A recent study established noninvasive prediction models for colorectal adenoma risk, potentially improving screening efficiency and enabling earlier identification of high-risk patients.
In the study, published in BMJ Open Gastroenterology, lead study author Yi-Lu Zhou and colleagues analyzed data from 9,747 participants who underwent colonoscopy across three cohorts. The study aimed to enhance colorectal cancer (CRC) prevention by integrating lifestyle and gut microbiota data into colorectal adenoma (CRA) risk prediction.
A multivariable logistic regression model based on 14 variables showed strong discriminatory ability, with a c-statistic of 0.79 (95% confidence interval [CI] = 0.75–0.82). Machine learning models—random forest and gradient boosting—demonstrated similar performance, each with a c-statistic of 0.78. When microbial biomarkers such as Fusobacterium nucleatum and polyketide synthase (pks)-positive Escherichia coli were included, model performance improved further, with c-statistics ranging from 0.84 to 0.86.
“The integrated mathematical modeling incorporating lifestyle parameters and gut microbial signatures provides an effective noninvasive strategy for CRA risk stratification,” the study authors said. “The accompanying machine learning–assisted prediction application enables cost-effective, population-level screening implementation to optimize CRC prevention protocols,” they added.
Based on age-specific CRA incidence data, the investigators recommended initiating colonoscopy screening at age 42 years for high-risk individuals and 53 years for those at lower risk.
Significant lifestyle and metabolic risk factors were identified, including male sex (odds ratio [OR] = 1.81, 95% CI = 1.30–2.53, P = .001), hypertension (OR = 2.66, 95% CI = 1.84–3.83, P = 4.87 × 10⁻⁶), smoking (OR = 1.68, 95% CI = 1.36–2.07, P = 1.40 × 10⁻⁶), alcohol use (OR = 1.44, 95% CI = 1.16–1.77, P = .001), and consumption of processed meat (OR = 1.48, 95% CI = 1.14–1.93, P = .003). Protective associations were observed with high intake of vegetables and fruits (OR = 0.50, 95% CI = 0.36–0.68, P < .001), cereals (OR = 0.62, 95% CI = 0.45–0.86, P = .004), and increased walking time (OR = 0.79, 95% CI = 0.65–0.97, P = .021).
Microbiota profiling revealed significantly higher fecal abundance of F nucleatum and pks-positive E coli in patients with CRA compared with healthy controls. Inclusion of these biomarkers improved model discrimination.
The study used R (version 4.1.0) and Python (version 3.8.10) for statistical analysis and model development. A digital self-assessment tool was developed and made publicly available to support individualized risk estimation and screening guidance.
The authors reported no competing interests.