两项试验范式下估计目标的变化与结论不一致的发生率:一项系统综述
该研究系统检索了2010至2024年ClinicalTrials.gov的3期试验,识别出498项重复试验(占5%)。其中12%的配对试验存在主要结局或一类错误控制方法的中途变更,17%的配对试验统计结论不一致,FDA仍批准了14种药物但其中8例要求补充随机对照试验。
要点
- 重复3期试验仅占2010–2024年3期试验的约5%,主要见于皮肤科和眼科。
- 约12%的重复试验配对存在主要结局或一类错误控制方法的中途变更。
- 统计结论不一致的配对试验占17%,FDA对不一致结果的上市决策缺乏一致性。
核心结论
- 两项试验范式下重复试验仅占3期试验约5%,结局变更少见但存在,且结论不一致时监管决策缺乏一致性。
研究设计与样本
- 系统综述,提取2010年1月1日至2024年1月7日ClinicalTrials.gov所有3期试验,自动检索重复试验并由两名研究者核对。
主要发现
- 在9925项合格试验中识别出498项重复研究(5%;246对或三胞胎),主要分布于皮肤科(n=96,19%)和眼科(n=58,12%)。
- 86项试验未报告主要终点(n=56,11%)或报告不恰当(n=30,6%)。
- 29对试验(12%)存在主要结局或一类错误控制方法的中途变更,其中17对(7%)涉及配对内所有试验。
- 9对试验(4%)在第一项试验揭盲后对另一项试验进行了变更。
- 206对试验(70%)有研究结果,其中35对(17%)存在统计结论不一致;FDA批准了14种药物,其中8例要求补充随机对照试验。
关键数值与推荐
- 原文未报告关键数值与推荐标准。
临床怎么用
- 临床解读重复3期试验证据时,需关注试验中途主要结局或统计方法的变更及配对试验结论不一致对稳健性的影响。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND AND OBJECTIVES: The United States Food and Drug Administration (FDA) has recommended that two adequate and well-controlled trials are conducted to establish efficacy. Although replication strengthens evidentiary robustness, conducting two trials may raise interpretative issues in case of discrepant findings, or if knowledge gained from the first trial influences the conduct of the second. The objectives of this study are to describe how the strict "two-trial rule" (ie, duplicated trials) is applied to phase 3 trials, especially regarding outcome modifications, and how discrepancies between trials are handled from a regulatory perspective. METHODS: We extracted all phase 3 trials from ClinicalTrials.gov between January 1, 2010, and January 7, 2024. We automatically retrieved duplicated trials by searching identical sponsor, disease, intervention, and study design. All pairs were checked by two investigators, and discrepancies were discussed among authors. We extracted all outcome modifications, study results, and subsequently searched how discrepant conclusions and outcome modifications were considered by the FDA. RESULTS: Among 9925 eligible trials, there were 498 duplicated studies (5%; 246 pairs or triplets), mainly in dermatology (n = 96, 19%) and ophthalmology (n = 58, 12%). Among them, 86 trials did not report their primary end point (n = 56, 11%) or did but not properly (n = 30, 6%). Changes in primary outcome or in the type 1 error control method during the trial were observed for 29 pairs (12%), and mostly concerned all trials within a pair (n = 17, 7%). For nine pairs (4%), changes were performed in one trial after unblinding of the first trial. Study results were available for 206 pairs (70%), among which discrepancies in statistical conclusion were found in 35 (17%). FDA authorization was granted for 14 drugs despite these discrepancies, among which an additional randomized controlled trial was required for eight cases. CONCLUSION: Although the two-trial paradigm has been the default rule for years, duplicated trials only represent about 5% of phase 3 trials between 2010 and 2024. Changes in outcomes that may seriously compromise the robustness of the results were observed in less than 5% of situations. However, there is a lack of consistency in marketing decisions in case of discrepant results.
引用
- Hlavaty A, Jambon-Barbara C, Manceau M, et al. Changes in estimands and prevalence of discrepancies in conclusions under the two-trial paradigm: a systematic review. J Clin Epidemiol. 2026. PMID: 42176998. doi:10.1016/j.jclinepi.2026.112341
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。