We investigated the long-term prognosis of borderline ovarian tumours and determined risk factors for recurrence. One hundred and twenty-one borderline ovarian tumours treated between 1994 and 2003 at the participating institutions in the Tohoku Gynecologic Cancer Unit were retrospectively investigated for clinical stage, histopathological subtype, surgical technique, postoperative chemotherapy, the presence or absence of recurrence, and prognosis. The median follow-up period was 57 months (1-126 months). One hundred and nine cases (90.6%) were at clinical stage I. The histopathological subtypes consisted of 91 cases of mucinous tumour (75.2%), 27 cases of serous tumour (22.3%), and three cases of endometrioid tumour. Conservative surgery was used in 53 cases (43.8%), radical surgery in 68 cases (56.2%), a staging laparotomy in 43 cases (35.5%), and postoperative adjuvant therapy in 30 cases (24.8%). Recurrence was found in eight cases, but no tumour-related deaths were reported. Although no significant difference in disease-free survival rate was seen between different clinical stages, the difference in disease-free survival rate between serous and non-serous (mucinous and endometrioid) types was significant (P<0.05). The 10-year disease-free survival rate was 89.1% for the radical surgery group and 57.4% for the conservative surgery group -- this difference was significant (P<0.05). In the conservative surgery group, cystectomy and serous tumour were independent risk factors for recurrence. Although recurrence was observed, the long-term prognosis of borderline ovarian tumour was favourable, without tumour-related deaths. Considering the favourable prognosis, conservative surgery can be chosen as far as the patient has a non-serous tumour and receive adnexectomy. However, in cases of serous type and/or receiving cystectomy special care should be given as relative risk rates of recurrence elevate by 2-4-folds.