The aim of the study was to determine the epidemiological background, clinical details and surgical outcome of patients presenting with urogenital fistulae to St Luke's Hospital, Uyo, and the associated VVF Unit at Mbribit Itam, Akwa Ibom State, Nigeria, between January 1970 and December 1994. A retrospective review of hospital operating theater records and case notes was carried out. Clinical details and outcome were assessed for the total cohort of 2484 patients. Epidemiological data were extracted from the case notes of 715 patients presenting between January 1990 and December 1994. Of these 92.2% were of obstetric etiology, 80.3% following neglected obstructed labor, 6.9% following cesarean section, and 5.0% followed ruptured uterus; 4.4% followed pelvic surgery and the remaining 3.4% of miscellaneous causes included malignancy, coital injury, infection and trauma; 8% had a coexisting rectovaginal fistula or third-degree perineal tear. Only 37.3% of patients were aware of their age; the median age of this group was 28 years. Literacy was difficult to judge reliably, although 29% were able to sign their name. Parity ranged from 0 to 17, and only 31.4% of fistulae related to first pregnancies. Although 73.1% were delivered in hospital, in 97.1% labor was initially managed at home, with a traditional birth attendant, in a maternity home, or in church; 34.1% were delivered by cesarean section, although the live-birth rate was only 10.3% in the causative pregnancy. For a variety of reasons 124 women were not operated upon: 1954 underwent only one operation, giving a presumptive cure rate at first operation of 81.2%; 247 underwent two, 116 three, 32 four, and 11 five operations during the study period. The ultimate closure rate was 97.7%, with only 0.6% undergoing urinary diversion. The type and distribution of fistulae recorded in this series is consistent with previous series of largely obstetric fistulae from the developing world. Surgical cure rates are also comparable. The epidemiological background is at variance with previous reports in several respects; this may reflect biosocial differences in the population studied.