Prognostic factors in colorectal liver metastases patients with various tumor numbers treated by liver resection: a single-center, retrospective study

Background Multiple liver metastases is considered a risk factor for overall survival of colorectal liver metastases patients (CRLM) after curative resection. However, whether the prognostic factors were constant in patients with various liver metastases (LM) numbers has not been adequately investigated. This retrospective study aimed to evaluate the changing of prognostic factors on overall survival (OS) in CRLM patients with various LM after curative resection. Methods Patients who underwent liver resection for CRLM between January 2000 and November 2020 were retrospectively studied. They were divided into three subgroups according to LM numbers by X-tile analysis. Multivariable analysis identified prognostic factors in each subgroup. Nomograms were built using different prognostic factors in three subgroups, respectively. Performance of the nomograms was assessed according to the concordance index (C-index) and calibration plots. The abilities of different scoring systems predicting OS were compared by calculating the area under the time-dependent receiver operating characteristic (ROC) curve (AUC). Results A total of 1095 patients were included. Multivariable analysis showed tumor number increasing was an independent risk factor. Patients were subsequently divided into 3 subgroups according to the number of LM by X-tile analysis, namely solitary (n = 375), 2-4 (n = 424), and >= 5 (n = 296). The 3-year and 5-year OS rates were 64.1% and 54.0% in solitary LM group, 58.1% and 41.7% in 2-4 LM group, and 50.9% and 32.0% in >= 5 LM group, respectively (p < 0.001). In multivariable analysis, RAS mutation was the only constant independent risk factor in all subgroups. The nomograms were built to predict survival based on independent factors in three subgroups. The C-index for OS prediction was 0.707 (95% CI 0.686-0.728) in the solitary LM group, 0.695 (95% CI 0.675-0.715) in the 2-4 LM group, and 0.687 (95% CI 0.664-0.710) in the >= 5 LM group. The time-dependent AUC values of nomograms developed using different risk factors after stratifying patients by tumor number were higher than the traditional scoring systems without patient stratification. Conclusions The prognostic factors varied among CRLM patients with different LM numbers. RAS mutation was the only constant risk factor. Building prediction models based on different prognostic factors improve patient stratification.