Background: In liver transplantation most studies were designed to predict massive transfusion rather than whether or not transfusion is required. We hypothesized that (presurgery) data from thromboelastometry may predict perioperative blood requirements.
Methods: A post-hoc analysis of data from a controlled trial was performed with the primary end point of predicting zero red blood cells. Of the 92 patients studied, 6 were excluded because of incomplete EXTEM and/or FIBTEM data. The multivariate models included preoperative variables with a P value <0.10 in the univariate model: age, MELD score, hemoglobin, plasma fibrinogen, platelet count, activated partial thromboplastin time, INR, EXTEM maximum clot amplitude after 10 minutes, EXTEM an FIBTEM maximum clot firmness, plasma creatinine, and donor data.
Results: Blood was transfused to 58% of patients during the surgical procedure and to 34% in the first 24 hours postoperatively. The final model was selected using a backward approach, and fractional polynomials were explored to assess model improvement for the prediction. Hemoglobin was a strong predictor: each 1 g/dL of hemoglobin increase reduced the risk of blood transfusion by 52%. An EXTEM maximum amplitude at 10 min was also a predictor of Red Blood requirement, showing a 64% risk reduction for values between the first quartile (35 mm) and the second quartile (41 mm) but no further improvement for the third and fourth quartiles and resulting in a prediction (ROC AUC of 0.815 [0.771-0.859]).
Conclusions: Presurgical EXTEM maximum amplitude at 10 min <35 mm is highly predictive of red blood administration during liver transplantation.