Leveraging Minimal Variables for Maximal Screening Yield: A Global Equity Perspective on Colorectal Cancer Risk Stratification

Colorectal cancer (CRC) ranks as the third most frequently diagnosed malignancy and the second leading cause of cancer‐related death worldwide [1]. Historically, the burden of CRC has been concentrated in high‐income countries; however, rapid westernization of lifestyles and population aging have driven a pronounced epidemiologic shift toward low‐ and middle‐income countries (LMICs) [2]. Although colonoscopy remains the gold standard for CRC screening, it is invasive, costly, and highly dependent on specialized personnel and infrastructure. As a result, screening uptake remains only 60%–70% in developed regions and falls below 5% in many LMICs [3]. More importantly, currently available risk stratification tools, such as the Asia‐Pacific Colorectal Screening (APCS) score [4] and European models incorporating fecal occult blood test (FOBT), fecal immunochemical test (FIT) or stool DNA testing [5], rely heavily on laboratory testing and reliable family history data. In LMICs, however, such resources are often unavailable or incompletely documented because of fragmented healthcare records and limited information accessibility. Against this backdrop of profound global inequity in screening access, Liu et al. [6] developed a predictive model based on multicenter data from asymptomatic individuals in Shandong, China. This model uses only six readily obtainable clinical variables (age, gender, smoking, drinking, chronic appendicitis, and hypertension) to identify advanced colorectal neoplasia (AN) in average‐risk populations. The model demonstrated acceptable discrimination (C‐statistic = 0.656) and calibration in internal validation. Moreover, by applying a simplified risk score, the authors successfully stratified participants into three risk categories and reduced the number needed to screen (NNS) from 28.9 in the low‐risk population to 8.7 in the high‐risk population. Beyond providing a nearly cost‐free prescreening strategy for resource‐limited settings, this study also raises an important methodological question: in the pursuit of advanced biomarkers and artificial intelligence–driven diagnostics, are we underestimating the predictive value of simple and accessible clinical variables?

The principal innovation of this model [6] lies in its deliberate emphasis on simplicity and broad applicability in variable selection. CRC screening in high‐income settings generally relies on FOBT/FIT or colonoscopy, both of which require supporting infrastructure, including laboratory services, cold‐chain logistics, and endoscopic facilities, resources that are often unavailable in primary care settings across LMICs [7]. By contrast, the proposed risk assessment can be completed using only six readily available variables, all of which can be collected through a brief interview by community healthcare workers without specialized equipment or consumable resources. This makes the model particularly suitable for settings in which endoscopic workforce density is below one physician per 100,000 population or where annual per capita healthcare expenditure remains below US$50. Notably, the inclusion of chronic appendicitis and hypertension as risk factors is of particular interest, as these readily identifiable clinical conditions may reflect biological processes that are relevant across diverse populations. Persistent local inflammatory activity associated with chronic appendicitis may contribute to colorectal neoplasia by disrupting intestinal microbial homeostasis and immune surveillance. Although the inflammatory–neoplastic transition has been well characterized in inflammatory bowel disease–associated CRC [8], its mechanistic relevance in the appendix as an immune‐regulatory organ remains to be clarified. Likewise, the association between hypertension and CRC may be mediated by shared biological pathways, including renin–angiotensin system–related proliferative signaling and chronic low‐grade inflammation [9, 10]. Evidence from a large Japanese nationwide cohort demonstrated that untreated hypertension was significantly associated with an increased risk of CRC. Compared with normotensive individuals, the hazard ratios were 1.07 (95% CI, 0.99–1.15) and 1.17 (95% CI, 1.08–1.28) for stage 1 and stage 2 hypertension, respectively. Notably, these associations remained evident even among metabolically healthy individuals [11]. Collectively, these findings suggest that, at a global level, hypertension and chronic appendicitis may serve as accessible clinical surrogates of the metabolic–inflammatory axis, thereby linking primary care practice with cancer screening. This feature is especially valuable in LMICs, where comprehensive cancer registries and accurate family history data are often lacking.

