Colon cancer is one of the most common and deadly cancers worldwide, yet it is also one of the most preventable. Before colon cancer develops, the colon typically goes through a series of earlier changes - precancerous lesions such as polyps and adenomas - that can be detected and removed before they become cancer. Catching the disease at this early stage dramatically improves survival.
Colonoscopy is the most effective tool for finding and removing these lesions. However, it is an invasive procedure requiring sedation, bowel preparation, and specialist time. It is also expensive and carries a small risk of complications. Performing colonoscopy on every adult over a certain age is impractical, and many people avoid it due to discomfort or inconvenience.
What if there were a simple, inexpensive way to identify which individuals are at highest risk - and therefore most likely to benefit from colonoscopy - using only information that can be gathered through a basic questionnaire and a weight measurement? This study set out to build exactly that tool: a nomogram, a visual risk-scoring chart based on modifiable lifestyle factors, and to test whether it works not just in one local population but across different countries and cultures.
Although colon and rectal cancers are often grouped together as colorectal cancer (CRC), this study made the specific choice to focus only on colon cancer and precancerous lesions in the colon. This distinction matters because colon and rectal cancers differ in where they arise, their embryonic development, and importantly, how they are influenced by lifestyle factors.
High-fat diets and metabolic conditions like obesity appear to have a stronger relationship with colon cancer specifically. By focusing on colon cancer, the researchers aimed to build a more precise, colon-specific model rather than a one-size-fits-all tool that might be less accurate for either cancer type individually.
Precancerous lesions covered in this study include conventional adenomas (such as tubular, tubulovillous, and villous adenomas), serrated adenomas, and dysplasia associated with inflammatory bowel disease. These are the recognized stepping stones from a healthy colon to frank cancer, and detecting and removing them - before they become malignant - is the primary goal of colon cancer screening programs.
The researchers used two entirely separate patient groups. The modeling cohort consisted of 400 patients who underwent colonoscopy at a hospital in Xinjiang, China between 2021 and 2024: 272 with normal findings and 128 who were diagnosed with colon cancer or precancerous lesions. Only non-invasive, easily obtainable information was used - no blood tests or genetic analysis were required.
The validation cohort came from a completely different source: the National Health and Nutrition Examination Survey (NHANES), a large U.S. population study. From the 2021-2023 NHANES data cycle, 284 individuals were identified: 191 healthy and 93 with self-reported colon cancer or related conditions. Testing the model in a different country on a different population is a stringent way to evaluate whether it will work broadly, not just in the specific setting where it was developed.
The predictors collected were: age (split at 55 years), body mass index or BMI (split at 25 kg/m2), smoking history, alcohol consumption, and high-fat diet (defined as fat providing more than 30% of total daily calories). These were deliberately chosen because they can be assessed without any laboratory equipment - just questions and a scale. A multivariable logistic regression analysis was then used to determine which factors independently predicted colon cancer or precancerous lesions.
Statistical analysis confirmed that five factors were each independently associated with colon cancer or precancerous lesions: age over 55, BMI over 25 kg/m2 (overweight), smoking history, alcohol consumption, and a high-fat diet. Crucially, none of these is a fixed, unalterable characteristic - all are either modifiable behaviors or reflect when to start screening.
Factors that were tested but did not independently predict risk included gender, education level, history of gallbladder surgery, high blood pressure, diabetes, high-fiber diet, and several blood test values including cholesterol, hemoglobin, platelet count, and fasting blood glucose. This result simplifies the tool considerably - it does not require blood tests.
In the disease group, 48% of patients were over 55, 69% had a BMI over 25, 53% had a smoking history, 55% drank alcohol, and 56% consumed a high-fat diet. All of these proportions were significantly higher than in the healthy group, confirming these factors cluster together in people who develop colon pathology.
The five predictors were combined into a nomogram - a graphical tool where each risk factor is represented as a scale with a point value. A clinician or patient simply identifies their status on each scale, adds up the points, and reads off a corresponding probability of having colon cancer or a precancerous lesion. This visual format makes risk calculation fast, intuitive, and usable without a computer.
In the modeling cohort, the nomogram achieved an AUC (area under the receiver operating characteristic curve) of 0.765. This measure ranges from 0.5 (no better than chance) to 1.0 (perfect prediction). A value of 0.765 indicates solid, clinically meaningful discrimination - the model correctly identifies higher-risk individuals substantially better than random chance.
