Construction and Validation of a Novel Forecasting Nomogram to the Risk of Colorectal Adenomas: Preventing Colorectal Cancer at Its Origin

Colorectal cancer ranks as the third most prevalent cancer and the second leading cause of cancer-related death worldwide (1). In recent years, the incidence and mortality rates of colorectal cancer in China have increased owing to economic development and lifestyle changes (2). In 2020, newly diagnosed colorectal cancer cases in China accounted for 28.8% of all new cases globally (555,477/1,931,590), and colorectal cancer–related deaths accounted for 30.6% of all colorectal cancer–related deaths worldwide (286,162/935,173; ref. 3). With the increasing understanding of the biological development of colorectal cancer, targeted therapies directed against key oncogenes and pathways have emerged, but these treatments are only effective in a small proportion of patients with colorectal cancer (4).

Approximately 60% of colorectal cancers result from the adenomatous pathway, which progresses from nonadvanced tubular adenoma to villous tubular adenoma and potentially to invasive carcinoma (5). Colonoscopy is an effective tool to prevent colorectal cancer; however, a previous study conducted in China revealed that only 14% (25,593/182,927) of high-risk participants with colorectal cancer underwent colonoscopy, suggesting that colonoscopy is not yet well-accepted in the general population (6). Current colonoscopy resources in China do not accommodate comprehensive colonoscopy screening for all people of appropriate age, and colonoscopy may miss approximately one-quarter of colorectal adenomas, which greatly increases the potential cancerous risk of colorectal adenomas (7).

Previous studies have identified the following as hypothesized factors to influence the risk of colorectal adenomas occurrence: sex (8), age (8), body mass index (BMI; ref. 9), obesity (8–10), smoking (11, 12), drinking (12, 13), red meat consumption (11), dietary fiber (14), physical activity (8), and diabetes mellitus (8).

With a deeper understanding of colorectal cancer and precancerous lesions, early screening of high-risk groups has become the emphasis of colorectal cancer prevention and treatment. Although research on colorectal adenoma risk factors is extensive, there are fewer studies on risk probability prediction based on risk factors. In this study, the colorectal adenoma risk nomogram constructed based on lifestyle factors is easier to be extended for early screening of colorectal adenoma, and patients with high-risk colorectal adenomas can receive timely colonoscopies for the prevention of colorectal cancer.

The study cases were obtained from participants who underwent complete colonoscopy at the Second Hospital of Jilin University from 2021 to 2022. These subjects underwent a complete colonoscopy for diarrhea, abdominal pain, dark stool, and other digestive symptoms or physical examination. The pathologic examination of polyps was completed by experienced pathologists. All included participants underwent colonoscopy for the first time. The exclusion criteria for cases were as follows: (i) age less than 16 years old; (ii) the presence or a history of any type of malignancy; (iii) previous colorectal surgery; (iv) a family history of colorectal cancer; (v) previous or current detection of inflammatory bowel disease; (vi) the presence or a history of severe cardiovascular disease; and (vii) pathologic type of polyps other than colorectal adenomas. A total of 1,133 participants were eligible for analysis. The study was conducted in accordance with the Declaration of Helsinki. The study was approved by the Ethics Committee of the Second Hospital of Jilin University. We obtained written informed consent from each participant. Only study-related data were used in this study, and identifiable personal information was removed by the researchers.

Data on physical factors, family history, personal disease history, diet, and lifestyle habits were collected by the investigators from the participants via telephone callback. Physical factors included age, sex, height, weight, and BMI (calculated from height and weight). Comorbidities included diabetes mellitus and hypertension. Family history mainly included a family history of colorectal cancer. Personal disease history was used to exclude participants who had any kind of malignancy or inflammatory bowel disease or who underwent colorectal surgery. Diet and lifestyle habits included smoking status, smoking duration, and quantity of cigarettes smoked; alcohol consumption status, alcohol consumption duration, and amount of consumed alcohol; spicy diet; high-fat diet; consumption of pickled products, vegetables, and fruits; regular aspirin use; regular NSAID use; water intake; physical activity; sleep duration; and defecation.

