Background
Removal of adenomas reduces colorectal cancer incidence and mortality; however, the benefit of surveillance colonoscopy on colorectal cancer risk remains unclear. We examined heterogeneity in colorectal cancer incidence in intermediate-risk patients and the effect of surveillance on colorectal cancer incidence.

Methods
We did this retrospective, multicentre, cohort study using routine lower gastrointestinal endoscopy and pathology data from patients who, after baseline colonoscopy and polypectomy, were diagnosed with intermediate-risk adenomas mostly (>99%) between Jan 1, 1990, and Dec 31, 2010, at 17 hospitals in the UK. These patients are currently offered surveillance colonoscopy at intervals of 3 years. Patients were followed up through to Dec 31, 2014.We assessed the effect of surveillance on colorectal cancer incidence using Cox regression with adjustment for patient, procedural, and polyp characteristics. We defined lower-risk and higher-risk subgroups on the basis of polyp and procedural characteristics identified as colorectal cancer risk factors. We estimated colorectal cancer incidence and standardised incidence ratios (SIRs) using as standard the general population of England in 2007. This trial is registered, number ISRCTN15213649.

Findings
253 798 patients who underwent colonic endoscopy were identified, of whom 11 944 with intermediate-risk adenomas were included in this analysis. After a median follow-up of 7·9 years (IQR 5·6–11·1), 210 colorectal cancers were diagnosed. 5019 (42%) patients did not attend surveillance and 6925 (58%) attended one or more surveillance visits. Compared to no surveillance, one or two surveillance visits were associated with a significant reduction in colorectal cancer incidence rate (adjusted hazard ratio 0·57, 95% CI 0·40–0·80 for one visit; 0·51, 0·31–0·84 for two visits). Without surveillance, colorectal cancer incidence in patients with a suboptimal quality colonoscopy, proximal polyps, or a high-grade or large adenoma (≥20 mm) at baseline (8865 [74%] patients) was significantly higher than in the general population (SIR 1·30, 95% CI 1·06–1·57). By contrast, in patients without these features, colorectal cancer incidence was lower than that of the general population (SIR 0·51, 95% CI 0·29–0·84).

Interpretation
Colonoscopy surveillance benefits most patients with intermediate-risk adenomas. However, some patients are already at low risk after baseline colonoscopy and the value of surveillance for them is unclear.