Davis C E
Department of Biostatistics, University of North Carolina, Chapel Hill 27599-7400, USA.
Control Clin Trials. 1997 Dec;18(6):557-60; discussion 561-7. doi: 10.1016/s0197-2456(96)00133-x.
Response variables from clinical trials are often divided into those considered primary and those considered secondary to the purposes of the study. If the difference between treatment groups on primary outcomes is not significant, the interpretation of significant differences in secondary response variables may be difficult because of multiple comparisons. It is proposed that the comparison on the primary outcome use a significance level of alpha and that the secondary endpoints be evaluated using a significance level alpha/(k + 1), where k is the number of secondary outcomes to be evaluated. Examples from recently completed clinical trials illustrate the use of this rule.
来自临床试验的反应变量通常分为那些被视为主要的和那些相对于研究目的被视为次要的变量。如果治疗组在主要结局上的差异不显著，由于进行了多次比较，那么对次要反应变量中显著差异的解释可能会很困难。建议对主要结局的比较使用显著性水平α，而对次要终点的评估使用显著性水平α/(k + 1)，其中k是要评估的次要结局的数量。最近完成的临床试验的例子说明了这条规则的应用。
