Long-term effects of once-only flexible sigmoidoscopy screening on colorectal cancer incidence and mortality: 21-year follow-up of the UK Flexible Sigmoidoscopy Screening randomised controlled trial.

Background: Flexible sigmoidoscopy screening reduces colorectal cancer incidence and mortality; however, uncertainty exists about the duration of protection and differences by sex and age. We assessed effects of once-only flexible sigmoidoscopy screening after 21 years' follow-up.

Methods: The UK Flexible Sigmoidoscopy Screening Trial is a multicentre randomised controlled trial that recruited men and women aged 55-64 years from general practices serving 14 hospitals. Among participants indicating that they would attend flexible sigmoidoscopy screening if invited, randomisation (2:1) to the control (no further contact) or intervention (invited to once-only flexible sigmoidoscopy screening) group was performed centrally in blocks of 12, stratified by centre, general practice, and household type. Masking of intervention was infeasible. Primary outcomes were colorectal cancer incidence and mortality. The Kaplan-Meier method estimated cumulative incidence. Primary analyses estimated intention-to-treat hazard ratios (HRs) and risk differences, overall and stratified by subsite, sex, and age. The trial is registered with ISRCTN, number 28352761.

Findings: Among participants recruited between Nov 14, 1994, and March 30, 1999, 170 432 were eligible and 113 195 were randomly assigned to the control group and 57 237 were randomly assigned to the intervention group. 406 participants were excluded from analyses (268 in the control group and 138 in the intervention group), leaving 112 927 participants in the control group (55 336 [49%] men and 57 591 [51%] women) and 57 099 in the intervention group (27 966 [49%] men and 29 103 [51%] women). Of participants who were invited to be screened, 40 624 (71%) attended screening. Median follow-up was 21·3 years (IQR 18·0-22·2). In the invited-to-screening group, colorectal cancer incidence was reduced compared with the control group (1631 vs 4201 cases; cumulative incidence at 21 years was 3·18% [95% CI 3·03 to 3·34] vs 4·16% [4·04 to 4·29]; HR 0·76 [95% CI 0·72 to 0·81]) with 47 fewer cases per 100 000 person-years (95% CI -56 to -37). Colorectal cancer mortality was also reduced in the invited-to-screening group compared with the control group (502 vs 1329 deaths; cumulative incidence at 21 years was 0·97% [0·88 to 1·06] vs 1·33% [1·26 to 1·40]; HR 0·75 [0·67 to 0·83]) with 16 fewer deaths per 100 000 person-years (-21 to -11). Effects were particularly evident in the distal colorectum (726 incident cancer cases in the invited-to-screening group vs 2434 cases in the control group; HR 0·59 [0·54 to 0·64]; 47 fewer cases per 100 000 person-years [-54 to -41]; 196 cancer deaths in the invited-to-screening group vs 708 deaths in the control group; HR 0·55 [0·47 to 0·64]; 15 fewer deaths per 100 000 person-years [-19 to -12]) and not the proximal colon (871 incident cancer cases in the invited-to-screening group vs 1749 cases in the control group; HR 0·98 [0·91 to 1·07]; one fewer case per 100 000 person-years [-8 to 5]; 277 cancer deaths in the invited-to-screening group vs 547 deaths in the control group; HR 1·00 [0·86 to 1·15]; zero fewer deaths per 100 000 person-years [-4 to 4]). The HR for colorectal cancer incidence was lower in men (0·70 [0·65-0·76]) than women (0·86 [0·79 to 0·93]; p_interaction=0·0007) but there was no difference by age.

Interpretation: We show that once-only flexible sigmoidoscopy screening reduces colorectal cancer incidence and mortality for two decades and provide important data to inform colorectal cancer screening guidelines.

Funding: National Institute for Health and Care Research Health Technology Assessment Programme and the Medical Research Council.