Calibration - how well predicted probabilities match actual outcomes - was also good. The Hosmer-Lemeshow goodness-of-fit test showed no significant deviation between predicted and observed outcomes. Decision curve analysis showed that using the nomogram to guide colonoscopy referrals provided greater net clinical benefit than either referring everyone or referring no one, specifically when the risk threshold exceeded 15%.
The most rigorous test of any predictive model is whether it works in a population it was never trained on. Despite the considerable differences between the Xinjiang Chinese cohort (clinical diagnoses, pathologically confirmed) and the American NHANES cohort (self-reported diagnoses, entirely different ethnic and geographic background), the model performed remarkably consistently.
In the NHANES validation cohort, the AUC was 0.761 - virtually identical to the 0.765 achieved in the development cohort. A statistical test confirmed there was no significant difference between these two AUC values (p = 0.899). Calibration also remained acceptable, with the Hosmer-Lemeshow test showing borderline results (p = 0.054) that were judged acceptable given the sample size.
Subgroup analyses confirmed that the five risk factors remained significant in both men and women separately, and adding gender to the model did not meaningfully improve it. This confirms the tool's broad applicability across different demographic groups. One notable adjustment: the optimal risk threshold was 15% in the Chinese cohort but 11% in the NHANES cohort, reflecting the different baseline risk in each population - a reminder that threshold calibration may be needed when applying the model in new settings.
Age raises risk because DNA repair becomes less efficient over time and cumulative exposure to carcinogens increases. Interestingly, early-onset colon cancer in adults under 50 is rising, linked to modern sedentary, high-calorie lifestyles - suggesting age-related risk is not just about biology but also lifestyle accumulated over decades.
Excess body weight (BMI over 25) promotes cancer through multiple pathways. Adipose (fat) tissue releases inflammatory signals like TNF-alpha and IL-6 that create a pro-cancer environment. Obesity also causes insulin resistance and elevated IGF-1 signaling, which stimulates cell growth and proliferation. Smoking introduces carcinogens directly into the colon lining, while alcohol generates acetaldehyde that damages DNA and impairs the cellular machinery that normally fixes genetic errors.
High-fat diet alters the gut microbiome, expanding populations of bacteria like Escherichia coli and Bacteroides fragilis that produce toxins damaging the intestinal lining. High fat also promotes obesity and metabolic syndrome, compounding the cancer risk. The fact that all five factors are potentially modifiable means this model doubles as a practical guide for colon cancer prevention: reducing weight, quitting smoking, moderating alcohol, and improving diet could measurably lower an individual's risk score.
The nomogram's key clinical advantage is simplicity. It requires no blood tests, no imaging, no genetic testing - just answers to five questions and a BMI calculation. This makes it practical for primary care settings, community health screenings, and low-resource environments where expensive diagnostic infrastructure is unavailable. It could serve as a first-line filter to identify who most urgently needs colonoscopy referral.
However, the study's limitations are important to acknowledge. The model omits family history of colorectal cancer, a well-known risk factor, because it was not available in the modeling dataset. This could cause the model to underestimate risk in genetically predisposed individuals. The NHANES validation relied on self-reported diagnoses, which are less precise than pathological confirmation - and people are more likely to recall a cancer diagnosis than asymptomatic precancerous polyps, potentially skewing the validation toward more advanced disease.
The sample sizes in both cohorts are modest, and the model has not yet been tested in prospective clinical settings to confirm it actually changes outcomes. The authors call for larger, multi-center, prospective studies - ideally using histologically confirmed precancerous lesion data - to fully validate the nomogram before it is adopted for routine use. They also emphasize that the optimal risk threshold (the cut-off above which colonoscopy is recommended) will likely need local calibration in each new population setting.
This study developed and validated a non-invasive, five-factor nomogram for predicting individual risk of colon cancer and precancerous lesions. Based entirely on age, BMI, smoking, alcohol, and dietary fat - information gathered through a simple interview and a scale - the model achieved AUC values around 0.76 in both a Chinese clinical cohort and a U.S. population survey.
The consistency of results across two very different populations suggests that modifiable lifestyle factors exert a fundamental, universal influence on colon cancer risk that transcends geography and ethnicity. This is both scientifically interesting and practically encouraging: it means a relatively simple tool may be broadly applicable.
The ultimate goal is integration into population-level colorectal cancer screening programs - allowing risk-stratified screening where high-risk individuals get colonoscopy sooner and more frequently, while lower-risk individuals may be safely monitored less intensively. This approach could make limited colonoscopy resources go further, reduce diagnostic delays, and ultimately detect more cancers at earlier, more treatable stages.