Surveillance colonoscopy at 5 years matched 3 years after risky polyps

After a colonoscopy removes a worrying polyp, guidelines call for another look in three years. A large European trial reported in the New England Journal of Medicine now suggests that, for these patients, waiting five years may be just as safe — at least on the evidence available so far [s1].

The question

Surveillance colonoscopy after polyp removal is a substantial share of all the colonoscopies done, and the intervals rest on limited direct evidence [s1]. Doing them too often exposes patients to a burdensome procedure and its small risks, and consumes scarce endoscopy capacity; doing them too rarely could let a cancer develop unseen. The EPoS II trial set out to test whether the recommended three-year interval could be stretched to five for people at higher-than-average risk [s1].

What they did

In this ongoing non-inferiority trial conducted in eight European countries, patients with high-risk adenomas were randomly assigned to a first surveillance colonoscopy five years after their polyp removal or at three years, the interval currently recommended in guidelines [s1]. High-risk adenomas were defined as at least one adenoma of 10 mm or larger, high-grade dysplasia, or villous growth, or 3 to 10 adenomas of any kind [s1]. The three-year group also underwent colonoscopy at five years [s1]. The trial, the second European Polyp Surveillance study, was registered as EPoS II [s2]. The primary endpoint is the cumulative incidence of colorectal cancer at 10 years, with a prespecified non-inferiority margin of 0.7 percentage points for the upper boundary of the confidence interval for the difference between the two groups [s1]. The results published now are an interim analysis, after 5.5 years of follow-up [s1]. A non-inferiority design is the appropriate frame for a de-escalation question: it does not try to show the longer interval is better, only that the longer wait does not raise cancer risk beyond an agreed margin.

What it showed

A total of 10,799 patients underwent randomisation — 5398 assigned to the five-year group and 5401 to the three-year group [s1]. The five-year cumulative incidence of colorectal cancer was 0.77% with the less-frequent surveillance and 0.82% with the more-frequent surveillance, a difference of −0.05 percentage points [s1]. The upper boundary of the confidence interval was 0.68, which met the criterion for non-inferiority [s1]. Deaths from colorectal cancer were rare and split almost evenly: 5 patients died of the disease, 3 (0.06%) in the five-year group and 2 (0.04%) in the three-year group [s1]. The distribution of cancer stage at diagnosis did not appear to differ substantially between the groups [s1].

How to read it

The most important word here is "interim." The trial is designed to answer its question at 10 years, and what has been reported is a snapshot at 5.5 years — the point at which the two groups have only just begun to diverge, since the five-year group had by then had roughly one surveillance colonoscopy and the three-year group two [s1]. To handle patients who missed surveillance, the analysis used inverse-probability weighting, a statistical adjustment for missing data rather than a direct measurement [s1]. And the confidence-interval boundary of 0.68 sits just under the 0.7-point margin: the interim result cleared the bar, but not by a wide margin [s1].

Two features strengthen the finding. It is a randomised comparison rather than an observational one, so the groups should be alike except for the assigned interval; and it was publicly funded, by the Research Council of Norway and other non-commercial backers [s1] — an independent test of doing fewer procedures, not more. That direction is unusual in screening research and worth noting, because there is no commercial incentive to fund a study whose success means less endoscopy.

What it does not say

The result speaks only to the interval between colonoscopies for people who have already had a high-risk adenoma removed. It is not a statement about who should be screened in the first place, nor about lower-risk polyps, and it does not change any guideline on its own. It also cannot yet address the endpoint that matters most — cancers arising between years five and ten — which is precisely what the full analysis is built to capture.

What to watch

The trial continues to its 10-year read-out, when the cumulative-incidence comparison will be assessed with the prespecified confidence interval [s1]. If the equivalence holds at that point, the case for a longer default interval after high-risk polyps strengthens considerably; if the curves separate later, the interim reassurance will look premature. Until then, the honest summary is that stretching the interval looks safe so far, in a trial not yet finished.

Sources

• [s1] Colonoscopy Intervals and Colorectal Cancer Incidence after Adenoma Removal. New England Journal of Medicine. 16 September 2026.
• [s2] ClinicalTrials.gov. European Polyp Surveillance Trial (EPoS II, NCT02319928). U.S. National Library of Medicine.