Most eligible candidates for primary tumor resection among metastatic colorectal cancer patients: a SEER-based population analysis.

Abstract

Background: Primary tumor resection (PTR) can improve the prognosis and survival of some patients with metastatic colorectal cancer (mCRC). However, selecting candidates that may benefit from this intervention may be challenging. Therefore, we aim to construct a predictive model to help identify the most eligible candidates for PTR.

Methods: Propensity score matching (PSM) was used to balance the baseline characteristics of the patients. Patients in the surgical group were further allocated to either a beneficial or a non-beneficial cohort based on whether their survival time exceeded the median overall survival (mOS) time of the non-surgical group. A multivariate Cox analysis was then conducted to select independent prognostic risk factors the surgical group. Finally, multivariate logistic regression was used to establish a predictive model based on the demographic characteristics, and the calibration curves, area under the curve (AUC) of the receiver operating characteristic (ROC) curve, and a decision curve analysis (DCA) were used to validate and assess the model accuracy and clinical prediction ability.

Results: A total of 11,763 mCRC patients were enrolled in the study, of whom 8,808 (74.88%) underwent PTR. After PSM, the median cancer-specific survival (CSS) was 29 months in the surgical group and 16 months in the non-surgical group (P<0.001). Based on the logistic regression, 10 covariates [age, ethnicity, negative or positive CEA, TNM staging, grade, bone metastasis, liver metastasis, histology, primary tumor site, distant metastasis surgery (or no surgery), and chemotherapy] were identified and used to construct the predictive model, using a training and a validation group. The AUC values of the nomograph were 0.727 in the training group and 0.742 in the validation group. The calibration curves, DCA and Kaplan-Meier (K-M) analysis results suggest that the predictive model was able to accurately predict the likelihood of a patient benefiting from PTR (P<0.001).

Conclusions: This study constructed and validated a predictive model to help clinicians identify patients with mCRC who are most likely to benefit from PTR.