Failure of preoperative data to accurately predict which patients undergoing major hepatic surgery will develop postoperative coagulation disturbances: a single-hospital retrospective cohort study.

Purpose: Epidural analgesia has been considered a highly effective analgesic modality for hepatobiliary surgeries, particularly within 72 hr postoperatively. Coagulation derangements are possible complications following liver resection that can be concerning in the setting of epidural analgesia given the inherent risk of spinal epidural hematoma. We sought to develop predictive models for postoperative coagulation disturbance (defined as an international normalized ratio > 1.5, a partial thromboplastin time > 40 sec, or a platelet count < 100,000 × 10⁶·L^-1) in patients eligible for epidural analgesia.

Methods: We conducted a retrospective cohort study of patients undergoing liver resection at the University of Iowa (Iowa City, IA, USA) between 2011 and 2023. We reviewed records for patient characteristics, operative parameters, preoperative coagulation labs, and postoperative coagulation labs up to seven postoperative days. We used three types of predictive modeling.

Results: Among 684 patients, 37 had a length of stay ≤ 72 hr of surgery or preoperative coagulation disturbance. Among the remaining 647 patients, 512 (79%) received a thoracic epidural. The incidence of postoperative coagulation disturbances within 72 hr was 25% (95% confidence interval, 22 to 28), mostly thrombocytopenia (20% of all patients), and was noted on postoperative day 1 for 11% and postoperative day 2 for 22%. The volume of liver resected was greater among patients with postoperative coagulation disturbance (P < 0.001; area under the receiving operating characteristic curve, 0.61). There was no predictive value for coagulation disturbance based on patients' sex, American Society of Anesthesiologists' Physical Status classification, body mass index, weight, age, adjuvant chemotherapy, estimated operative duration, or year of data (all standardized differences < 0.24). Classification tree modeling had a single node (i.e., no useful preoperative prediction). Stepwise backward logistic regression using P < 0.05 for inclusion had just two patients (0.3%) with a predicted probability of postoperative coagulation disturbance < 10% and none < 5%.

Conclusions: Coagulation disturbances occur commonly in the context of hepatic surgery. Preoperative data commonly used to qualify a patient to receive epidural analgesia are insufficient to predict which patients are likely to develop postoperative coagulation disturbance. Regardless of the predictive modeling or criterion, the postoperative risk of coagulopathic disturbance will exceed 5% by 72 hr postoperatively. Enhanced recovery protocols recommending early epidural catheter removal need to consider the period of incidence of coagulation disturbance.