NordICC Trial Results in Line With Expected Colorectal Cancer Mortality Reduction After Colonoscopy: A Modeling Study

Abstract

Colonoscopy screening is a widely recommended method for detecting colorectal cancer (CRC) in countries across the world.1Schreuders E.H. et al.Gut. 2015; 64: 1637-1649Crossref PubMed Scopus (843) Google Scholar However, until recently, no randomized controlled trials demonstrated its effectiveness in average-risk individuals. Recently, Bretthauer et al2Bretthauer M. et al.N Engl J Med. 2022; 387: 1547-1556Crossref PubMed Scopus (149) Google Scholar published preliminary results of a multicenter randomized controlled trial, the Nordic-European Initiative on Colorectal Cancer (NordICC) trial, that investigated the effects of once-only colonoscopy screening on CRC incidence and mortality.2Bretthauer M. et al.N Engl J Med. 2022; 387: 1547-1556Crossref PubMed Scopus (149) Google Scholar In the intention-to-screen analysis, which compared participants not offered screening to those offered screening regardless of participation, they found that the invited group had an incidence and mortality reduction at 10 years of 18% and 10%, respectively. The investigators noted that although the incidence and mortality reductions were clinically important, they were lower than anticipated based on observational and modeling studies. The publication of the NordICC trial results induced media attention and controversy regarding the effectiveness of colonoscopies.3Krz S. et al.Nature. 2023; 613: 235-237Crossref PubMed Scopus (1) Google Scholar Experts advised people to interpret the results cautiously, noting aspects of the NordICC trial that could contribute to the underwhelming findings. A critical issue was the low screening uptake (42%). In the adjusted per-protocol analyses, which compared participants not offered screening to those offered screening who received colonoscopy, incidence and mortality reductions at 10 years increased to 31% and 50%, respectively. Another important consideration was the relatively short 10-year follow-up period. This study aimed to evaluate whether the NordICC trial results are lower than expected based on modeling and to what extent the results could be explained by screening uptake and follow-up period. We used 3 Cancer Intervention and Surveillance Modeling Network CRC models to simulate NordICC trial outcomes: Colorectal Cancer Simulated Population Model for Incidence and Natural History (CRCSPIN), Microsimulation Screening Analysis Colorectal Cancer (MISCAN-Colon), and Simulation Model of Colorectal Cancer (SimCRC). Using these models, we simulated the NordICC trial population,2Bretthauer M. et al.N Engl J Med. 2022; 387: 1547-1556Crossref PubMed Scopus (149) Google Scholar with 42% of the invited group simulated to receive a 1-time colonoscopy and a usual-care group remaining unscreened (Supplementary Table 1). Our modeling assumptions included random selection into screening unrelated to CRC risk, full adherence to US guidelines for adenoma surveillance,4Gupta S. et al.Gastrointest Endosc. 2020; 91: 463-485.e5Abstract Full Text Full Text PDF PubMed Scopus (142) Google Scholar and high sensitivity of colonoscopy (Supplementary Materials). We compared model predictions to reductions in CRC incidence and mortality observed in the trial. Additionally, we simulated 5 hypothetical scenarios: 42% adherence with 15- and 20-year follow-up and 100% adherence with 10-, 15-, and 20-year follow-up. With 42% uptake and 10-year follow-up, the models predicted CRC incidence and mortality reductions of 11%–28% and 24%–32% (ranges are across models), respectively (Figure 1A and B). These estimates overlap the 95% confidence intervals (Cis) of the decreases observed in the NordICC intention-to-screen analyses, which were 18% (95% CI: 7–30) and 10% (95% CI: –16 to 36), respectively. The level of screening uptake had the largest impact on the findings: with 100% uptake, the model-predicted incidence and mortality reductions more than doubled to 26%–61% and 53%–70%, respectively (Figure 1C and D). These estimates compared well with reductions of 31% (95% CI: 17–45) and 50% (95% CI: 23–73), respectively, in the per-protocol NordICC analyses. Although the relative differences in risk reduction are substantial, the absolute incidence and mortality reduction only increased from 0.14%–0.29% to 0.31%–0.64% and from 0.10%–0.12% to 0.22%–0.26%, respectively, with 42% vs 100% uptake (Supplementary Figure 1). With 42% uptake, the predicted incidence reduction increased to 18%–33% and 19%–35% at the 15- and 20-year follow-up, respectively (Figure 1A). With 100% uptake, these reductions increased to 40%–73% and 43%–77%, respectively (Figure 1C). Combining 100% uptake with a 15-year follow-up resulted in expected incidence and mortality reductions of 40%–73% and 59%–79%, respectively (Figure 1C and D). In this study, we show that, in spite of suggestions otherwise,2Bretthauer M. et al.N Engl J Med. 2022; 387: 1547-1556Crossref PubMed Scopus (149) Google Scholar model predictions are consistent with the NordICC trial results. As experts have pointed out, the results of the NordICC trial are largely determined by the screening uptake and the follow-up duration. Prior observational studies reported that colonoscopy was associated with a pooled CRC mortality reduction of 62% (range, 11%–88%) at an average follow-up of 8 years.5Zhang J. et al.J Cancer. 