Characteristics of women aged 40-59, recorded at an initial breast screen, were compared with the outcome of incident, invasive breast cancer occurring up to 3 years after the initial screen. The study design was case-control, nested within the study population of the Canadian National Breast Screening Study. Screening consisted of a two-view mammographic examination and physical examination of the breasts; additional risk factor information was obtained from self-administered questionnaires. Of the etiological risk factors considered only age at entry or years menstruating were significant risk factors for breast cancer (P < 0.0025). Years menstruating accounted for much of the age effect and all of the menopausal status effect. Risk factors obtainable from mammography and physical examination were more significant than self-reported risk factors, other than age or years menstruating. In spite of possible misclassification of the variable parenchymal pattern, women with a mammographic film classified as either P2 or DY had a 2-fold risk of breast cancer (odds ratio = 2.1; 95% confidence interval = 1.5-2.9, P = 2.9E-05). An abnormality reported by either the radiologist [odds ratio = 1.7 (1.3-2.3)] or nurse examiner [odds ratio = 1.9 (1.4-2.6)] was also associated with an independent significant risk.