临床试验中分类数据的处理方法比较

Comparison of Procedure for Processing Categorical Data in Clinical Trials

目的通过模拟比较,探讨分析多中心临床试验中高维列联表资料的有效方法。方法采用Monte Carlo模拟比较CMH检验与logistic回归处理多分类单向有序、无序资料以及CMH检验与一般线性模型GLM(General Linear Model)处理双向有序资料的区别。结果多分类单向有序和无序资料采用CMH检验和logistic回归分析结果相同,检验效能基本一致。双向有序资料采用GLM与CMH检验分析结果相同,检验效能基本一致。结论在多中心临床试验中,高维列联表资料的统计分析可以用logistic回归或GLM代替CMH卡方检验。

Objective To probe appropriate procedure to analyze high dimension contingency tables in multicenter clinical trials. Methods Monte Carlo simulations are adopted to compare the logistic regression and CMH or GLM and CMH when they are used to analyze one-way ordinal and nominal data or two-way ordinal data, respectively. Results Same results will be obtained when both logistic regression and CMH test are used simultaneously to complete the analysis of ordinal or nominal data. So will the results when both General Linear Model（GLM） and CMH test are used to process two-way ordinal data. Conclusion Logistic regression or GLM can replace CMH test when we analyze categorical data from high dimension contingency tables in multicenter clinical trials.