Benefit and prognosis prediction of radiation therapy in colorectal cancer liver metastases: a retrospective, population-based, propensity score-matched study.

[BACKGROUND] The benefits and optimal sequencing of radiation therapy (RT) for colorectal cancer liver metastases (CRLM) remain unclear. Therefore, using a large population-based dataset, this retrospective study aimed to evaluate the impact of RT on overall survival (OS) and determine its optimal timing. Furthermore, a nomogram was developed to predict OS in CRLM patients receiving RT.

[METHODS] This population-based retrospective cohort study used data from the Surveillance, Epidemiology and End Results database, including 5141 patients diagnosed with CRLM between 2010 and 2019. OS rates were conducted using Kaplan-Meier curves and log-rank tests. Propensity score matching (PSM) was employed to reduce baseline confounding between the RT and non-RT groups. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic factors for OS in CRLM patients receiving RT. A nomogram for predicting OS was constructed using R software.

[RESULTS] After PSM, 536 patients were included in both the RT and non-RT groups. The mean OS time was 26 months in the RT group and 22 months in the non-RT group, with a statistically significant difference confirmed by the log-rank test (p<0.001). Cox regression analysis identified RT as an independent protective factor for OS (HR=0.55, 95% CI 0.41 to 0.73). Subgroup analysis based on RT timing revealed that the combined preoperative and postoperative RT group had the highest OS, followed by the preoperative RT group, and then the postoperative RT group. Additionally, chemotherapy, brain, bone and other site metastases, and distant lymph node metastases were identified as independent factors affecting OS in CRLM patients receiving RT. The nomogram exhibited moderate discriminative performance in predicting OS, with Area Under the ROC Curve (Receiver Operating Characteristic Curve) (AUC) values of 0.634 and 0.637 in the training and validation sets, respectively.

[CONCLUSIONS] RT, particularly when administered in a combined preoperative and postoperative approach, is associated with long-term survival benefits in CRLM patients. The constructed nomogram, while demonstrating moderate predictive accuracy, serves as an initial tool for individualised OS estimation. These hypothesis-generating findings highlight the potential value of RT and warrant further validation through larger, prospective studies to develop more robust predictive models for personalised treatment strategies.