Abstract
This retrospective cohort study aimed to identify baseline patient characteristics involving modifiable lifestyle factors that are associated with the development of colorectal adenomas, and establish and validate a nomogram for risk predictions among high-risk populations with negative index colonoscopy. A total of 83,076 participants who underwent an index colonoscopy at the Tianjin Union Medical Center between 2004 and 2019 were collected. According to meticulous inclusion and exclusion criteria, 249 subjects were enrolled and categorized into the primary and validation cohorts. Based on the primary cohort, we utilized the LASSO-Cox regression and the univariate/multivariate Cox proportional hazards (Cox-PH) regression parallelly to select variables, and incorporated selected variables into two nomogram models established using the multivariate Cox-PH regression. Comparison of the Akaike information criterion and the area under the receiver operating characteristic curve of the two models demonstrated that the nomogram model constituted by four covariates retained by the LASSO-Cox regression, including baseline age, body mass index, physical activity and family history of colorectal cancer (CRC) in first-degree relatives, performed better at predicting adenoma-free survival probabilities. Further validation including the concordance index, calibration plots, decision curve analysis and Kaplan-Meier survival curves also revealed good predictive accuracy, discriminating ability, clinical utility and risk stratification capacity of the nomogram model. Our nomogram will assist high-risk individuals with negative index colonoscopy to prevent colorectal adenoma occurrence and CRC morbidity with improved cost-effectiveness.
Introduction
Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide, despite being highly preventable through the detection and removal of precancerous adenomas. Colonoscopy is the gold standard for screening; however, clinical resources are often strained. A significant challenge lies in managing patients who receive a negative "clean" index colonoscopy. Current guidelines often categorize these individuals broadly, leading to "one-size-fits-all" surveillance intervals that may result in either the over-utilization of resources or the failure to catch rapidly developing lesions in high-risk individuals. This study addresses the need for personalized risk assessment by incorporating modifiable lifestyle factors into a predictive nomogram.
Methods and Datasets
The study utilized a retrospective cohort design based on electronic medical records from the Tianjin Union Medical Center, China, covering the period from 2004 to 2019. The initial pool consisted of 83,076 participants. Through a rigorous selection process, the researchers identified 249 high-risk subjects (defined by positive High-Risk Factor Questionnaires or Fecal Immunochemical Tests) who had a negative index colonoscopy followed by at least one surveillance colonoscopy.
The cohort was split chronologically into a primary cohort (n=174) for model development and a validation cohort (n=75). Two statistical approaches for variable selection were compared: a traditional univariate-to-multivariate Cox proportional hazards (Cox-PH) pipeline and a LASSO-Cox regression method. The LASSO-Cox approach, which effectively handles overfitting by shrinking less significant variables, proved superior in identifying the most robust predictors for the final nomogram.
Key Results
The final predictive model, Nomogram-1, identified four key variables: baseline age, Body Mass Index (BMI), physical activity, and family history of CRC in first-degree relatives. Notably, the inclusion of modifiable factors like BMI and physical activity significantly enhanced the model's predictive performance.
The model demonstrated reliable discriminating power with a C-index of 0.682 in the primary cohort and 0.663 in the validation cohort. Kaplan–Meier survival curves indicated a statistically significant difference (p < 0.0001) in adenoma-free survival between high-risk and low-risk groups as defined by the nomogram. Furthermore, Decision Curve Analysis (DCA) confirmed that the nomogram provides a higher net clinical benefit compared to universal screening or no screening strategies.
Discussion and Limitations
This study highlights that lifestyle choices directly influence the rate at which new adenomas appear following a negative scan. By providing a personalized score, clinicians can move toward tailored surveillance intervals—for example, recommending a check-up at two years rather than five based on a patient's specific risk profile.
However, several limitations exist. The final sample size of 249 is relatively small due to strict exclusion criteria. As a retrospective study, lifestyle data relied on patient recall, which may introduce bias. Additionally, the data is from a single center in China, necessitating future multi-center prospective trials to validate the model across diverse populations. Future iterations could also benefit from including dietary data and medication history (e.g., aspirin use) to further refine accuracy.
Summary updated