Partners MS Center, Brigham & Women's Hospital, Brookline, MA, USA.
Mult Scler. 2010 Jul;16(7):840-7. doi: 10.1177/1352458510371409. Epub 2010 Jun 7.
Many phase I/II clinical trials in multiple sclerosis use gadolinium-enhanced lesions as the outcome measure. The best scanning interval and analysis for this outcome has not been determined. The objective of this study was to compare timing schemes and analysis techniques in terms of power for phase I/II clinical trials. Data were simulated under four scenarios assuming a negative binomial distribution for the number of new lesions and an exponential distribution for the duration of enhancement. The first scenario assumed an immediate treatment effect on the number of new lesions, while the second scenario assumed a delayed treatment effect. The third scenario assumed a higher proportion of patients had no new lesions, and the final scenario assumed an immediate treatment effect on the duration of enhancement. For each scenario, power for a six-month trial with 100 patients per arm was calculated using 10 analysis strategies. The scanning intervals tested were monthly scans, bimonthly scans and a single end-of-study scan. In addition, cost-effectiveness of each trial design and analysis was compared. Negative binomial regression models for the total number of new lesions were the most powerful analyses under an immediate treatment effect, and repeated measures models with a categorical time effect were the most powerful analyses under a delayed treatment effect. Although monthly scans generally provided most power, this design was also most costly. Designs with fewer scans per patient provide similar power and are more cost-effective. Negative binomial regression models are more powerful than non-parametric approaches.
许多多发性硬化症的 I/II 期临床试验使用钆增强病变作为结局指标。尚未确定这种结局的最佳扫描间隔和分析方法。本研究的目的是比较不同时间方案和分析技术在 I/II 期临床试验中的效能。在假设新病变数量呈负二项分布和增强持续时间呈指数分布的四种情况下模拟数据。第一种情况假设对新病变数量有即刻治疗效果,而第二种情况假设有延迟治疗效果。第三种情况假设更多的患者没有新病变,最后一种情况假设对增强持续时间有即刻治疗效果。对于每种情况,使用 10 种分析策略计算了 100 例患者每组的六个月试验的效能。测试的扫描间隔为每月扫描、每两个月扫描和单次研究结束扫描。此外,还比较了每种试验设计和分析的成本效益。在立即治疗效果下,用于新病变总数的负二项回归模型是最有力的分析方法,而在延迟治疗效果下,具有分类时间效果的重复测量模型是最有力的分析方法。虽然每月扫描通常提供最大的效能,但这种设计也最昂贵。每个患者的扫描次数较少的设计提供了相似的效能,并且更具成本效益。负二项回归模型比非参数方法更有优势。