OBJECTIVE: To estimate the prevalence of severe stress urinary incontinence (SUI) among perimenopausal women and to examine potential obstetric risk factors. DESIGN: Mail survey of female volunteers for epidemiological research. SETTING: Postal questionnaire on SUI. POPULATION: Three thousand one hundred and fourteen women aged 49-61 years who comprised the GAZEL cohort. METHODS: Logistic regression using data from the entire cohort to estimate the impact of risk factors. A second logistic regression using data from women who had given birth included obstetric history. MAIN OUTCOME MEASURE: Prevalence of severe SUI defined by the response 'often' or 'all the time' to the question 'Does urine leak when you are physically active, cough or sneeze?' RESULTS: Two thousand six hundred and twenty-five women (85%) completed and returned the questionnaire The frequency of SUI reported in the preceding four weeks was as follows: 'never' 32%, 'occasionally' 28%, 'sometimes' 26%, 'often' 10% and 'all the time' 5%. Prevalence of severe SUI was lowest among nulliparous women (7%), but it was similar among parous women regardless of birth number (14-17%). The prevalence of severe SUI was not associated with mode of delivery (14% for women delivered by caesarean only vs 16% for vaginal births). Significant risk factors for severe SUI were high body mass index (BMI >30), diabetes mellitus, previous incontinence surgery, parity and first delivery under the age of 22 years. CONCLUSION: Previous pregnancy itself is a risk factor for severe SUI among women who reach the age of 50. In this age group the impact of the mode of delivery (spontaneous, forceps or caesarean) on severe SUI is slight.