CHENG Qing-qing, WANG Rui, WANG Jing-jing, TANG Dai-mao, ZHANG Ling-xiao, XU Ling, YI Fang, YANG Juan, LV Ying-hua, ZHENG Qing-shan, SHEN Yi-feng, LI Hua-fang, SI Tian-mei
Objective: Clinical trials of antipsychotic drugs involve unique patient management and ethical considerations, leading to a growing number of non-inferiority trials, where the selection of the non-inferiority margin (M2) poses a significant challenge. This study uses aripiprazole, an active control drug, as an example to explore a rational method for selecting the non-inferiority margin, providing a reference for the design of similar studies in antipsychotic drugs. Methods: The total score reduction in the Positive and Negative Syndrome Scale (PANSS) was chosen as the primary endpoint. Following the FDA-recommended fixed margin method, and in consultation with clinical expert consensus, the M2 was determined. First, based on literature data from aripiprazole versus placebo studies, a model-based meta-analysis (MBMA) was conducted to obtain the pharmacodynamic characteristics of the net effect of aripiprazole (after deducting the placebo effect) and to explore the influencing factors, providing a basis for pooled data analysis across different treatment durations. Second, a meta-analysis was conducted using data from aripiprazole versus placebo studies to estimate the treatment effect of aripiprazole and its 95% confidence interval (CI), with the lower bound of the 95%CI used as the entire effect (M1). Subsequently, the range of M2 values in antipsychotic clinical trials reported in the literature was collected, and the corresponding sample sizes were estimated. Combined with expert consensus on clinically acceptable sample sizes, M2 was determined. Results: MBMA suggested that the therapeutic effect of aripiprazole approaches a plateau after 6 weeks of treatment. Meta-analysis estimated the net effect of aripiprazole after 6 weeks of treatment as 11.39 (95%CI: 9.72, 13.06), with M1 being 9.72. The current clinical literature shows that M2 values for non-inferiority trials in schizophrenia are widely distributed, ranging from 5 to 8. For generic drugs, assuming a one-sided α=0.025, a power of 80%, and a common standard deviation (SD) of 18, the estimated sample size per group is 122, corresponding to an M2 of 6.5, which is similar to FDA's guidelines for clinical endpoint bioequivalence studies and can achieve bridging purposes. For innovative drugs, using approximately 50% of M1 (9.72) as the non-inferiority margin, i.e., M2=5, the corresponding sample size per group should be 205. Considering the difficulty of patient management and care, this is a clinically acceptable sample size. Conclusion: For non-inferiority trials using PANSS score reduction as an efficacy endpoint and aripiprazole as the control drug, an M2 of 6.5 is appropriate for generic drugs, and an M2 of 5.0 is appropriate for innovative drugs. This study provides important references and insights for selecting non-inferiority margins in similar trials involving other active control drugs.