External, Multicentre Validation of a Machine-Learning Model to Predict Colonic Adenoma in Indian Adults

External, Multicentre Validation of a Machine-Learning Model to Predict Colonic Adenoma in Indian Adults-A Prospective, Observational, Multicentre Study

Colorectal adenomas are precursors to colorectal cancer (CRC). Accurate pre-procedure risk stratification could optimize colonoscopy yield and resource allocation in India, where adenoma prevalence varies by age, sex, and lifestyle/metabolic factors. ML models can integrate multiple predictors to estimate individualized risk.

Existing risk scores are largely Western; performance and calibration may not be appropriate in Indian populations with different socio-demographic and metabolic profiles. External, prospective, multicentre validation is essential before clinical implementation.

Inclusion Criteria:

• Adults ≥18 years undergoing diagnostic colonoscopy.
• Adequate bowel preparation (Boston Bowel Preparation Scale total ≥6 with each segment ≥2).
• Complete examination (cecal intubation; withdrawal time ≥6 min when no therapy).
• Availability of all model predictors per CRF.

Exclusion Criteria:

• • Known CRC or polyp, prior colectomy, polyposis syndromes, known IBD, or strong hereditary CRC syndromes (e.g., Lynch) if excluded in derivation.

  • Inadequate prep, incomplete colonoscopy, obstructing lesions preventing optical diagnosis beyond obstruction.
  • Emergency colonoscopies, therapeutic-only procedures without diagnostic intent.