Smoking status was categorized into two types: smoking (continuous smoking for more than 2 years with an average of at least two cigarettes per day, excluding those who had quit smoking) and nonsmoking (never smoked). Alcohol consumption status was categorized into two types: drinking (continuous drinking for more than 2 years with an average daily intake of more than 20 g of ethanol, excluding those who had quit) and nondrinking (never drinking or with an average daily intake of less than 1 g of ethanol). The frequency of each dietary intake indicated the subject's dietary habits, categorizing spicy diets, high-fat diets, pickle consumption, and vegetable and fruit consumption into three separate statuses: negligible, occasionally (no more than 2 days per week), and frequently (at least 3 days or more per week). Regular use of aspirin or NSAIDs meant that the subject had taken them on average at least once a week for at least the past 5 years or continuously for at least 2 months each year for at least the past 5 years. The prevalence status of hyperglycemia and hypertension was divided into two groups: present and absent. Water intake status was divided into three groups: less than 500 mL/day, 500 to 1,500 mL/day, and more than 1,500 mL/day. Physical activity status was divided into three groups: negligible, occasional, and frequent (moderate intensity exercise at least once a week for at least 30 minutes). Sleeping time was divided into three groups: less than 6 hours/day, 6 to 8 hours/day, and more than 8 hours/day. Defecation was divided into two groups: regular (almost daily bowel movements at a fixed time) and irregular (daily but irregular or nondaily bowel movements). The collected information was collated, and all subjects were grouped after screening based on inclusion and exclusion criteria.

The predictive model was built by logistic regression analysis. We assessed health factors influencing the occurrence of colorectal adenomas in the training cohort by univariate stepwise analysis. The independent variables in the multiple logistic regression analysis were health factors that were significantly associated with the risk of colorectal adenomas in the univariate analysis. Based on the results of multivariate logistic regression analysis, we established a nomogram for predicting the occurrence risk of colorectal adenoma. We evaluated the probability of consistency between the predicted results of the model and the real situation through the plotting of the ROC curve and the C-index. Generally, we consider that the model has no predictive ability if the C-index is less than 0.5, and the model has weak or strong predictive ability if the C-index is between 0.5 and 0.7 or greater than 0.7.

We evaluated the consistency of the training cohort and validation cohort predicted probabilities with the actual results by calibration curves. The clinical benefit of the model at different decision thresholds was predicted by the decision curve analysis (DCA) curve.

We used SPSS, RRID:SCR_002865, and R4.2.2 software for statistical analysis. Quantitative variables that followed a normal distribution were expressed as (x̄ ± s) and analyzed using independent Student t tests. Categorical variables were expressed as frequencies and percentages (n%) and were analyzed using the χ² test. The reported levels of statistical significance were two-sided, with significance less than 0.05 considered statistically significant.

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

A total of 1,518 subjects participated in this study, of which 385 were excluded by exclusion criteria. After the screening, 792 participants were included in the training cohort and 341 in the validation cohort (Fig. 1). The baseline clinical characteristics of subjects in the colorectal adenoma and no adenoma groups are shown in Supplementary Table S1. Baseline clinical characteristics of the training and validation cohorts are shown in Supplementary Table S2. There was no statistical difference in the characteristics of the patients in the training and validation cohorts (P > 0.05).

In the training cohort, 792 participants were divided into colorectal adenoma group (433) and no adenoma group (359). All variables were analyzed by univariate analysis, and statistically significant variables were included in the multivariate analysis. The following six variables were finally identified as independent factors influencing the occurrence of colorectal adenomas: advanced age, male, smoking, drinking, frequent consumption of pickles, and irregular defecation (Table 1). We constructed a nomogram forecasting the occurrence risk of colorectal adenoma from these six variables (Fig. 2). Each factor had a different score depending on the variable, and the score for each participant was the sum of the scores for all factors. The predicted probability of the occurrence risk of colorectal adenoma is obtained by plotting a vertical line from the total score line to the risk line. The greater the total score, the higher the risk of colorectal adenoma. The ROC curve analysis of the subjects showed that the AUC of the model was 0.778 (95% confidence interval, 0.746–0.810). Meanwhile, the efficacy of the nomogram at the cutoff point showed 74.4% specificity, 68.4% sensitivity, 76.3% positive predictive value, and 66.1% negative predictive value. This indicates that the model has good discrimination ability (Fig. 3).

