Construction and Evaluation of Metastatic Risk Model in Left and Right Colorectal Cancer

Abstract

Objective To identify the influence factors and construct a predicted model for liver, lung, bone, or brain metastasis among patients with left or right colorectal cancer. Methods Patients with colorectal cancer with information on liver, lung, bone, and brain metastasis were retrospectively filtered and analyzed from 2010 to 2018 from the SEER database. These patients were divided into three groups based on their primary tumor location. Multivariate logistic regression analysis was conducted to identify the influence factors on metastasis. A nomogram that could predict metastasis was established and further validated by the AUC of ROC curves. Results A total of 49335 eligible patients were chosen from the SEER database. N stage and CEA were identified as risk factors for all metastases, which were unrelated to primary tumor location. By contrast, race had varying effects on liver metastasis between different groups (P < 0.05). The nomogram model predicting liver metastasis was successfully established, and the AUCs based on the three groups were 0.821 (95%CI: 0.813-0.830), 0.841 (95%CI: 0.833-0.848), and 0.796 (95%CI: 0.782-0.811), respectively. Conclusion The influence factors and predictive models on liver metastasis were different in patients with colorectal cancer and different primary tumor locations.

Keywords

• colorectal cancer
• primary tumor location
• influence factor
• metastasis
• predicted model