Abstract
Objective: This research examines the diagnostic performance of the Japanese version of the GAD-7 (J-GAD-7) based on analysis of two independent studies. The first assesses the accuracy of the J-GAD-7 in identifying generalized anxiety disorder (GAD) in primary care (Study 1). Next, a meta-analysis of diagnostic accuracy studies in five language versions of the GAD-7 (English, Spanish, Dutch, Turkish, and German) is conducted (Study 2), the results of which are compared with those of the J-GAD-7.
Methods：In Study 1, subjects (n=161; male=55, female=106) returned the self-rating questionnaire (J-GAD-7) to their health care professional within 48 hours and underwent a diagnostic evaluation interview based on the Japanese version of M.I.N.I-Plus. Most (n=151) also responded to the SF-8, and there were negligible missing data in either set of questionnaires. Of the 161 participants, 63 were diagnosed as suffering GAD. The meta-analysis of diagnostic test accuracy (MA-DTA) conducted in Study 2 focused on identifying GAD at the optimal cut-off value 10 using integration of the eight selected studies.
Results: The optimal cut-off value for the J-GAD-7 (Study 1) was determined as ≥10, which displayed sensitivity of 70.6 %, specificity of 93.5 %, and positive likelihood ratio of 10.9. The area under the ROC curve (AUC) was 0.939, with an SE of 0.019 (P <0.0001). Stratum-specific likelihood ratios (SSLRs) of the J-GAD-7 scores of 0-4, 5-9, 10-14, and 15-21 for GAD were 0.04 (95% CI: 0.01-0.14), 1.17 (95% CI: 0.68-2.01), 10.94 (95% CI: 4.29-27.8), and 10.94 (95% CI: 3.00-39.8), respectively. Findings of the Kruskal-Wallis test showed a significant relationship between the severity of the J-GAD-7 and MCS of the SF-8 (χ２＝71.73, df=3, p ≤0.0001). In Study 2, sensitivity was 0.754 (95%CI :0.604 - 0.861), specificity was 0.816, and AUC was 0.828. The heterogeneity τ2 was 0.709 (95% CI: 0.064-1.626) for sensitivity and 1.153 (95% CI: 0.264-2.332) for specificity.
Conclusions: The Japanese version of the GAD-7 was validated as an adequate tool for identifying GAD in primary care. Moreover, its performance at the optimal cut-off value 10 was equivalent to that of the MA-DTA for five other language versions.