Screening programme audits are recommended in the European Guidelines and considered an ethically required part of population-based screening. We conducted a systematic programmatic audit in order to identify the magnitude of different screening failures and their impact on the remaining cervical cancer burden in Norway.
The case population consisted of all invasive cancers recorded in the Cancer Registry of Norway with incidence dates in 1993-2013, and amounted to 6467 cases. Ten age-matched controls per case were drawn from the National Population Register. Cases and controls were individually linked to the screening databases for screening history analysis and categorisation. There were 17 case and 121 control women who had opted-out of screening registration and these were excluded from further analysis for a final dataset of 6450 cases and 64,549 controls. Screening exposure variables were constructed using both three and five year windows of exposure. Cervical tests in the last 6 months before diagnosis were considered diagnostic, non-preventive, and therefore discounted. Screening history was categorised by participation, result of primary screening test and management of positive tests. Odds ratios for cervical cancer according to screening history were calculated with 95% confidence intervals (CIs).
Most, 63%, of the case women had not participated in screening in the three-year interval ending 6 months before diagnosis, compared with 47% of the controls. Corresponding proportions for a five-year interval was 53 and 37%. The risk of cervical cancer was similarly reduced among women that had participated in screening in either a three- (OR 0.47, 95% CI 0.45-0.50) or five-year interval (OR 0.45, 95% CI 0.43-0.48). Of the cases, 22% had a negative last primary cytology in the three-year interval, compared with 50% of the controls. The risk of cervical cancer was low among these women compared to non-participants (OR 0.30, 95% CI 0.28-0.32). Among women screened within the recommended three-year interval, the risk for cervical cancer was much higher for those with a borderline (OR 11.5, 95% 10.3-12.8) or high-grade primary cytology (OR 18.0, 95% CI 15.2-21.3) compared with those with normal smears. Women with borderline and high-grade primary cytology constituted 10 and 4.3% of all women with cervical cancer, compared with only 2.0 and 0.5% of the controls.
Non-participation was the most important contributor to cervical cancer risk in Norway. However, women with positive screening tests had significantly elevated risks of cervical cancer, indicating possibilities for improvement also in the management and treatment of screen-positive women.