT without an interim-analysis has significance level α=0.05, power=0.86 and sample size 800. RESULTS: Each combination of sample size and significance level, which is called a scenario, is investigated by simulating 2,000 trials of 800 patients. Per simulated scenario we report, among others, the “Probability of positive final analysis test result GIVEN negative interim analysis test result” (= wrongly stopped) and the “Probability of negative final analysis test result GIVEN positive interim analysis test result” (= wrongly continued). The results of the first scenario, i.e. an interim analysis at 300 included patients and significance level α=0.05, are as follows. If the actual improvement of the EFS hazard rate is 0%, the abovementioned probabilities are 3.65% and 7.85%. An actual improvement effect of 5% changes the probability values to 20.25% and 12.80, while an actual improvement effect of 10% causes the probability values 28.50% and 5.55%. CONCLUSIONS: The simulated probabilities were mainly influenced by the actual EFS improvement. The smaller the actual outcome improvement, the greater the probability of continuing the trial up to 800 included patients without getting a positive final test result (i.e. wrongly continued). The developed software (i.e. R-codes) can easily be applied to other cases.
Conference/Value in Health Info
2014-11, ISPOR Europe 2014, Amsterdam, The Netherlands
Value in Health, Vol. 17, No. 7 (November 2014)
Code
PRM217
Topic
Study Approaches
Disease
Oncology