Logistic回归、CMH检验与Meta分析在多中心临床试验中应用探讨

目的探讨临床随机对照试验多中心效应比较的统计方法。方法以一项多中心临床随机对照试验数据为例,运用χ2检验、CMH检验、Meta分析及logistic回归分析。结果CMH检验显示各中心间效应值的一致性检验差异有统计学意义(P<0.05),扣除中心效应后,组间比较差异有统计学意义(P<0.05);Meta分析异质性检验差异无统计学意义(P>0.05),采用固定效应模型,合并后效应值组间差异有统计学意义(P<0.05),logistic回归分析,各中心效应值差异无统计学意义(P>0.05),组间效应差异有统计学意义(P<0.05),且存在可能影响效应的协变量。结论多中心临床随机对照试验研究中,如果存在分中心组间疗效差异趋势不一致时,可选择Meta分析及logistic回归分析,然后对三种分析方法的结果作出客观的评价。如果logistic回归分析存在影响效应的协变量,建议对这些协变量进行再分析。

Application of Logistic Regression,CMH Test and Meta Analysis in Multi-center RCT

Objective To explore the statistic method for comparison for multiple center effects in multi-center randomized control trail(RCT).Methods The data acquired from a multi-center RCT were used as the case for study to process with Chi-square test,CMH test,Meta analysis and logistic regression.Results The outcome of CMH test indicates the effect discrepancy among different centers which is statistically significant(P<0.05).When the center-induced effects were excluded,difference still exists among groups(P<0.05)with statistic significance.However,the heterogeneity test in Meta analysis shows no statistic significance(P<0.05).Among the data discrepancy,but has significance(P<0.05)on inter-group discrepancy which implies a hidden variable with merged effects on the fixed effect model(P<0.05).Conclusion In the multi-center RCT clinical research,Meta analysis and logistic regression should be used if the therapeutic effect inclines to have discrepancy among different centers.Then based on the results from Meta analysis and logistic regression and combined with other possible factors that may have potential effects,the outcome of multi-center RCT can be properly evaluated.