An invitation to a single screening colonoscopy reduced colorectal cancer (CRC) incidence by 19% over 13 years but did not significantly reduce CRC mortality in the multicountry NordICC randomized trial. The absolute incidence difference remained modest, and the investigators reported that CRC mortality in the group not offered screening was substantially lower than anticipated when the trial was designed.
Published in The Lancet, the analysis extends previously reported 10-year findings by three years. Investigators analyzed 84,583 men and women aged 55 to 64 years from Norway, Poland, and Sweden who were randomly assigned in a 1:2 ratio to an invitation for a single screening colonoscopy or no screening. That cohort represents 89.1% of individuals originally enrolled in the trial; data from the Netherlands were excluded because of continued restrictions on the availability of trial endpoints. The primary outcomes were CRC incidence and mortality after 10 to 15 years in intention-to-screen analyses. Secondary outcomes included CRC incidence and mortality in per-protocol analyses, all-cause mortality, and subgroup analyses by age, sex, and cancer location. The trial is ongoing.
Of 28,217 participants assigned to the screening group, 11,841, or 42%, underwent colonoscopy; 56,366 participants were assigned to no screening. Participation varied by country, from 33% in Poland to 61% in Norway. Median follow-up was 13 years in both groups. Participants were followed through linkage with national cancer and cause-of-death registries.
Incidence lower with screening
At 13 years, 375 CRCs occurred among 28,217 participants invited to screening, compared with 912 among 56,366 participants not offered screening. The corresponding 13-year risks were 1.46% and 1.80%, respectively, representing a 19% relative reduction and a 0.34-percentage-point absolute reduction in CRC incidence in the intention-to-screen analysis. The number needed to invite to screening to prevent one CRC within 13 years was 294.
The additional three years of observation increased the number of CRC cases by 45% in the screening group and 47% in the no-screening group compared with the 10-year analysis, allowing more detailed assessment of prespecified subgroups. The relative incidence reduction changed little — from 18% at 10 years to 19% at 13 years — while the absolute difference between groups increased. Because fewer than half of invited participants underwent screening, investigators also conducted per-protocol analyses estimating the effect if all participants assigned to screening had received colonoscopy. These analyses used an instrumental variable approach intended to avoid reliance on measurement of all confounders associated with screening participation. Under an assumption of additive homogeneity, estimated CRC incidence was 1.00% with screening versus 1.80% without screening, an absolute difference of 0.80 percentage points and a 45% relative reduction. Sensitivity analyses evaluated alternative assumptions of multiplicative homogeneity and monotonicity.
Mortality difference remains uncertain
CRC deaths occurred in 106 participants assigned to screening and 236 assigned to no screening, corresponding to 13-year risks of 0.41% and 0.47%, respectively. The absolute difference was 0.06 percentage points, corresponding to an estimated 12% relative reduction, but the difference was not statistically significant. All-cause mortality was also similar, at 16.30% in the screening group and 16.34% in the no-screening group, for a risk ratio of 1.00.
In per-protocol analyses, CRC mortality was 0.33% in the screening group and 0.47% in the no-screening group. The authors described the per-protocol estimates of screening benefit on CRC mortality as still imprecise.
The investigators reported that CRC mortality in the no-screening group was substantially lower than anticipated during trial planning, with the observed rate about half the expected rate.
In contrast, observed CRC incidence was close to the rate predicted when the trial was designed. The authors cited improvements in CRC treatment and prognosis as a potential explanation for the lower mortality observed during the trial.
Greater reduction for distal cancers
Anatomic subgroup analyses showed a larger estimated effect of screening for distal than proximal CRC. Distal CRC occurred in 224 participants assigned to screening and 563 assigned to no screening, corresponding to a 21% relative reduction. Proximal CRC occurred in 129 and 283 participants, respectively, with a smaller and imprecise estimated effect. Mortality differences were also small when cancers were separated by proximal and distal location.
Among men, 214 CRCs occurred among 14,154 participants assigned to screening versus 541 among 28,247 assigned to no screening, corresponding to a 23% relative reduction. Among women, there were 161 cases among 14,063 participants in the screening group versus 371 among 28,119 in the no-screening group; the estimated reduction was smaller. The investigators wrote that colonoscopy could be more effective in men than women but that the difference was not statistically significant.
Age-stratified findings similarly showed a larger estimated effect among participants aged 60 to 64 years at enrollment. In that group, 208 CRCs occurred among 13,850 participants assigned to screening versus 550 among 27,636 assigned to no screening, a 25% relative reduction. Among participants aged 55 to 59 years, there were 167 cases among 14,367 participants assigned to screening and 362 among 28,730 assigned to no screening, with a smaller estimated reduction.
The investigators did not observe what they considered a clinically relevant shift toward earlier-stage cancers in the screening group. They noted that screening-detected CRCs accounted for 16% of the 375 CRCs in the screening group at 13 years. Cancers detected later were diagnosed following clinical symptoms, as screening colonoscopy was offered only once and was not available outside the trial.
Among the 11,841 participants who underwent colonoscopy, 65 CRCs, or 0.55%, were diagnosed after screening over the 13-year follow-up. The authors compared that figure with a nationwide analysis of the Polish screening program, which reported a 10-year risk of cancer after colonoscopy of 0.38% among all participants, and wrote that the interval cancer estimates confirmed a high quality of colonoscopies performed in NordICC.
Limitations and longer follow-up
The authors identified the randomized design, large sample, screening-naive population, minimal screening contamination of the control group, and complete long-term follow-up as strengths.
Limitations included the participation rate of 42% (which was expected when planning the trial) and the absence of individual data on colonoscopies after screening. The investigators also could not exclude some self-initiated screening among participants in the control group, although registry data from Norway and Poland suggested such activity was negligible.
Continued NordICC follow-up is intended to determine more definitively whether colonoscopy screening reduces CRC mortality and how long the preventive effect of a single colonoscopy persists. The authors concluded that the 13-year findings support a reduction in CRC incidence even though a mortality benefit has not been established: “Screening might be justified if it prevents people from getting cancer even without reducing the risk of dying.”
Several authors reported consulting fees, honoraria, or equipment from endoscopy manufacturers including Olympus and Fujifilm, or research funding from government and nonprofit sources; the remaining authors declared no competing interests.
Expert Insight
GI & Hepatology News invited Aasma Shaukat, MD, MPH, of the Division of Gastroenterology, Department of Medicine, NYU Grossman School of Medicine, New York, to comment on the study.
She said: “The study has important implications in understanding the benefit that may be derived from screening colonoscopy, and that the benefit is risk based. It’s important to note that the study is underpowered for endpoint of colorectal cancer mortality, and longer follow up is needed. The study should be placed in context with other published evidence on the effectiveness of colonoscopy in reducing the risk of colorectal cancer.”
Dr. Shaukat also authored separate commentary on the NordICC findings published in The Lancet. Regarding the small absolute difference in colorectal cancer mortality, she wrote: “From a clinical perspective, the question becomes not just one of whether colonoscopy saves lives, but one of how many procedures, with what opportunity costs, are required to avert one death in contemporary practice.”