RESUMEN
AIM: To identify the risk, the host-related prognostic factors and their predictive value for anastomotic leakage after colorectal resections following cancer. METHOD: 993 patients who underwent large bowel resection and primary anastomosis above 12 centimeters from the anal verge, without a temporary or permanent stoma at the Surgical Hospital No.3 (Cluj-Napoca, Romania) were retrospectively reviewed. RESULTS: 32 (3.22 percent) anastomotic leaks were confirmed. Univariate analysis showed that the preoperative variables significantly associated with anastomotic leakage included: weight loss, smoking, cardiovascular disease, lung disease, hypoproteinemia, diabetes, anemia, leukocytosis, presence of two or more underlying diseases. Alcohol use, cerebrovascular disease, bowel preparation, type of anastomosis, tumor location, stage and histology were not significant variables. Hypoproteinemia (total serum protein level < or = 6 g/dl) and anemia (serum hemoglobin level < or = 11 g/dl) remained significant in the logistic regression model. The prognostic role of serum hemoglobin and proteins for the anastomotic leak was assessed using ROC curve analysis. For the cut-off value of serum protein level = 5.5 g/dl, a sensitivity of 61.6 percent and a specificity of 84.2 percent were calculated. The area under the curve was 0.703 (p= 0.0024). The area under the curve for serum hemoglobin was 0.616 (p=0.028). A sensitivity of 64.0 percent and a specificity of 64.7 per cent were obtained for a cut-off value of 9.4 g/dl. CONCLUSION. A serum protein level lower than 5.5 g/dl and serum hemoglobin lower than 9.4 g/dl could be considered as host-related predictive markers for anastomotic leak in large bowel resections for cancer.
