A New Procedure of One-Sided Test in Clinical Trials with Multiple Endpoints

This paper considers clinical trials with multiple endpoints, in which the efficacy of a test treatment is confirmed only when the superiority of the test treatment to control is evidenced in at least 1 endpoint and non-inferiority is observed in the remaining endpoints. Perlman and Wu (2004) proposed a one-sided testing procedure that was adaptable to this type of trials. This paper proposes a modification of this procedure, in which the likelihood ratio test is replaced with another test similar to that proposed by Tang et al. (1989). The performance of the proposed procedure was examined through theoretical consideration and Monte Carlo simulations assuming normality and homoscedasticity. The simulation study demonstrated that the power of the proposed procedure was higher than that of the procedure proposed by Perlman and Wu; in this procedure, type I error rates are maintained within nominal significance levels unless primary endpoints are highly correlated.