The purpose of this study was to investigate the one-year test-retest reliability and the demographic correlates of a self-administered web-based depression section of the World Health Organization-Composite International Diagnostic Interview (WHO-CIDI) in a working population. Overall, 1060 out of all employees (N=1279) from a manufacturing company in Japan responded to two web-based surveys of depression of the WHO-CIDI within a one-year interval in 2009 and 2010. The concordance between lifetime diagnoses of major depressive disorder on two occasions was calculated as percent agreement (%), Gwet's AC1, and Yule's Q indicators were compared by gender, age, education, and marital status. For the total sample, percent agreement was 94%, AC1 was 0.93, and Yule's Q was 0.82. The concordance rate was low (0.15) among those who were diagnosed at either time or both times. The concordance differed significantly across education and marital status. While the agreement indicators were relatively high, consistent with previous reports based on face-to-face interviews conducted within a shorter interval, the low stability of positive cases may challenge the accuracy of lifetime diagnosis of major depressive disorder using a web version of the WHO-CIDI. Education and marital status might affect the test-retest reliability. Copyright