上市后不良反应评价中似然比检验方法的比较

Comparison of likelihood ratio tests in post-marketing safety surveillance

目的对上市后安全性监测中各类似然比检验方法(likelihood ratio test,LRT)进行概述及比较。方法通过文献检索对上市后不良反应评价中LRT方法的原理进行综述并对进一步的探索方向进行展望。结果现有研究表明与报告比值比、信息成分方法相比LRT方法能更好控制I型错误发生率,并且有多种方法适用于不同的应用场景。结论LRT方法是上市后不良反应评价可行的方法之一,尤其适用于需要严格控制伪发现率,或者需要权衡伪发现率/I型错误发生率与灵敏度/检验功效的情况,但仍有许多问题需进一步探索。

Objective To survey and compare likelihood ratio test(LRT)approaches to post-marketing safety surveillance.Methods The fundamentals of likelihood ratio tests as well as the advantages and disadvantages of each method were summarized.Meanwhile,priorities of research in the future were predicted.Results Existing evidence suggested that LRT methods could better control type I errors than the reporting odds ratio and information component methods.Besides,LRT approaches employed a class of precise statistical tools which enabled objective evaluation under various scenarios.Conclusion LRT methods are feasible for post-marketing safety surveillance,especially when rigorous control of the false discovery rate is required or when a trade-off between the false discovery rate/type I error incidence and sensitivity/power is needed.However,there are still some problems that remain unsettled.