Optimizing symptomatic triage to colonoscopy: evaluation of FIT-centered multivariable selection strategies.

<title>Abstract</title> <p>Background and Aims: Fecal Immunochemical Testing (FIT) can be used at fixed thresholds to select symptomatic patients for colonoscopy, although this selection misses 10% of colorectal cancer (CRC) cases. We evaluated multivariable risk models incorporating demographic variables, FIT and blood-based biomarkers, to triage symptomatic colonoscopy referral in the Urgent Cancer Pathway. This was done with CRC as outcome and CRC combined with advanced adenomas (AA) and inflammatory bowel disease (IBD). Methods Clinical data and samples for fecal and blood-based biomarkers were collected prior to colonoscopy. Three selection strategies were evaluated: FIT only model, FIT demographic model (including age and sex), and an extended model incorporating also blood-based biomarkers based on stepwise selection. Models were internally validated using 5-fold cross-validation. Performance was assessed using AUC and by estimating specificity and colonoscopy reduction at fixed sensitivities. Results Among 1,586 patients, colonoscopy identified 84 CRC (5.3%), 111 AA (7.0%), and 29 IBD (1.8%) cases. The FIT demographic model achieved an AUC of 0.91 (95% CI, 0.88–0.94), compared with 0.88 (95% CI, 0.84–0.92) for the FIT only model (p &lt; 0.001). The FIT extended model (final stepwise model including FIT, age, sex, prior positive FIT screening, and selected blood-based biomarkers) achieved an AUC of 0.92 (95% CI, 0.89–0.95), representing only a modest additional improvement. At 95% sensitivity, the FIT extended model achieved 57% specificity, corresponding to a 54% reduction in colonoscopies. At a sensitivity threshold of 100%, colonoscopies could be reduced by 23%. For CRC/AA/IBD, the FIT extended model also outperformed the FIT only model. Conclusion Multivariable FIT-centered selection strategies improve risk stratification compared with FIT alone and may optimize symptomatic triage to colonoscopy. Blood-based biomarkers added limited benefit.</p>