Preoperative inflammatory markers and tumor markers in predicting lymphatic metastasis and postoperative complications in colorectal cancer: a retrospective study.

Abstract

Objective: To analyze the impact of preoperative inflammatory markers and tumor markers on lymphatic metastasis and postoperative complications in colorectal cancer patients, and explore their predictive value for these outcomes. Furthermore, based on the preoperative inflammatory and tumor marker indicators with significant effects, predictive models for the risk of lymphatic metastasis and the incidence of postoperative complications will be constructed.

Methods: This study retrospectively analyzed the clinical data of CRC patients who underwent surgical treatment at Shanxi Bethune Hospital between January 2021 and June 2024. Preoperative inflammatory markers and tumor markers were compared between the lymph node-positive and lymph node-negative groups. Variables were selected using Lasso regression, and independent factors influencing lymph node metastasis were identified through multivariate logistic regression analysis. Based on these results, a Nomogram prediction model was constructed, and its accuracy was evaluated using a calibration curve. The discriminatory ability of the model was assessed with the ROC curve, and its clinical applicability was analyzed using the DCA curve. Similarly, for predicting postoperative complications, Pearson correlation analysis was used to examine the relationships between preoperative inflammatory markers, tumor markers, and complications. ROC curves were employed to calculate the AUC and optimal cutoff values for each marker. Kaplan-Meier (KM) curves were used to analyze the impact of these markers on DFS. Independent factors were identified through univariate and multivariate logistic regression analyses, and a Nomogram model was constructed and validated.

Results: A total of 196 patients were included in the study. The NLR, PLR, FAR, CEA, CA199, and CA724 levels were significantly elevated in the lymph node metastasis group (P < 0.05). Lasso regression identified smoking history, NLR, FAR, and CA724 as non-zero coefficient variables. Multivariate logistic regression further confirmed smoking history (HR = 4.20), NLR (HR = 2.52), FAR (HR = 1.18), and CA724 (HR = 1.32) as independent predictors of lymph node metastasis (P < 0.05). The Nomogram prediction model constructed based on these results showed high prediction accuracy, with a ROC curve AUC of 0.880, indicating excellent discriminatory ability. The DCA decision curve demonstrated good clinical applicability. In postoperative complication prediction, Pearson correlation analysis revealed a positive correlation between NLR, PLR, FAR, CA199, and CA724 with complication rates (P < 0.05), with correlation coefficients of 0.24, 0.34, 0.16, 0.19, and 0.19, respectively, with PLR showing the strongest correlation. ROC curve analysis showed that the AUCs for NLR, PLR, LMR, FAR, and CAR were 0.633, 0.675, 0.467, 0.580, and 0.559, with optimal cutoff values of 4.29, 261.71, 3.39, 18.20, and 11.26, respectively. The AUCs for CEA, CA199, and CA724 were 0.567, 0.612, and 0.609, with optimal cutoff values of 11.87, 10.27, and 6.85. KM curve analysis showed that higher levels of NLR, FAR, CAR, CEA, CA199, and CA724 were associated with poorer DFS. Univariate and multivariate logistic regression further confirmed NLR (HR = 1.53) and CA724 (HR = 1.11) as independent predictors of complications (P < 0.05). The calibration curve indicated high prediction accuracy, with a ROC curve AUC of 0.729, demonstrating excellent discriminatory ability, and the DCA decision curve showed good clinical applicability.

Conclusion: Preoperative inflammatory markers and tumor markers have a significant impact on the occurrence of lymphatic metastasis and postoperative complications in colorectal cancer patients, demonstrating certain clinical value in predicting lymphatic metastasis and postoperative complications. The predictive models developed in this study provide a reference for personalized diagnosis and treatment, but their practical application needs to be further validated through large-scale clinical studies.