Background: The possibility of outlining a risk profile for perioperative blood transfusion of cirrhotic patients submitted to hepatic resection can help to rationalize transfusion policy.
Methods: Data from 323 hepatic resections, performed in cirrhotic patients, were reviewed. Bootstrap and a leave-one-out logistic regressions were applied to test the accuracy of available risk scores for peri-operative transfusion identified from PubMed search of the last 20 years, to refine them, and to provide internal validation for present results.
Results: One-hundred-six patients (32.8%) required blood transfusions during either intra- and/or postoperative. The predictive accuracy of three identified risk scores was poor with the area under receiver operating characteristics (AUROC) curves <0.70 in all cases. Tumor diameter, hemoglobin and presence of coronary artery disease were confirmed, in the present cohort, as predictors of blood transfusion together with serum albumin and bilirubin. The leave-one-out logistic regression results in an AUROC of 0.80, and of 0.79 for internal validation, significantly higher than that of the three scores tested (P<0.001). A Maximal Surgical Blood Order Schedule stratification was proposed.
Conclusion: The risk profile for transfusion of cirrhotic patients undergoing hepatectomy can be better assessed with a model that combines already known clinical factors and hepatic function indexes.