OBJECTIVES: To assess the cost-effectiveness of paroxetine (PX) compared with conventional tricyclic antidepressants (TCA's) for major depressive disorder in Japan. METHODS: A decision analytic model was built in order to compare three regimens: 1) initiating with PX followed by a TCA (PX-TCA's); 2) a TCA followed by PX (TCA's-PX); and 3) TCA's only. The decision tree incorporated the probabilities of dropping out of therapy obtained from a meta-analysis of randomized controlled trials in Japan, and also relapse or recurrence probabilities of depressive symptoms employed from published literature. Resource use and costs were estimated by the national health insurance (NHI) scoring system. Health related quality of life was incorporated in the model using utility values in the literature. The analysis was performed from a patient perspective during nine months including the follow-up period. The sensitivity analyses were conducted to test the robustness of the results. RESULTS: The average costs per patient for each regimen of the PX-TCA's, the TCA's-PX and the only TCA's were estimated with 4676€, 4541€ and 5045€, respectively. Regarding the utility analysis, the PX-TCA's best contributed to gain with incremental cost-effectiveness ratio of 5587€/QALY. The average cost per patient for each regimen of the PX-TCA's or the TCA's-PX remained lower than that of the only TCA's over the ranges of the sensitivity analyses. CONCLUSIONS: A treatment for major depression using PX combined with TCA's was cost-saving in terms of average costs per patient, comparing to the treatment by the TCA's alone. The first choice of PX did not reduce the average cost towards the first choice of TCA's in the combined treatment. When quality of life of patients incorporated, the first choice of PX was best recommended among the three strategies.