RESUMO
OBJECTIVE: To observe the clinical value of prognostic nutritional index (PNI) combined with carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 242 in early prediction of anastomotic leakage after radical gastrectomy for gastric cancer. METHODS: We retrospectively collected clinical data of 350 gastric cancer patients who underwent radical gastrectomy in Gansu Provincial Hospital of Traditional Chinese Medicine between January 2018 and May 2022. According to the occurrence of anastomotic leakage, patients were divided into an occurrence group (n=34) and a non-occurrence group (n=316). The clinical value of PNI combined with CEA and CA242 on the 3rd day after surgery in predicting anastomotic leakage was explored. Lasso regression analysis was used to screen predictive indicators of anastomotic leakage and establish a risk model. RESULTS: In the 350 patients who underwent radical gastrectomy for gastric cancer, anastomotic leakage was observed in 34 cases, with an incidence rate of 9.7%. A higher proportion of patients in the occurrence group exhibited diabetes, hand-sewn anastomosis, advanced tumor node metastasis (TNM) staging, and intraoperative bleeding, when compared to those in the non-occurrence group (P<0.05). Moreover, on the 3rd postoperative day, patients in the occurrence group demonstrated a significantly lower PNI than those in the non-occurrence group, along with elevated levels of CEA and CA242 (P<0.05). The area under the curve (AUC) for PNI, CEA, and CA242 were 0.827, 0.601, and 0.504, respectively, while the AUC for the combination was 0.829. As per the LASSO regression analysis, history of diabetes and PNI were identified as key factors correlating with anastomotic leakage (P<0.05). Employing the risk score formula, we obtained individual risk scores for each sample. Notably, risk scores in the occurrence group significantly surpassed those in the non-occurrence group (P<0.0001). The AUC for the risk score in predicting patient lung infection was 0.854. The internal verification C-index emerged as 0.863 (0.806-0.920), indicating a good model fit. Furthermore, the DeLong test revealed a significantly greater AUC of the risk model, compared to the combination and PNI (P<0.05). CONCLUSION: CEA and CA242 are not promising predictive indicators for anastomotic leakage after surgery in patients with gastric cancer, but the prediction model we established can improve the predictive efficiency of anastomotic leakage in these patients.