Effects of Multiple Comparisons on Integration of Stratified Analysis: On the Serum TG Decreasing Effect of EPA / DHA in Healthy Subjects.

When the data of the study participants, who met the renewed inclusion criteria, were extracted from the existing data, the sample size of the analysis set decreased, thus decreasing the statistical power. To address this concern, an integrated analysis weighted by the sample size of individual studies, in which a stratified sampling from multiple independent studies was performed, was proposed. However, little is known about the adjustment for multiple comparisons of longitudinal data. Thus, we herein investigated the effect of the multiple comparison method on the results of the integrated analysis after a stratified sampling. We used two clinical data, which evaluated the effects of EPA/DHA on the decrease of serum triglyceride; stratified the participants into healthy and hyperlipidemic groups for sampling; and performed an integrated analysis. The adjustments of the multiple comparisons of longitudinal data were made using the Bonferroni procedure and its improved procedures, i.e., Holm, Shaffer, and Holland-Copenhaver procedures, and the results were compared. Results showed that the Holm, Shaffer, and Holland-Copenhaver procedures had no significant differences. Because the Bonferroni procedure is very conservative, it should not be used with a large number of comparisons. [ABSTRACT FROM AUTHOR]