Prediction of advanced colonic neoplasm in symptomatic patients: a scoring system to prioritize colonoscopy (COLONOFIT study)

BACKGROUND: Fast-track colonoscopy to detect patients with colorectal cancer based on high-risk symptoms is associated with low sensitivity and specificity. The aim was to derive a predictive score of advanced colonic neoplasia in symptomatic patients in fast-track programs. METHODS: All patients referred for fast-track colonoscopy were evaluated. Faecal immunological haemoglobin test (3 samples; positive> 4 μg Hb/g), and a survey to register clinical variables of interest were performed. Colorectal cancer and advanced adenoma were considered as advanced colonic neoplasia. A sample size of 600 and 500 individuals were calculated for each phase 1 and phase 2 of the study, respectively (Phase 1, derivation and Phase 2, validation cohort). A Bayesian logistic regression analysis was used to derive a predictive score. RESULTS: 1495 patients were included. Age (OR, 21), maximum faecal-Hb value (OR, 2.3), and number of positive samples (OR, 28) presented the highest ORs predictive of advanced colonic neoplasia. The additional significant predictive variables adjusted for age and faecal-Hb variables in Phase 1 were previous colonoscopy (last 5 years) and smoking (no, ex/active). With these variables a predictive score of advanced colonic neoplasia was derived. Applied to Phase 2, patients with a Score > 20 had an advanced colonic neoplasia probability of 66% (colorectal cancer, 32%), while those with a Score ≤ 10, a probability of 10% (colorectal cancer, 1%). Prioritizing patients with Score > 10, 49.4% of patients would be referred for fast-track colonoscopy, diagnosing 98.3% of colorectal cancers and 77% of advanced adenomas. CONCLUSIONS: A scoring system was derived and validated to prioritize fast-track colonoscopies according to risk, which was efficient, simple, and robust. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1186/s12885-019-5926-4) contains supplementary material, which is available to authorized users.