To investigate the relationship between serum tumor markers and colorectal cancer liver metastasis (CRLM), thus to establish a prediction model for CRLM and evaluate its application value.

A total of 371 patients with colorectal cancer admitted to Cangzhou Central Hospital from April 2021 to October 2022 were included in this study. They were divided into two groups according to whether there were liver metastases. There were 59 patients in the CRLM group and 312 patients in the non-CRLM group. Clinical data, test results of pre-treatment blood and serum tumor markers of patients were collected retrospectively. Binary Logistic regression was used to analyze the independent influencing factors of CRLM, and the diagnostic efficacy of the prediction model was evaluated using receiver operating characteristics (ROC) curve and goodness-of-fit test.

Compared with non-CRLM group, neutrophil to lymphocyte ratio (NLR), carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9), and sialic acid/hydroxyproline (SA&Hyp) levels were significantly higher (all P<0.01). Binary Logistic regression analysis showed that the high levels of serum CEA (OR=1.021, 95% CI: 1.012-1.030), CA19-9 (OR=1.003, 95% CI: 1.001-1.005) and SA&Hyp (OR=1.055, 95% CI: 1.033-1.077) were independent risk factors for CRLM after correction (all P<0.01). ROC curve analysis revealed that the predictive modeling (AUC=0.867, 95% CI: 0.818-0.916, sensitivity was 0.831, specificity was 0.785) jointly constructed by CEA, CA19-9 and SA&Hyp was more effective in predicting CRLM compared with any single indicator (all P<0.05). The goodness-of-fit test results indicated that the prediction model had good matching (χ²=8.441, P=0.392).

High levels of serum CEA, CA19-9 and SA&Hyp are independent risk factors for CRLM. The prediction model of CRLM based on the joint detection of risk factors shows a reasonable predictive value and may provide new insights for the early diagnosis of CRLM in clinical work.