Abstract
Objective: To establish a nomogram prediction model (based on clinicopathological and radiological features) for the development of metachronous liver metastasis (MLM) in patients with colorectal cancer (CRC). Methods: This retrospective study included patients with CRC who underwent surgery at Changshu No.1 People's Hospital and the Second Affiliated Hospital of Soochow University between January 2016 and December 2018. The clinical, pathological, and radiological features of each patient were investigated. Risk factors for MLM were identified by univariable and multivariable analyses. The predictive nomogram for MLM development was constructed. The predictive performance of the nomogram was estimated by the receiver operating characteristics curve, calibration curve, and decision curve analysis. Results: This study included 161 patients with CRC [median age: 66 (range, 33-87) years]. Fifty-nine developed MLM after a median of 12 (range, 2-52) months after surgery. The multivariable logistic regression analysis showed that age >66 years (OR=3.471, 95% CI: 1.272-9.473, P=0.015), N2 stage (OR=6.534, 95% CI: 1.456-29.317, P=0.014), positive vascular invasion (OR=2.995, 95% CI: 1.132-7.926, P=0.027), positive tumor deposit (OR=4.451, 95% CI: 1.153-17.179, P=0.030), and linear (OR=6.774, 95% CI: 1.306-35.135, P=0.023) and nodal pericolic fat infiltration patterns (OR=8.762, 95% CI: 1.521-50.457, P=0.015) were independently associated with MLM. These five factors were used to create a nomogram. The area under the receiver operating characteristics curve of the nomogram was 0.866 (95% CI: 0.803-0.914), indicating favorable prediction performance. The calibration curve of the nomogram showed a satisfactory agreement between the predicted and actual probabilities. Conclusions: A nomogram prediction model based on five clinicopathological and radiological features might have favorable prediction performance for MLM in patients who underwent surgery for CRC. Hence, the present study proposes a nomogram that can easily be used to predict MLM after CRC surgery based on readily available features.