2020; 11: 5953Crossref PubMed Scopus (22) Google Scholar This is within the CI of the per-protocol NordICC trial results and in line with the modeling results, which estimated an average 63% CRC mortality reduction with 100% uptake and the 10-year follow-up. A limitation of our study is that we assumed similar CRC risk for screening participants and nonparticipants. The trial results could be influenced by the healthy screenee effect, with people participating in screening at lower risk of CRC. If participants have a lower CRC risk than nonparticipants, this means that our models overestimate the effectiveness of screening. On the other hand, NordICC trial results show that the noninvited group had a higher risk of CRC than nonparticipants in the invited group,2Bretthauer M. et al.N Engl J Med. 2022; 387: 1547-1556Crossref PubMed Scopus (149) Google Scholar suggesting that there was self-selection of higher-risk individuals participating in screening (eg, those with a family history of CRC or symptoms). A higher CRC risk in participants than nonparticipants implies we may have underestimated screening effectiveness in our models. Systematic differences between screened and unscreened participants in the intervention group might explain differences between trial estimates of CRC mortality and model predictions. A second limitation concerns our assumed colonoscopy sensitivity. If the colonoscopy sensitivity achieved in the trial was lower than the sensitivity assumed when making model projections, then the projected benefits would be optimistic. Lower colonoscopy sensitivity would allow more adenomas to progress to cancer, reducing the effectiveness of colonoscopy. The trial’s 42% uptake aligns with the 5%–59% uptake reported in previous population-based studies.6Holden D.J. et al.Ann Intern Med. 2010; 152: 668-676Crossref PubMed Scopus (234) Google Scholar Low participation, exemplified by the 42% uptake, may mute the population-level benefits of CRC screening, leading some to perceive it as disappointing. Nevertheless, it is crucial to emphasize that the individual-level benefit for participants, which is closer to the NordICC trial’s adjusted per-protocol results, is more reassuring and reaffirms the effectiveness of the test. It is important to highlight that individuals who choose not to participate in screening do not receive any screening benefits, underlining the value of screening. Moreover, individuals should be aware that more favorable outcomes may be expected in the long term, especially beyond 15 years of follow-up, and that larger benefits could be achieved with repeated 10-yearly colonoscopy screening, as recommended in the United States. In conclusion, our findings show that NordICC trial results are consistent with anticipated mortality reductions from screening colonoscopy, and that with further follow-up higher benefits may be realized, especially in the NordICC’s per-protocol analyses. The CISNET-Colon Group includes Ann G. Zauber,1 Anne I. Hahn,1 Fernando Alarid Escudero,2 Christopher E. Maerzluft,3 Alexandra Katsara,4 Karen M. Kuntz,5 John M.Inadomi,6 Nicholson Collier,7 Jonathan Ozik,7 Luuk A. van Duuren,4 Rosita van den Puttelaar,4 Matthias Harlass,4 Claudia Leigh Seguin,8 Barak Davidi,8 Carlos Pineda-Antunez,9 Eric J. Feuer,10 and Lucie de Jonge4; from the 1Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York; 2Department of Health Policy, School of Medicine, and Stanford Health Policy, Freeman-Spogli Institute for International Studies, Stanford University, Stanford, California; 3Fred Hutchinson Cancer Center, Hutchinson Institute for Cancer Outcomes Research, Biostatistics Program, Public Health Sciences Division, Seattle, Washington; 4Department of Public Health, Erasmus MC, University Medical Center, Rotterdam, The Netherlands; 5Division of Health Policy and Management, University of Minnesota School of Public Health, Minneapolis, Minnesota; 6Department of Internal Medicine, University of Utah School of Medicine, Salt Lake City, Utah; 7Decision and Infrastructure Sciences, Argonne National Laboratory, Lemont, Illinois; 8Institute for Technology Assessment, Department of Radiology, Massachusetts General Hospital, Boston, Massachusetts; 9Global Health Cost Consortium (CISIDAT), Morelos, Mexico; and 10Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, Maryland. Danica van den Berg, MSc (Conceptualization: Equal; Formal analysis: