A 9-Year Follow-up Study of Participants and Nonparticipants in Sigmoidoscopy Screening: Importance of Self-Selection

Background: Self-selection may compromise cost-effectiveness of screening programs. We hypothesized that nonparticipants have generally higher morbidity and mortality than participants. Methods: A Swedish population-based random sample of 1,986 subjects ages 59 to 61 years was invited to sigmoidoscopy screening and followed up for 9 years by means of multiple record linkages to health and population registers. Gender-adjusted cancer incidence rate ratio (IRR) and overall and disease group-specific and mortality rate ratio (MRR) with 95% confidence intervals (95% CI) were estimated for nonparticipants relative to participants. Cancer and mortality rates were also estimated relative to the age-matched, gender-matched, and calendar period-matched Swedish population using standardized incidence ratios and standardized mortality ratios. Results: Thirty-nine percent participated. The incidence of colorectal cancer (IRR, 2.2; 95% CI, 0.8-5.9), other gastrointestinal cancer (IRR, 2.7; 95% CI, 0.6-12.8), lung cancer (IRR, 2.2; 95% CI, 0.8-5.9), and smoking-related cancer overall (IRR, 1.4; 95% CI, 0.7-2.5) tended to be increased among nonparticipants relative to participants. Standardized incidence ratios for most of the studied cancers tended to be >1.0 among nonparticipants and <1.0 among participants. Mortality from all causes (MRR, 2.4; 95% CI, 1.7-3.4), neoplastic diseases (MRR, 1.9; 95% CI, 1.1-3.5), gastrointestinal cancer (MRR, 4.7; 95% CI, 1.1-20.7), and circulatory diseases (MRR, 2.3; 95% CI, 1.2-4.2) was significantly higher among nonparticipants than among participants. Standardized mortality ratio for the studied outcomes tended to be increased among nonparticipants and was generally decreased among participants. Conclusion: Individuals who might benefit most from screening are overrepresented among nonparticipants. This self-selection may attenuate the cost-effectiveness of screening programs on a population level. (Cancer Epidemiol Biomarkers Prev 2008;17(5):1163–8)