Rik van Eekelen1, Esther de Hoop2, Ingeborg van der Tweel2. 1. Department of Biostatistics & Research Support, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, PO Box 85060, 3508 AB Utrecht, The Netherlands. Electronic address: r.vaneekelen@umcutrecht.nl. 2. Department of Biostatistics & Research Support, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, PO Box 85060, 3508 AB Utrecht, The Netherlands.
Abstract
OBJECTIVE: Usually, sequential designs for clinical trials are applied on the primary (=efficacy) outcome. In practice, other outcomes (e.g., safety) will also be monitored and influence the decision whether to stop a trial early. Implications of simultaneous monitoring on trial decision making are yet unclear. This study examines what happens to the type I error, power, and required sample sizes when one efficacy outcome and one correlated safety outcome are monitored simultaneously using sequential designs. STUDY DESIGN AND SETTING: We conducted a simulation study in the framework of a two-arm parallel clinical trial. Interim analyses on two outcomes were performed independently and simultaneously on the same data sets using four sequential monitoring designs, including O'Brien-Fleming and Triangular Test boundaries. Simulations differed in values for correlations and true effect sizes. RESULTS: When an effect was present in both outcomes, competition was introduced, which decreased power (e.g., from 80% to 60%). Futility boundaries for the efficacy outcome reduced overall type I errors as well as power for the safety outcome. CONCLUSION: Monitoring two correlated outcomes, given that both are essential for early trial termination, leads to masking of true effects. Careful consideration of scenarios must be taken into account when designing sequential trials. Simulation results can help guide trial design.
OBJECTIVE: Usually, sequential designs for clinical trials are applied on the primary (=efficacy) outcome. In practice, other outcomes (e.g., safety) will also be monitored and influence the decision whether to stop a trial early. Implications of simultaneous monitoring on trial decision making are yet unclear. This study examines what happens to the type I error, power, and required sample sizes when one efficacy outcome and one correlated safety outcome are monitored simultaneously using sequential designs. STUDY DESIGN AND SETTING: We conducted a simulation study in the framework of a two-arm parallel clinical trial. Interim analyses on two outcomes were performed independently and simultaneously on the same data sets using four sequential monitoring designs, including O'Brien-Fleming and Triangular Test boundaries. Simulations differed in values for correlations and true effect sizes. RESULTS: When an effect was present in both outcomes, competition was introduced, which decreased power (e.g., from 80% to 60%). Futility boundaries for the efficacy outcome reduced overall type I errors as well as power for the safety outcome. CONCLUSION: Monitoring two correlated outcomes, given that both are essential for early trial termination, leads to masking of true effects. Careful consideration of scenarios must be taken into account when designing sequential trials. Simulation results can help guide trial design.
Authors: Ruben P A van Eijk; Stavros Nikolakopoulos; Kit C B Roes; Bas M Middelkoop; Toby A Ferguson; Pamela J Shaw; P Nigel Leigh; Ammar Al-Chalabi; Marinus J C Eijkemans; Leonard H van den Berg Journal: J Neurol Neurosurg Psychiatry Date: 2019-07-10 Impact factor: 10.154