Beyond simple adjustment: integrating endoscopic screening into the natural history of colorectal cancer risk models

Colorectal cancer (CRC) prevention has been transformed over the last decades through the widespread adoption of endoscopic screening, and the detection and removal of precancerous polyps. Yet, integrating these interventions into epidemiological risk models remains a persistent challenge. In this issue of Cancer Epidemiology, Biomarkers & Prevention, Wei and colleagues (1) present a nuanced methodological framework that moves beyond simple covariate adjustment. Their work offers a clearer understanding of how screening coverage and adenoma resection interacts with traditional risk factors to shape CRC incidence.

Historically, most CRC risk models have treated endoscopic screening as a confounding variable to be controlled for in a multivariable model (2). However, as the authors rightly argue, screening is not merely a confounder; it is a fundamental disruptor of the disease’s natural history. By removing adenomas, screening essentially resets the “biological clock” of carcinogenesis. This creates a heterogenous population where the influence of lifestyle and genetic risk factors may manifest differently depending on whether an individual has recently undergone an endoscopy.

The researchers categorize the study population into “screen-covered” (SC), defined as having a colonoscopy within the past 10 years, and “not screen-covered” (NSC), and estimate risk factor associations separately for these two distinct situations. Also, among the SC population, the authors accounted for high-risk adenoma resection, trying to differentiate within each risk factor the adenoma-mediated effect and the direct effect on CRC.

Another methodological strength of this work is the development of a fully adjusted hazard ratio (HR) estimate, which uses a weighted average of the log-transformed HRs from the SC and NSC models, allowing researchers to generate a single, screening-adjusted risk estimate that represents the actual population’s screening behaviour. This approach departs from “standard” models that assume risk factor effects remain static regardless of screening status, by acknowledging that the presence of screening modifies the relationship between the exposure and the outcome (Table 1).

The relevance of the study by Wei et al. lies in its recognition that the magnitude of established risk factors, such as age, smoking, body mass index (BMI), or family history, is generally stronger among the NSC population, which suggests that the true biological effect of a risk factor is best observed when the natural progression of the disease is not interrupted by clinical intervention. Relying on risk estimates derived from populations with high screening rates without accounting for this modification risks underestimating the true impact of lifestyle factors. For instance, the HR for age became notably stronger after full adjustment, moving from an unadjusted HR of 1.59 [95% confidence interval (CI): 1.48–1.71] to a fully adjusted HR of 1.81 (95% CI: 1.67–1.97). This shift suggests that because older individuals are more likely to be screened and have high-risk adenomas removed, their observed risk in simpler models is artificially suppressed. Conversely, the protective association of current menopausal hormone therapy (MHT) became weaker after full adjustment. The HR for estrogen-only therapy moved to 0.72 (95% CI: 0.60–0.87) compared to an unadjusted 0.62 (95% CI: 0.54–0.73). This indicates that a portion of the “protection” previously attributed to MHT likely reflects the higher screening propensity among MHT users.

Furthermore, the study provides empirical evidence for the clinical impact of screening adherence on cancer staging. Those diagnosed during NSC time were significantly more likely to present with advanced disease; 14.9% of NSC cases were stage 4, compared to 11.4% of SC cases (P=0.048). This underscores the premise that regular screening not only prevents cancer but shifts the remaining diagnoses toward earlier, more treatable stages.

While the study benefits from high-quality assessments, the unique characteristics of the Nurses’ Health Study (NHS) cohort must be considered (3). The cohort is comprised of health professionals who are predominantly white and inherently more health conscious. This ‘healthy volunteer’ effect suggests a level of health literacy and preventive adherence (inherent to health professionals) that may not map directly onto the general population, potentially skewing these weighted estimates.

The study by Wei et al. (1) establishes a framework for how to handle screening in older populations, but it also highlights the “blind spot” regarding populations where screening has not even begun. We are seeing a global spike in CRC among individuals under the age of 50 years (4), a demographic for whom screening history is often non-existent and where risk factors potentially differ from those captured in older populations, involving the microbiome and ultra-processed diets.

From a clinical perspective, this work aligns with recent findings from the German screening registry, which suggest that the 10-year interval for repeat colonoscopy might even be safely extended for individuals with a negative baseline result (5). However, this approach has several limitations. Firstly, we must confront the fact that a significant portion of interval CRCs are not the result of aggressive new biology, but of lesions that were simply missed during the initial procedure (6). Secondly, the model in the current study excludes serrated polyps. Given that the serrated pathway accounts for a significant proportion of interval cancers and may be less effectively interrupted by colonoscopy compared to the conventional adenoma pathway (7), integrating serrated histology is a requirement for future iterations of this model. Lastly, to truly implement these models in clinical practice, we will require to feed them with data on the quality of bowel preparation, and adenoma and serrated polyp detection rates.

Finally, the different modalities and varying sensitivities of screening programs, like faecal immunochemical tests (FIT), sigmoidoscopy, or multi-target stool DNA tests (8), must be considered. Next-generation risk models must differentiate between screening modalities: a 2-year FIT “coverage” window yields a risk-resetting effect distinct in duration and magnitude from the 10-year window of a high-quality colonoscopy.

Moving away from a static “10-year window” to a “dynamic risk score” based on both the modality and quality of the last screening procedure is the next logical step. By reclassifying screening as a modifier of the natural history rather than a mere confounder, Wei and colleagues have cleared a path for more accurate, biologically grounded risk prediction. In the pursuit of precision prevention, this level of methodological rigor is no longer optional, it is essential.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Annals of Cancer Epidemiology. The article did not undergo external peer review.

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://ace.amegroups.com/article/view/10.21037/ace-2026-1-0003/coif). R.J. serves as the consultant for MSD, Norgine, Alpha Sigma, GI Supply. The other author has no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Wei EK, Wu K, Colditz GA, et al. Accounting for Endoscopic Screening in Colorectal Cancer Risk Models. Cancer Epidemiol Biomarkers Prev 2025;34:2049-57. [Crossref] [PubMed]
2. Wei EK, Colditz GA, Giovannucci EL, et al. A Comprehensive Model of Colorectal Cancer by Risk Factor Status and Subsite Using Data From the Nurses' Health Study. Am J Epidemiol 2017;185:224-37. [Crossref] [PubMed]
3. Bao Y, Bertoia ML, Lenart EB, et al. Origin, Methods, and Evolution of the Three Nurses' Health Studies. Am J Public Health 2016;106:1573-81. [Crossref] [PubMed]
4. Siegel RL, Wagle NS, Cercek A, et al. Colorectal cancer statistics, 2023. CA Cancer J Clin 2023;73:233-54. [Crossref] [PubMed]
5. Heisser T, Kretschmann J, Hagen B, et al. Prevalence of Colorectal Neoplasia 10 or More Years After a Negative Screening Colonoscopy in 120 000 Repeated Screening Colonoscopies. JAMA Intern Med 2023;183:183-90. [Crossref] [PubMed]
6. Pohl H, Robertson DJ. Colorectal cancers detected after colonoscopy frequently result from missed lesions. Clin Gastroenterol Hepatol 2010;8:858-64. [Crossref] [PubMed]
7. Rex DK, Ahnen DJ, Baron JA, et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012;107:1315-29; quiz 1314, 1330.
8. Imperiale TF, Ransohoff DF, Itzkowitz SH, et al. Multitarget stool DNA testing for colorectal-cancer screening. N Engl J Med 2014;370:1287-97. [Crossref] [PubMed]