Does an invitation to colonoscopy reduce the incidence of colorectal cancer and colorectal cancer mortality compared with usual care?
In the first randomized trial of CRC screening using colonoscopy, a smaller than expected reduction in CRC incidence was seen in both the intention-to-treat (absolute risk reduction [ARR] = -0.22% over 10 years; NNI = 455) and adjusted per-protocol analyses (ARR = -0.38%; P < .05; NNS = 263). The same was true for reduction in CRC mortality in the intention-to-treat analysis (ARR = -0.03%; P = NS) and per-protocol analysis (ARR -0.15%; P < .05). The lower-than-expected mortality reduction may be explained in part by improvements in treatment and the modest duration of follow-up. The authors were careful to try to adjust for differences between invitees who accepted colonoscopy and those who did not (to avoid the healthy volunteer bias), although unmeasured confounding is still possible. Trials comparing fecal immunochemical test with colonoscopy are nearing their conclusion, and the results may add further clarity.
Randomized controlled trial (single-blinded)
Despite widespread use as a screening test for colorectal cancer (CRC) in the United States, colonoscopy has never been subjected to a randomized trial. These authors identified 94,959 healthy men and women, aged 55 to 64 years, from the Netherlands, Norway, Sweden, and Poland who had not previously been screened for CRC. None of these countries had organized programs for CRC screening using colonoscopy, at least not in the regions from which participants were recruited. Unfortunately, follow-up data for 10,374 Dutch participants could not be included because of changes in (overly) restrictive European data protection laws that made it impossible to obtain data for uninvited persons from the general population. The remaining 84,585 participants were randomized in a 1:2 ratio to receive an invitation to a single screening colonoscopy or usual care. The median age at enrollment was 59 years, half the participants were women, and most came from Poland or Norway. Colonoscopy was performed at dedicated centers with training and quality assurance programs.
Only 11,843 of the 28,220 persons invited to screening (42%) actually underwent colonoscopy. The median follow-up was 10 years, 91% had a good or very good bowel preparation, 97% achieved intubation of the cecum, and 30.7% had an adenoma detected. The risk of CRC was higher in the screened group for the first 5 years after colonoscopy (due to cancer diagnoses during the exams and heightened surveillance for pre-cancerous lesions, presumably), but was then less likely thereafter. In the intention-to-treat analysis, the incidence of CRC was significantly lower in the screened group (0.98% vs 1.20%; relative risk [RR] 0.82; 95% CI 0.70 - 0.93; number needed to invite [NNI] = 455 over 10 years). CRC mortality was not significantly lower in the screened group (0.28% vs 0.31%; RR 0.9; 0.64 - 1.16). There was no difference in all-cause mortality (11.03% vs 11.04%). The authors performed a separate per-protocol analysis to estimate the apparent benefits if everyone invited to colonoscopy had been screened, adjusting for baseline differences between those accepting the invitation and those who ignored it (important to at least partially adjust for the healthy volunteer bias). They estimate a lower incidence of CRC (0.84% vs 1.22%; RR 0.69; 0.55 - 0.83; number needed to screen [NNS] = 263) and a greater reduction in CRC mortality (0.15% vs 0.30%; RR 0.50; 0.27 - 0.77; NNS = 667). Complications were rare: 15 episodes of major bleeding (0.13%, 0 fatal) and no perforations.
Mark H. Ebell, MD, MS
University of Georgia