Authors
Chang Liu, Yuxing Shuai, Huizhen Ye, Chong Shan Ng, Fubin Hu, Longxiu Fan, Shaoyu Cheng, Beiping Zhang, Haiyan Zhang, Tianwen Liu, Xin-Lin Chen, Cailing Zhong
Published in
International journal of general medicine. Volume 19. Pages 640776. Epub Nov 05, 2026.
Abstract
To develop and validate an integrative prediction model for colorectal neoplasm (CN) screening to enhance detection effectiveness and provide a noninvasive diagnostic method.
This cross-sectional study included 2,047 participants undergoing colonoscopy at Guangdong Provincial Hospital of Chinese Medicine between March 2022 and July 2023. Predictive models were constructed using epidemiological and Traditional Chinese Medicine factors via five machine learning algorithms. Performance was assessed using the area under the curve (AUC), sensitivity, specificity, positive predictive value, negative predictive value, DeLong test, calibration curves, and decision curve analysis.
Among 1,814 included participants (39.25% with CN), significant predictors included male, age, hypertension, drinking, tongue features, and pigmented nevus of the facial and perioral triangle (all P <0.05). Ten features were selected through univariate and multivariate logistic regression. The Integrative Medicine Colorectal Screening (IMCS) model using extreme gradient boosting (XGBoost) achieved the highest training-set AUC (0.817, 95% CI: 0.794-0.841), while the logistic regression (LR) model performed best on the validation set (AUC = 0.690, 95% CI: 0.644-0.736). The IMCS (LR) model outperformed existing models (APCS, M-APCS, CNP; Delong test, all P <0.05), demonstrated strong calibration (mean square error = 0.212, P = 0.425), and showed potential clinical utility (net benefit: 0.144-0.265).
The IMCS model improves predictive capability for CN risk and may assist in CN risk stratification. Further large-scale prospective studies are warranted to validate its clinical applicability.
PMID:
43005915
Bibliographic data and abstract were imported from PubMed on 10 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 8
- Comments 0