過去有許多臨床研究探討認知行為取向對憂鬱症狀之療效。個別認知行為治療對輕度至中度憂鬱症狀緩解的成效已確立，然針對認知行為團體治療療效進行統合分析的研究僅有少數幾篇，且並無文獻對認知行為團體治療近十年趨勢、有效之療程、團體內容及對不同病症或群體之療效，提供系統性分析。本研究蒐集1997至2007年間電子資料庫中與認知行為團體治療相關之期刊論文共126篇，以含認知行為治療團體、具量化資料、具憂鬱症狀之測量等六項篩檢標準，篩出32篇研究樣本，其中共16篇具足夠研究數據進行統合分析。結果顯示本研究之研究樣本平均效果量為-0.68，以次群體效果量分析整理出近十年認知行為團體治療之研究趨勢，並提供最可能在認知行為團體治療中受益者的特性：在受試者特徵上，兒童青少年群體的效果量高於成人；在團體設計上，有發治療手冊的效果量較佳。本研究結果再度支持認知行為治療團體對憂鬱情緒具緩解效果。

In the past decades, the effects of the individual cognitive-behavioral therapy on mild to moderate depression have been supported. With regards to the cognitive-behavioral group therapy (CBGT), a small amount of research has addressed the impacts of different group designs on its therapeutic effect. Moreover, only few meta-analysis studies have tried to integrate these results. However, no recent study has provided systematic information regarding the group characteristics of the CBGT for the past ten years. The purpose of this meta-analysis is to find out the most effective group design of the CBGT to maximize its therapeutic effect for different populations and diagnoses. One hundred and twenty-six studies included in this study were retrieved from Academic search premier, PsycINFO and Medline, during January 1997 to June 2007. These retrieved papers were screened via the six inclusion criteria by four independent reviewers. Finally, a total of 32 papers were selected in this study, but only half of them had sufficient data to compute the effect size. Two different types of effect sizes -d and eta squared values -were used to compare the magnitudes of the treatment effects across studies. The mean standardized effect size d for 11 selected studies was -0.68, with high heterogeneity. Due to high heterogeneity among these studies, subgroup analyses were used to determine a possible moderator which could put these studies into homogenous groups. Subgroup analyses indicated a stronger effect for the child and adolescent group than the adult group. Also, providing a treatment manual for group members seemed to be an important characteristic of the group design to increase the effect of the CBGT. Five studies with more complex designs were compared by the magnitudes of their eta squared values. Moderate to large eta squared values for the time main effect indicated a significant improvement of participants, depression symptoms after the CBGT treatment. A small and a large eta squared values for the group main effect suggested a significant but unstable CBGT group effect on depression when comparing with the control group. Moderate eta squared values for the time-group interaction effect implied a possible time-group influence when explaining the CBGT treatment effect. Furthermore, all selected studies in this study are journal articles, which could lead to overestimate its effect size. The publication bias of this study was evaluated via the funnel plot and the fail-safe N. The result of the funnel plot and the fail-safe N examination of this study suggested a positive publication bias that was not stronger enough to overrule the significant treatment effect of this study. Due to the diversity of the CBGT studies in the past ten years, the present study was unable to determine key moderators of the group designs to maximize the CBGT treatment effect. More research is needed to ascertain the characteristics of group design to reach the best effect for different populations and diagnoses.