A predictive model identifies patients less likely to have adenomas after a colon cancer

Background. Patients with prior colorectal cancer (CRC) are at slightly increased risk of metachronous colorectal neoplasms, therefore endoscopic surveillance is indicated. Current recommendations of repeating examinations at 1, 3 and 5 years after surgery, are not tailored according to risk stratification. Our aim was to find predictive factors of colorectal neoplasms to build a predictive model, to spare colonoscopies for lowrisk patients. Methods. Multicenter retrospective study including patients with colon carcinoma surgically resected from 2001 to 2008 (training cohort) and from 2009 to 2013 (validation cohort). A predictive model for neoplasms occurrence at second surveillance colonoscopy was developed and externally validated. Results. 396 and 131 patients were included in training and validation cohort respectively. Patients with >=1 adenoma at the 2nd surveillance colonoscopy were 113/396 (28.5%) and 21/131 (16.5%) in the two groups. In the validation cohort, 3 cancers were found. Four variables were associated with higher risk of metachronous colorectal adenomas at 2nd surveillance colonoscopy on multivariate analysis: age >65 years old, left colectomy, >=1 advanced adenoma at basal colonoscopy and >=1 adenoma at first surveillance colonoscopy (table 1). The predictive model showed fair discrimination, with an area under the ROC curve of 0.69 and 0.64, in training and validation cohort respectively. In validation group, If patients with a low-risk profile (i.e. none of the risk factors) skip the 2nd surveillance colonoscopy, 25/131 (19.1%) exams would be saved while missing 2/ 21 (9.5%) patients with 1 adenoma; no cancer would be missed. Conclusions We provided a risk-stratification tool for adenoma occurrence after colon surgery, which could prove cost-effective to select patients who could skip the second surveillance colonoscopy.