Four RCTs (437,600 participants, range 55,736 to 170,432) and 12 observational studies (four cohort studies and eight case control studies; number of patients not reported) were included in the review. Median follow-up ranged from seven to 12 years in the RCTs. Participation rates in the RCTs ranged from 58.3% to 86.6%. RCTs fulfilled eight or nine out of 11 quality criteria; none reported estimates adjusted for contamination of the control group. One RCT had a short follow-up duration and one RCT did not report per protocol analyses. Observational studies met between four and nine out of nine quality criteria.
RCTs: Screening with flexible sigmoidoscopy statistically significantly reduced the incidence of cancer in participants compared to controls on an intention-to-screen basis at any site (RR 0.82, 95% CI 0.75 to 0.89; Ι²=52%; four RCTs), proximal (RR 0.91, 95% CI 0.83 to 0.99; Ι²=0%; three RCTs) and distal (RR 0.69, 95% CI 0.63 to 0.74; Ι²=24%; three RCTs).
Screening with flexible sigmoidoscopy statistically significantly reduced mortality (RR 0.72, 95% CI 0.65 to 0.80; Ι²=0%; four RCTs). When subgroup analyses were performed by cancer site, the results remained statistically significant for distal sites (RR 0.54, 95% CI 0.42 to 0.67; Ι²=0%; three RCTs) but were no longer statistically significant for proximal sites (RR 0.95, 95% CI 0.77 to 1.17; Ι²=0%; two RCTs). Sensitivity analysis reduced heterogeneity but did not significantly alter the estimates of effect. Results on a per protocol basis showed considerably stronger effect estimates.
Observational studies: Sigmoidoscopy reduced cancer incidence (RR 0.51, 95% CI 0.39 to 0.65; three studies; Ι²=0) and mortality (RR 0.53, 95% CI 0.30 to 0.97; three studies; Ι²=68%). Colonoscopy also reduced cancer incidence (RR 0.31, 95% CI 0.12 to 0.77; five studies; Ι²=94%) and mortality (RR 0.32, 95% CI 0.23 to 0.43; three studies; Ι²=0). Results for distal and proximal incidence and mortality were reported and showed that reductions were no longer statistically significant for proximal mortality with sigmoidoscopy or proximal and distal cancer incidence with colonoscopy.
Results from indirect comparisons were reported. There was no evidence of significant publication bias.