Accounting for differential exclusions in the NordICC trial discloses stronger-than-reported effects of screening colonoscopy

Objective: Recently, results on colorectal cancer (CRC) incidence and mortality reduction by the offer of screening colonoscopy were reported for the first time from a randomized controlled trial (RCT), the Nordic-European Initiative on Colorectal Cancer (NordICC) trial. Despite randomization, there was a substantially lower proportion of post-randomization exclusions of CRC cases due to cancer registry-recorded date of diagnosis before recruitment in the invited group than in the usual-care group. We aimed to evaluate the impact of such differential exclusions on the trial's effect estimates on CRC risk.

Study design and setting: We compared reported post-randomization exclusions of CRC cases due to cancer registry-recorded date of diagnosis, and we derived adjusted effect estimates on CRC risk accounting for the reported differential post-randomization exclusion of CRC cases in the invited group and the usual-care group.

Results: Reported post-randomization exclusion proportions of CRC cases were originally reported as 52/31,472 (0.17%) and 159/63,133 (0.25%) in the invited and usual-care group (p<0.005) in an analysis including participants from all four NordICC study countries, and as 52/28,277 (0.20%) and 164/56,529 (0.29%) in the recent analysis of 10-year follow-up data from three of the countries (p=0.018). Accounting for the differential exclusion proportions increased the estimated CRC risk reduction (95% CI) from originally reported 18% (7-30%) to 25% (95% CI 13-35%) in intention-to-screen analysis. Estimated reduction of CRC risk among screening attenders increased from originally reported 31% (17-45%) to 50% (25-69%) in adjusted per-protocol analysis.

Conclusion: Accounting for differential post-randomization exclusions of CRC cases leads to stronger-than-reported effect estimates in the so far only RCT on long-term effects of screening colonoscopy.