Meta分析在中药的多中心临床试验数据分析中的应用

A Meta Analysis in Multi-center Clinical Trials of Traditional Chinese Drugs

本文通过对４个中药复方制剂的多中心临床试验数据分别进行Ｍｅｔａ分析，以期对临床试验各亚单位统计推断结果不一致时如何解释进行初步的探讨。Ｍｅｔａ分析结果表明，虽然各多中心临床试验内各亚单位的显著性检验结果不一致，但每个多中心临床试验Ｍｅｔａ分析的结果均拒绝无效假设，提示Ｍｅｔａ分析的结果均与将各亚单位的资料合并后进行Ｒｉｄｉｔ分析统计的结果一致。Ｆａｉｌ—ｓａｆｅ研究结果表明，欲使４个多中心临床试验Ｍｅｔａ分析结论逆转，分别至少需要４个～３１个无效果的独立临床试验，提示本研究的阳性结局尚属可靠。由于Ｍｅｔａ分析方法是以综合多个研究发现为目的，对于有相同假设的多个独立的研究结果进行再分析的一种定量分析方法，尤其当各研究结果不一致时，可综合评价多个研究结果，从而减少偏倚，提高估计精度。因此，我们认为多中心临床试验资料应用合并进行统计后仍有必要采用Ｍｅｔａ分析方法再进行综合分析。

This paper dealst with how to conduct a meta-analysis on multi-center clinical trials with randomly blinded controlled design in traditional Chinese drugs.Multi-center clinical trials of four traditional Chinese drugs were carried out.In this study there were different statistical significances among sub-cemters in every trial.An attemption was made to expain scientifically the different results by using meta-analysis.Rosenthal Chi-square test was used for homogeneity test whish revealed that data in all four clinical trials were hemogeneous (χ 2=7.7349,2.89,1.6958,0.983 respectively).The Stouffer combined test and the Fisher combined test were used to estimate singnificance levels for combined studies with resuats of Stouffer's Zs being 3.5313,2.29,4.82, 4.50 respectively and Fisher's χ 2 were 38.1881,18.9515,40.2192,36.1079 respectively. The Fail-safe N(N fs0.05 ) was used to estimate the number of additional studies averaging zero effect size that would reverse the probabilities obstained from our combined tests.The number of additional studies averaging zero effect size that would reverse the probabilities obstained from our combined tests. The number of additional studies obtained from Nf s0.05 ranged from 4 to 31. Althoug the conclusion drawn from meta-analysis consists with that from conventional statistical analysis,we hold the point of view that it is necessary to do a meta-analysis when multi-center clinical trial is conductted because of the advantage of meta-analysis.