Despite its promise, the global implementation of this model faces several important challenges that merit careful consideration. The foremost concern is the lack of external validation. To date, the model has not been tested in diverse regions of China, including southern, northwestern, and ethnic minority populations, nor in cohorts from Africa, South Asia, and Latin America, where ethnic composition, dietary patterns, and infectious exposures differ substantially. Indeed, CRC epidemiology varies markedly across regions. For example, CRC in African populations is often characterized by younger age at onset, a higher prevalence of right‐sided lesions, and stronger associations with infectious etiologies such as schistosomiasis [12, 13], which may limit the transportability of models derived from eastern Chinese cohorts. Another point deserves clarification. Smoking was not statistically significant in the final multivariable logistic regression model, yet it was retained because it was consistently selected by Least Absolute Shrinkage and Selection Operator (LASSO) regression. This is not inherently contradictory, as LASSO prioritizes overall predictive performance at the model level rather than variable retention based solely on individual p values. Nevertheless, extensive evidence from international studies has consistently identified smoking as a major CRC risk factor. A meta‐analysis encompassing 106 studies found that smoking was associated with an 18% increased risk of CRC compared with never smoking, with risk increasing with smoking intensity and duration [14]. The lack of significance in the present study may reflect limited statistical power due to the relatively small number of AN‐positive cases, crude exposure categorization of smoking based only on a 30 pack‐year threshold without distinguishing current from former smoking, and potential confounding or effect modification related to sex disparities in smoking prevalence within the Chinese population. Therefore, future validation in heterogeneous populations should re‐examine the predictive contribution of smoking; if justified, smoking may be retained as a mandatory variable to preserve generalizability and clinical utility.

A further limitation is the relatively low prevalence of chronic appendicitis in this cohort (approximately 0.86%) [6], which may reduce coefficient stability and lead to variability in predictive performance across settings with different appendicitis burdens. In addition, the model does not include family history of CRC, a well‐established risk factor consistently emphasized in international guidelines [15]. In well‐resourced healthcare systems with reliable documentation and well‐informed patients, omission of family history may reduce the model's ability to identify individuals with clearly elevated baseline risk. However, the intended application of this model is primary care in LMICs, where collection of family history is often compromised by incomplete records, limited familial knowledge, and low patient awareness. In this context, omitting family history may represent a pragmatic compromise between predictive performance and real‐world feasibility. Even so, the scope of application should remain clearly defined. Individuals for whom reliable family history information is available should undergo further risk assessment in accordance with established recommendations rather than relying exclusively on this model.

Notwithstanding these limitations, the model occupies a meaningful niche within global CRC screening strategies. Rather than replacing FOBT/FIT or colonoscopy, it should be viewed as a first‐line risk‐filtering instrument that supports preliminary triage in resource‐constrained settings. The reduction in NNS from 28.9 to 8.7 indicates that screening approximately nine high‐risk individuals would identify one case of advanced colorectal neoplasia, a performance approaching that of FIT‐based programs in high‐income settings but at minimal cost [16]. In practical terms, the model could be integrated into World Health Organization (WHO)‐aligned primary healthcare systems, enabling community health workers to input the six clinical variables via mobile platforms during home visits and obtain immediate risk estimates. Low‐risk individuals could receive lifestyle counseling and routine annual follow‐up; intermediate‐risk individuals could be prioritized for FIT‐based screening or closer surveillance; and high‐risk individuals could be referred directly to tertiary centers for colonoscopic evaluation. Such a tiered strategy would promote more efficient allocation of scarce healthcare resources, reduce unnecessary invasive testing among low‐risk individuals, and improve both the cost‐effectiveness and benefit–risk balance of population‐based screening [17].

Looking ahead, this study highlights three pressing priorities for global CRC screening efforts. First, robust external validation in multiethnic and multinational populations is essential. Ideal validation cohorts should include regions such as Africa, South Asia, Latin America, and the Middle East, where CRC incidence is rising and screening infrastructure remains limited. Additional validation in independent cohorts and assessment of integration with FOBT/FIT‐based screening would help define the model's role in routine practice more precisely. Second, further research should clarify the interplay between metabolic and inflammatory mechanisms in colorectal carcinogenesis, particularly potential ethnic differences in the metabolic–inflammatory axis reflected by hypertension and chronic appendicitis. Future studies could incorporate inflammatory markers such as C‐reactive protein, interleukin‐6, and tumor necrosis factor‐α, together with metabolic indicators such as fasting glucose and insulin resistance indices [18, 19], to build hybrid models that combine clinical variables with biological markers and thereby enhance both predictive accuracy and mechanistic insight. Third, future work should examine whether machine learning methods, including random forests, gradient boosting, and neural networks, can better capture complex nonlinear relationships among predictors. Conventional logistic regression assumes linear and independent effects, yet interactions may exist between chronic appendicitis and hypertension or between alcohol consumption and sex. By identifying such latent patterns, machine learning approaches may yield further gains in predictive performance.

Overall, this study offers an important lesson for global CRC screening: in the pursuit of increasingly sophisticated technologies, readily obtainable clinical information from primary care settings should not be overlooked. For billions of individuals worldwide who remain beyond the reach of colonoscopy and FOBT/FIT‐based screening, a simple six‐item questionnaire may represent an accessible gateway to earlier detection, embodying a meaningful intersection between equity and precision in global cancer control.

Author Contributions

Shulun Nie: writing – original draft, conceptualization. Cunliang Wang: writing – review and editing. Xiaolan Cao: writing – review and editing. Shuyi Song: investigation. Yu Wang: writing – original draft, conceptualization.

Conflicts of Interest

The authors declare no conflicts of interest.