Combining the Platelet-to-Albumin Ratio with Serum and Pathologic Variables to Establish a Risk Assessment Model for Lymph Node Metastasis of Gastric Cancer and Evaluation of Its Predictive Value
Abstract Background The preoperative platelet count and serum tumor markers have been shown to correlate with the lymph node metastasis (LNM) of gastric cancer (GC).The aim of this study was to establish a risk assessment model that incorporated the platelet-to-albumin ratio (PAR) for LNM of GC and to evaluate its clinical significance. Methods The clinical data of 314 patients with GC diagnosed by postoperative pathology were collected in our hospital. According to whether there was LNM in the pathological specimens of the operation, the patients were divided into the group without LNM and the group with LNM. Univariate analysis and multivariate logistic regression were used to analyze the relevant factors affecting LNM of GC and to identify independent risk factors for LNM of GC. The random forest algorithm was used to extract the important risk factors of LNM in GC. A nomogram model of the risk assessment of LNM of GC was constructed by the “rms” package of R software. The receiver operating characteristic (ROC) curve was used to evaluate the accuracy, sensitivity and specificity of the model for predicting LNM of GC. Results Univariate analysis showed that the factors associated with LNM of GC were sex (P=0.015), smoking (P=0.027), lesion size (P=0.000), pathological type (P=0.001), differentiation degree (P=0.000), infiltration depth (P=0.000), PAR (P=0.005), carbohydrate antigen (CA) 19-9 (P=0.017), CA125 (P=0.000) and CA72-4 (P=0.005). Multivariate logistic regression showed that lesion size [odds ratio (OR): 1.322; P = 0.000], differentiation degree (OR: 0.582; P = 0.001), and depth of invasion (OR: 1.734; P = 0.000) were independent risk factors for LNM in GC. The risk assessment model of LNM in GC was established according to the ranking of variables shown by the random forest algorithm. The C statistic of the model evaluation was 0.827, the sensitivity was77.2%, and the specificity was 74.8%. Conclusion Lesion diameter larger than 2.65 cm, poor differentiation and deep infiltration were high-risk factors for LNM in GC. The nomogram model constructed by PAR, lesion size, infiltration depth, CA125, CA19-9, CA72-4, and differentiation degree, can well predict the risk of LNM in GC.