BACKGROUND: Studies from the early 90s have suggested that patients selected for and compliant with treatment at specialized lithium clinics have lower-than-expected suicide rates. The present study examines whether such findings can be replicated under less select treatment conditions. METHOD: All 362 patients in Göteborg, Sweden, with DSM-III-R mood or schizoaffective disorders, hospitalized at least once during an 8-year period and treated with lithium for a minimum of 1 year, were followed. The study included 3911 patient-years on lithium and, because of permanent or temporary discontinuation, 1274 patient-years off lithium. RESULTS: The risk of suicide was significantly increased on (standard mortality ratio [SMR] = 6.1) as well as off lithium (SMR = 29.0), but the relative risk of suicide was 4.8 times higher during periods off lithium (p