Background: Several factors can affect the risk of recurrence after curative resection of colorectal cancer (CRC). We aimed to develop a risk model for recurrence after definitive treatment of Stage I-III CRC using data from a nationally representative database and to develop an individualized web-based risk calculator.
Methods: A random sample of patients who underwent resection for Stage I-III CRC between 2006 and 2007 at Commission on Cancer (CoC) accredited centers were included. Primary data regarding first recurrence was abstracted from medical records and merged with the National Cancer Database. Multivariable cox regression analysis was used to test for factors associated with cancer recurrence, stratified by stage. Model performance was tested by c statistic and calibration plots. Hazard Ratios were utilized to develop an individualized web-based recurrence prediction tool.
Results: A total of 8249 patients from 1175 CoC centers were included. Of these, 1656 (20.1%) patients had a recurrence during 5 years of follow-up. Median time to recurrence was 16 months. The final predictive models displayed excellent discrimination and calibration with concordance indexes of 0.7. The online calculator included 12 variables, including tumor site, stage, time since surgery, and surveillance intensity. Output is displayed numerically and graphically with an icon array.
Conclusions: Using primarily abstracted recurrence data from a random sample of patients treated for CRC at CoC accredited centers across the United States, we successfully created an individualized CRC recurrence risk assessment tool. This web-based calculator can be used by physicians and patients in shared decision making to guide management discussions.
Trial registration: ClinicalTrials.gov Registration Number: NCT02217865.