BACKGROUND - Screening for ovarian cancer is based on ultrasound, colour Doppler, and tumour markers. There is only limited evidence on their discriminatory performance and no evidence on their effectiveness in reducing mortality. OBJECTIVE - To investigate the discriminatory performance of CA 125 as a screening test for ovarian cancer. METHODS - A registry of 15 093 serum samples drawn in 1968-72 was linked to the cancer registry. During follow up between 1968 and 1980 24 ovarian cancers were identified. One or two matched case-control design nested within the sample bank was applied and the concentrations of CA 125 were assessed. RESULTS - Case-control differences (relative risk 4-0, 95% confidence interval 1.0 to 15.5 at 20 kU/1) were found. Detection rate of the CA 125 test was 21-33% and the true negative rate was 75-98% depending on the cut off level and interval between drawing of the blood sample and diagnosis of the cancer. CONCLUSION - CA 125 is not a valid screening test if used alone. Case-control differences of borderline significance were found in CA 125 before diagnosis of ovarian cancer, but they were not large enough to provide a sufficient detection rate.