To validate the applicability of the nomogram to other data sets, we performed a validation using data from 341 subjects in the validation cohort. The validation cohort included 169 subjects without adenomas and 172 subjects with colorectal adenomas. The C-index for the validation cohort was 0.754 (95% confidence interval, 0.703–0.805). Meanwhile, the efficacy of the validation cohort at the cutoff point showed 66.3% specificity, 69.9% sensitivity, 66.9% positive predictive value, and 66.3% negative predictive value, suggesting that the nomogram also showed good predictability in the validation cohort. The calibration curves for the training and validation cohorts illustrated that the nomogram also showed a high degree of agreement between the predicted probability of colorectal adenoma and the true situation (Fig. 4). DCA demonstrated that the net benefit of intervention by the model was significantly better than if all subjects had either the intervention or no intervention, suggesting that there is a higher benefit to using the model for the prediction of the risk of colorectal adenoma occurrence (Fig. 5).

Adenomatous polyps are thought to be the precursor lesions for most colorectal cancers and mainly develop via the adenoma–carcinoma sequence (5, 15). Therefore, early detection and endoscopic removal of newly occurring adenomas is an important strategy to prevent colorectal cancer. In our study, we combined contemporary lifestyle habits and geographic dietary characteristics to evaluate possible risk factors for colorectal adenomas by classifying risk factors into physical factors, lifestyle factors, comorbidities, and medications.

We found gender and age to be nonmodifiable factors in the development of colorectal adenomas in our study, and this association remained significant after multivariate analysis. This may be because men are more prone to unhealthy lifestyle behaviors such as smoking, drinking alcohol, or consuming more fat than women, which are risk factors for colorectal adenomas (12). Estrogen reduces the risk of colorectal adenomas in women to some extent. Female hormones inhibit the formation of colorectal adenomas by binding to the estrogen receptor to lower secondary bile acid levels and by downregulating the expression of insulin-like growth factor-1 (16).

Several studies have explained the association between tobacco and colorectal polyps, given that tobacco contains numerous carcinogens that can damage the intestinal mucosa, leading to polyp development (17). Smoking is a well-known risk factor for colorectal polyps and colorectal cancer (18). Relative to nonsmokers, smoking significantly increased the risk of colorectal polyps. Some studies suggest that the duration and quantity of smoking showed a rough dose–response relationship with the risk of disease. The risk of colorectal polyps increased with increasing duration and quantity of smoking (19, 20). A previous study reported that the prevalence of polyps was 2.01 times higher in drinkers than that in nondrinkers (21). Some researchers believe that alcohol disrupts the stable state of the flora in the gut when the body is healthy, resulting in intestinal damage and the depletion of bacteria with anti-inflammatory activity (22). Our study also demonstrated that the occurrence risk of colorectal adenomas is associated with alcohol consumption.

In many areas of the world, due to the climatic conditions and geographical location, people generally prefer a high-fat, high-salt diet. Vegetables are stored for a longer time by salting them. We also investigated the effect of pickled product consumption on the occurrence risk of colorectal adenomas. The process of pickling vegetables requires the addition of a large amount of salt or sauce, which produces nitrites. Nitrites react with amines, amides, and other nitroso precursors to form N-nitroso compounds, which are associated with a high risk of colorectal and gastrointestinal cancers (23). Some studies have revealed that the risk of colorectal adenomas was 1.487 times higher with frequent pickle consumption than that without pickle consumption (24). Our findings are consistent with it.

Some studies suggest that dietary fiber prevents polyp formation by swelling the stool and enhancing bowel movements, thereby possibly reducing the colonic surface area exposed to cancer-causing toxins and bile acids in the stool (25). Our study explored the relationship between regular daily defecation or moderate water intake and the occurrence risk of colorectal adenomas. The researchers suggest that regular daily defecation or moderate water intake may exert an effect similar to that of dietary fiber, decreasing the residence time of fecal toxins in the colorectum and reducing their absorption. The results also confirm that regular defecation reduces the occurrence risk of colorectal adenomas. However, moderate water intake was not significantly associated with the risk of colorectal adenoma.

