R. Jover, M. Bretthauer, J. Cubiella, S.J.E. Barry, A. Macios, L. Barnett, R. Erichsen, M.F. Kaminski, H.-O. Adami, Ø. Holme, J. Ijspeert, M. Ferlitsch, M. Dinis-Ribeiro, M. Pellise, A. Martínez-Roca, S. Baile-Maxía, L. de-Castro, L. Bujanda, P. Díez-Redondo, J.A.N. Nilsen, H.D. Nguyen, L. Bernklev, J. Tilma, U.S. Løve, C. Jaensch, N.D. Pilonis, M. Szura, A. Bialek, A. van Berkel, M.C.W. Spaander, F. ter Borg
9 min
Current clinical guidelines recommend frequent surveillance colonoscopies (beginning at 3 years) for patients who have had high-risk adenomas removed. This practice is intended to prevent interval colorectal cancers, but it places a significant burden on patients and healthcare systems. The European Polyp Surveillance (EPoS) II trial is a large, randomized, noninferiority study involving 10,799 patients across eight European countries, designed to determine if a less-frequent surveillance interval (beginning at 5 years) is as effective as the standard 3-year interval in preventing colorectal cancer.
The interim analysis, conducted at 5.5 years of follow-up, found that the 5-year cumulative incidence of colorectal cancer was 0.77% in the 5-year surveillance group and 0.82% in the 3-year surveillance group. This difference of -0.05 percentage points met the prespecified criteria for noninferiority. Additionally, the study found no substantial differences in the distribution of cancer stages at diagnosis between the two groups. Although the detection of advanced adenomas and other polyps was higher in the 3-year group—likely due to the increased frequency of examinations leading to higher detection of previously missed lesions—this did not translate into a lower incidence of colorectal cancer.
These results challenge the current necessity of 3-year surveillance intervals for patients categorized as having high-risk adenomas. By demonstrating that 5-year intervals are noninferior, this study suggests that many patients could safely undergo less frequent, less invasive, and more cost-effective follow-up. This shift could reduce the administrative and physical burden on patients without compromising clinical safety, pending final results at 10 years.
Sam: How did they distinguish a genuine cancer-risk difference from simply finding cancer earlier in the more frequently examined group?
Alex: Both groups were offered an examination at the same final checkpoint. The earlier-start group had its additional examination before that. Sharing the final checkpoint helps avoid lead-time bias: a distorted comparison caused by detecting disease earlier rather than by changing its occurrence.
Sam: So it wasn’t simply one group getting examined while the other remained unchecked. Both schedules ended with a chance to detect otherwise hidden cancers.
Alex: Yes, and the comparison was randomized across centers in multiple European countries. Trial personnel also checked health records annually for cancer diagnoses and deaths. The main endpoint is cancer incidence over ten years, so this report captures only the earlier part of the planned follow-up.
Sam: Let’s get to the central result. How different were the cancer risks, rather than just the number of examinations?
Alex: In the analysis adjusted for missing examinations, cancer incidence was 0.77 percent with less-frequent surveillance, compared with 0.82 percent with more-frequent surveillance. Both were low. The small observed difference favored waiting longer, but that is not evidence that waiting prevents more cancers.
Sam: A tiny observed difference doesn’t settle noninferiority. How much extra risk could the uncertainty still allow?
Alex: The prespecified margin allowed an increase of 0.7 percentage points in absolute cancer risk. The upper confidence boundary was 0.68 percentage points. So the result cleared the rule, but only narrowly.
Sam: That distinction matters. The best estimate was close to no difference, while the uncertainty still allowed a higher risk with less-frequent surveillance.
Alex: And the threshold is the authors’ chosen criterion, not a finding that every possible increase would be clinically unimportant. They used a stringent confidence interval for this interim look, preserving most of their statistical error allowance for the final analysis.
Sam: Cancer stage could reveal a disadvantage that total incidence misses. Were cancers found later in their course when surveillance started later?
Alex: The stage distributions did not appear to differ substantially. Deaths from colorectal cancer were rare, and the authors did not perform inferential comparisons for that outcome. Those findings are reassuring descriptively, but they do not establish equal survival.
Sam: What did the less-frequent schedule actually save? The service burden is part of the reason to ask this question.
Alex: It involved almost half as many colonoscopies during follow-up. That’s a substantial reduction in procedures. Colonoscopy-related adverse events were uncommon, however, with no evidence of a substantial difference between groups, so this trial did not demonstrate a clear complication reduction.
Sam: The earlier-start group also had more chances to find polyps. Did that translate into a stronger prevention signal?
Alex: More adenomas were identified with more-frequent surveillance, including advanced adenomas, which have more concerning features. The authors suggest repeated examinations probably detect some lesions missed on a single examination. Whether that additional detection benefits patients remains unresolved.
Sam: So finding more possible precursors is not the same result as preventing more cancers.
Alex: That’s the distinction here. The authors propose that slow growth could leave time to detect clinically significant lesions later, before cancer develops. But that is an interpretation to be tested by longer follow-up, not something this interim analysis establishes.
Sam: Now the missing examinations. If people who skipped the final check had undiagnosed cancers, wouldn’t the apparent risk be too low?
Alex: That is the central limitation. A substantial minority in each group did not attend the final scheduled examination. Health records captured diagnosed cancers well, but could not rule out cancers that remained undiagnosed.
Sam: How did the authors account for those patients without pretending they knew the missing results?
Alex: They used inverse probability weighting: giving observed patients different weights to account for missing examinations. Attendance probabilities were estimated from measured characteristics, including age, sex, country, and the reason for the original colonoscopy. Observed patients represented more or fewer similar eligible patients according to those probabilities.
Sam: That adjustment still needs an assumption. What does it require about people who did not attend?
Alex: After accounting for those measured characteristics, missing examination data must be missing at random. If unmeasured differences linked attendance to hidden cancer, weighting might not remove the bias. A prespecified unadjusted analysis also supported noninferiority, but it cannot establish what the missing examinations would have found.
Sam: Given the randomized design but incomplete attendance and unfinished follow-up, who should read the full article, and where should they start?
Alex: Gastroenterologists, surveillance researchers, and guideline developers should read it closely. Start with Trial Design and Patients to understand eligibility and the shared final examination. Then read Statistical Analysis for the missing-data assumptions and the Discussion for the limitations. My assessment is that this is substantial interim evidence, not a settled long-term equivalence claim.
Sam: And for researchers outside this area, what is the line worth keeping?
Alex: After high-quality removal of high-risk adenomas, later surveillance met the interim cancer-risk criterion with far fewer procedures. The longer-term answer is still pending.
Sam: Fewer procedures, with a conclusion that still has a time limit.