Back to 2025 Abstracts
PREDICTIVE VARIABLES FOR LIVER FAILURE AFTER MAJOR HEPATECTOMY: A NSQIP ANALYSIS WITH INTERNATIONAL EXTERNAL VALIDATION
Francisco Tustumi
*2, Fabricio Ferreira Coelho
2, Vagner Birk Jeismann
2, Gilton Marques Fonseca
2, Jaime Pirola Kruger
2, Paulo Herman
2, Mihir Shah
1, David Kooby
1, Juan Sarmiento
1, Felipe B. Maegawa
11Surgery, Emory University School of Medicine, Atlanta, GA; 2Universidade de Sao Paulo, São Paulo, SP, Brazil
Background: Post-hepatectomy liver failure (PHLF) remains the leading cause of death after major liver resections. Despite current advances in perioperative care, PHLF after major hepatectomy is still prevalent, suggesting that patient selection tools can be improved. Herein, we examined the predictive factors for PHLF following major hepatectomy using multicenter data.
Methods: The National Surgical Quality Improvement Program (NSQIP) database (2017 to 2023) was utilized to identify patients who underwent major hepatectomies. These patients were divided into training and internal validation cohorts in a ratio of 7:3. The external validation cohort was derived from the institutional database from Hospital das Clínicas, São Paulo, Brazil. A forward logistic regression selection method identified the independent predictors of PHLF. Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) were built to assess the model's predictive ability.
Results: In the training cohort (n=6246), the independent predictors of PHLF included age (OR: 1.03, 95%CI 1.02–1.04), race (other vs white OR: 1.50, 95%CI 1.12-2.00), albumin-bilirubin (ALBI) grade (OR: 1.62, 95%CI 1.26-2.10), Model for End-Stage Liver Disease (MELD) (OR: 1.08, 95%CI 1.02-1.14), receipt of neoadjuvant therapy (OR: 1.49, 95%CI 1.12-2.00), bleeding requiring transfusion (OR: 2.83, 95%CI 2.12-3.77) and surgery extent (right vs left hepatectomy OR: 5.20, 95%CI 3.03-8.91; trisegmentectomy vs left hepatectomy OR: 5.79, 95%CI 3.33-10.1). The predictive model was constructed with these variables. The ROC curve of the model produced an AUC of 0.77. The Youden index determined the best cut-off point for ALBI to be -2.49. The internal validation (n=2725) confirmed the predictive ability of the model with an AUC of 0.76. Multivariable logistic regression revealed that ALBI > -2.49 was independently associated with PHLF (OR: 1.37, 95%CI 1.04-1.80). In the external validation (n=465), the model had a similar performance with an AUC of 0.77, and ALBI > - 2.49 was independently associated with PHLF, OR: 3.16, 95%CI 1.41-7.07.
Conclusions: By incorporating evidence from North American and South American populations, these findings provide robust and geographically diverse support for predictors of PHLF in major hepatectomy. The development of prediction models that can improve the risk stratification for PHLF, and enhance patient management and outcomes globally, is warranted.
Figure 1. Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) values for predicting post-hepatectomy liver failure are shown for the training, internal validation, and external validation groups.
Table 1. Multivariable logistic regression analysis for predictors if post-hepatectomy liver failure. Results are reported as odds ratios (OR) with their corresponding 95% confidence intervals (95% CI).
Back to 2025 Abstracts
