A nested case-control study was undertaken in the Maribo County cohort of 27,811 women with negative cervical smears. Fifty-three women who later developed invasive cervical cancer constituted the cases, and five matched controls were selected from the cohort for each case. A total of 633 previous negative smears for the cases and controls were reviewed independently by two pathologists. The review showed misclassification to be frequent in these smears, which were collected in the period 1966-82, and the odds ratio for patients compared with controls for having at least one positive smear was 22.12 (95% CI 7.54-64.94). The study thus shows that more cancer cases could have been prevented by the screening programme were the test to have been more sensitive. The study also shows, however, that the participating women would have to pay a considerable price in the form of unnecessary extra tests if the sensitivity was too high.