BMI is an international measure of obesity and health status, and many studies have discovered a positive correlation between BMI and the risk of developing colorectal adenoma (9). Abdominal obesity has a stronger association with the risk of colorectal adenoma than BMI (26). However, in our study, although the mean BMI and the frequency of high-fat diet were higher in the colorectal adenoma group than those in the no adenoma group, there was no statistical significance between the two groups. The relationship between red meat consumption and colorectal cancer has been investigated in several studies, and there is consensus that processed red meat consumption increases the risk of colorectal cancer and colorectal adenomas. However, there is a lack of consensus on the risk associated with unprocessed red meat consumption. The association between red meat and colorectal adenomas was not investigated in this study. This is because the distinction between processed and unprocessed red meat is difficult to define, and the health effects of each type of red meat may differ significantly. Therefore, without strict categorization, it may not be reasonable to group processed and unprocessed red meat (27).

Diabetes and hypertension increase the risk of colorectal adenoma (28). The use of aspirin or NSAIDs has a protective effect on the risk of colorectal adenoma (29, 30). However, the results confirmed that these potential risk factors did not show a significant correlation, probably because fewer participants had these characteristics.

Our nomogram is designed to screen people for high-risk colorectal adenomas and to guide people to healthy living, so they should be limited in the number of variables included in the nomogram. Excessive variables can make calculating predictive scores cumbersome and incomprehensible to people. Our nomogram showed good predictive power in the training cohort. The calibration curves showed similar distributions with the ideal reference line, and the DCA showed that relying on this model for prediction yielded better results at any threshold. Applying this model to the validation queue also shows good predictive power in its calibration curve and decision analysis curve. Thus, our nomogram can be applied to predicting the occurrence risk of colorectal adenomas.

Compared with other colorectal adenoma risk prediction models, this model has well predictive accuracy as well as excellent calibration and clinical utility. In an American study, a clinical scoring system was constructed to predict the risk of colorectal adenoma based on age, sex, family history of colorectal cancer, and smoking history, which had a C-index of 0.64 (31). Spanish scholars examined 88 SNPs associated with the risk of colorectal cancer and finally identified five SNPs associated with the occurrence of colorectal adenomas for calculating genetic risk scores. Combining sex and age with genetic risk scores to construct a colorectal adenoma risk prediction model was useful for colorectal adenomas risk stratification and screening program adjustment, and the C-index of this model was 0.655 (32). Researchers in China proposed a tool to predict the risk of colorectal adenomas based on advanced age, male, hyperlipidemia, smoking, high red meat intake, high salt intake, high dietary fiber intake, Helicobacter pylori infection, nonalcoholic fatty liver disease, and chronic diarrhea, which has a favorable discriminatory power, with a C-index of 0.775 (33).

However, there are some limitations to this study. First, collecting information on each participant’s daily diet or energy intake in a retrospective case–control study was restrictive, and participants were limited in providing realistic descriptions of health factors, which is prone to recall bias. Second, our study did not include related diseases such as nonalcoholic fatty liver disease and H. pylori infection as study factors. Because the subjects were all outpatients of our hospital, it was not possible to judge the authenticity of the illnesses, which may cause inclusion bias. The judgment of whether one has hyperglycemia or hypertension is easier and can reduce the risk of bias. Finally, the sample size of this study was narrow and we did not investigate the economic income and education level of the participants. Better-off and better-educated people are generally more health-conscious. Therefore, the risk of colorectal adenomas may be relatively lower. We can reduce model prediction bias by increasing the sample size to cover more population types (34).

In conclusion, we exploited a simple and reliable model for predicting the occurrence risk of colorectal adenoma based on dietary and lifestyle factors. The model has good predictive accuracy and clinical applicability, which is relevant for identifying patients with asymptomatic colorectal adenomas and selecting high-risk groups for colonoscopy screening.

No disclosures were reported.

M. Li: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, project administration. M. Cui: Data curation, formal analysis, supervision. X. Zhou: Data curation, formal analysis. Y. Song: Conceptualization, supervision, project administration, writing–review and editing.

The authors would like to thank all the staff of the Gastroenteric Medicine and Digestive Endoscopy Centre of the Second Hospital of Jilin University for their help in the data collection of this study.