OBJECTIVE: Urinary incontinence (UI) is often considered to be an age-related disease that develops gradually as women grow older. Much remains to be learnt about factors that promote its incidence or its remission. Our objective was to assess its incidence and risk factors. DESIGN: Longitudinal cohort study. SETTING: French GAZEL cohort. POPULATION: A cohort of 4127 middle-aged women (aged 47-52 years at baseline) over an 18-year period (1990-2008). METHODS: UI was defined as 'difficulty retaining urine'. The question was asked at baseline and repeated every 3 years over an 18-year period. Two groups (UI incidence and remission) were analysed according to status at baseline (continent or incontinent). A multivariable analysis (Cox model) was used to estimate the risk factors for UI incidence and remission. MAIN OUTCOME MEASURES: Annual incidence and remission rates and risk factors for UI incidence and remission. RESULTS: The annual incidence and remission rates for UI were 3.3% and 6.2%, respectively. High educational level (hazard ratio [HR] = 1.28; 95% confidence interval [95% CI] = 1.05-1.55), parity, i.e. at least one baby versus no baby (HR = 1.64; 95% CI = 1.19-2.27), menopause (HR = 5.44; 95% CI = 4.47-6.63), weight gain, i.e. for each kilogram change in weight (HR = 1.00; 95% CI = 1.00-1.02), onset of depressive symptoms (HR = 1.31; 95% CI = 1.09-1.57) and impairment in health-related quality of life incidence (social isolation dimension [HR = 1.29; 95% CI = 1.04-1.60] and energy dimension [HR = 1.41; 95% CI = 1.17-1.70]) were associated with an increased probability of UI. The factors associated with persistent UI were age (HR = 0.58; 95% CI = 0.55-0.61), weight gain (HR = 0.99; 95% CI = 0.98-0.99) and transition to menopausal status (HR = 1.54; 95% CI = 1.19-1.99). CONCLUSIONS: Our study suggests that, in our population of middle-aged women, age, menopause, weight gain, onset of depression and impaired health-related quality of life may promote UI.