Lead; Writing – original draft: Lead). Pedro Nascimento de Lima, PhD (Conceptualization: Equal; Formal analysis: Supporting; Writing – review & editing: Equal). Amy B. Knudsen, PhD (Conceptualization: Equal; Formal analysis: Supporting; Writing – review & editing: Equal). Carolyn M. Rutter, PhD (Conceptualization: Equal; Writing – review & editing: Equal). David Weinberg, MD, MSc (Conceptualization: Equal; Writing – review & editing: Equal). Iris Lansdorp-Vogelaar, PhD (Conceptualization: Equal; Supervision: Lead; Writing – review & editing: Equal). Supplementary Table 1CRC Age-Adjusted (World Population) Incidence Rates for SEER 1975–1979, Norway 2009–2010, and Poland 2009–2010 and CRC Incidence Rate Ratios Among Norway and Poland 2009–2010 vs SEER 1975–1979PeriodCRC cases per 100,000CRC rate ratioSEER, 1975–197939.31Norway, 2009–201039.81.0Poland, 2009–201025.30.6NOTE. Source for Poland and Norway data: European Cancer Information System (https://ecis.jrc.ec.europa.eu). Open table in a new tab NOTE. Source for Poland and Norway data: European Cancer Information System (https://ecis.jrc.ec.europa.eu). This study used 3 independently developed microsimulation models: Microsimulation Screening Analysis for Colorectal Cancer (MISCAN-Colon), Simulation Model of Colorectal Cancer (SimCRC), and Colorectal Cancer Simulated Population Model for Incidence and Natural History (CRCSPIN). Each model has a natural history and a screening component, summarized below. All models describe the natural history of CRC in an average-risk unscreened population. We assumed that all CRC develops through the adenoma-carcinoma pathway and that simulated persons are free of diagnosed CRC until screening in 2012. Each simulated individual can develop 1 or more colorectal lesions. Lesions may proceed through 3 phases: a noninvasive adenoma phase, a preclinical cancer phase, and a clinical cancer phase. Persons may die of other causes at any time. Each model’s natural history component was initially calibrated to Surveillance, Epidemiology, and End Results (SEER) data for the period 1975–1979. To adjust the models for CRC risk differences between SEER 1975–1979 and Norway and Poland, we compared CRC incidences in the countries. The magnitude of the difference was estimated by the ratio of CRC incidence in Norway and Poland between 2009 and 2010 relative to SEER data from 1975–1979 (Supplementary Table 1). We assumed that the decreased risk arises from changes in adenoma onset, not from slower progression of adenomas to CRC. Screening will alter some of the simulated life histories: some cancers will be prevented by the detection and removal of adenomas; other cancers will be detected in an earlier stage with a more favorable survival. The ability of a test to detect lesions depends on its sensitivity. These sensitivities are lesion based. We assumed a colonoscopy sensitivity of 0.75, 0.85, 0.95, and 0.95 for adenomas of 1 to <6 mm, adenomas of 6 to <10 mm, adenomas of ≥10 mm, and CRC, respectively. We assumed the same sensitivities for surveillance colonoscopy as for screening colonoscopy. Moreover, we assumed that individuals with an adenoma detected undergo colonoscopy surveillance according to the Multi-Society Task Force guidelines.5Zhang J. et al.J Cancer. 2020; 11: 5953Crossref PubMed Scopus (22) Google Scholar We assumed persons with adenoma findings are perfectly adherent with the surveillance colonoscopy schedules. Additionally, we assumed that persons in whom adenoma(s) have been detected remain on surveillance until age 85 years, provided that no adenomas are detected at the last surveillance colonoscopy. If adenomas are detected, then surveillance continues according to the clinical findings at the last colonoscopy until the person has a colonoscopy with no adenomas detected. Outcomes were simulated for 10 different birth cohorts (birth years ranging from 1948 to 1957). For each cohort, we simulated 2 strategies: (1) no screening and (2) once-only colonoscopy in 2012 with 100% adherence and surveillance for people with adenomas detected. The results of the different strategies were consolidated afterward in the postprocessing based on the screening participation of the trial participants.3Krz S. et al.Nature. 2023; 613: 235-237Crossref PubMed Scopus (1) Google Scholar All outcomes were tallied by year from 2012 onward. The primary outcomes included the number of CRC cases and CRC deaths. NordICC’s 10-Year Interim Results Are Unexpected and Inconsistent With Modeling PredictionsGastroenterologyPreviewWe commend van den Berg et al on their comparison1 of the 10-year interim results of NordICC2 to predictions from 3 well established colorectal cancer (CRC) screening models. The authors conclude that NordICC’s results are in line with modeling predictions. Our interpretation is different—that NordICC’s results differ considerably from model estimations. Full-Text PDF

Topics & Concepts