An APRI+ALBI-Based Multivariable Model as a Preoperative Predictor for Posthepatectomy Liver Failure

Abstract

Objective and background: Clinically significant posthepatectomy liver failure (PHLF B+C) remains the main cause of mortality after major hepatic resection. This study aimed to establish an aspartate aminotransferase to platelet ratio combined with an albumin-bilirubin grade (APRI+ALBI), based multivariable model (MVM) to predict PHLF and compare its performance to indocyanine green clearance (ICG-R15 or ICG-PDR) and albumin-ICG evaluation (ALICE).

Methods: A total of 12,056 patients from the National Surgical Quality Improvement Program database were used to generate a MVM to predict PHLF B+C. The model was determined using stepwise backwards elimination. The performance of the model was tested using receiver operating characteristic curve analysis and validated in an international cohort of 2525 patients. In 620 patients, the APRI+ALBI MVM, trained in the National Surgical Quality Improvement Program cohort, was compared with the MVM's based on other liver function tests (ICG clearance, ALICE) by comparing the areas under the curve (AUC).

Results: A MVM including APRI+ALBI, age, sex, tumor type, and extent of resection was found to predict PHLF B+C with an AUC of 0.77, with comparable performance in the validation cohort (AUC: 0.74). In direct comparison with other MVM's based on more expensive and time-consuming liver function tests (ICG clearance, ALICE), the APRI+ALBI MVM demonstrated equal predictive potential for PHLF B+C. A smartphone application for the calculation of the APRI+ALBI MVM was designed.

Conclusion: Risk assessment through the APRI+ALBI MVM for PHLF B+C increases preoperative predictive accuracy and represents a universally available and cost-effective risk assessment before hepatectomy, facilitated by a freely available smartphone app.

Keywords: 90-day mortality; ALICE grade; APRI+ALBI; ICG clearance; liver surgery.

FIGURE 1

A multivariable model based on the APRI+ALBI score, APRI combined with ALBI, for the prediction of PHLF grade B and C (B+C). The model is tested in the entire cohort (A) and tested in only patients undergoing major resection (B), documenting the nonlinear increase of PHLF B+C with a rising APRI+ALBI score. The model was calculated using the NSQIP cohort (predicted). The predictive performance of the model was validated using the validation cohort (international multicenter cohort, observed). APRI+ALBI score is given in deciles on the x-axis. PHLF B+C risk is given in % on the y-axis (A and B). To visualize the probability for PHLF B+C development for the complete cohort, a major resection subgroup, a CRCLM, primary liver cancer, and other malignancies with metastases in the liver subgroup and, respectively, subgroups for patients undergoing major liver resection in the different tumor subgroups, a bubble plot was generated (C). APRI+ALBI score is given in deciles and bubble size at each decile and for each patient groups indicates number of patients. The PHLF B+C probability is reflected in the bubble color and is calculated by the APRI+ALBI-based multivariable model. The associated PHLF B+C probability for each bubble is explained in the figure legend (C).

FIGURE 2

Evaluation of the performance of an APRI+ALBI, APRI combined with ALBI, based multivariable model (APRI+ALBI, age, sex, tumor type, extent of resection) for posthepatectomy liver failure grade B and C (PHLF B+C) prediction. The performance of the model is evaluated using ROC curve analysis and AUC (A). Validation of the model is done through ROC curve analysis and AUC calculation as well (B).

FIGURE 3

Descriptive comparison of the performance of several univariate models for prediction of PHLF B+C (A), PHLF C (B) and 90-day mortality (C). The different models are calculated using APRI+ALBI, APRI combined with ALBI, ICG clearance retention 15 min after administration (ICG-R15) and plasma disappearance rate (ICG-PDR), ALICE grade, and FIB-4 index. Models are calculated for 620 patients from an international multicenter cohort. The performance of each model is indicated as the AUC from ROC curve analysis. 95% CI and tests between the AUC of APRI+ALBI versus the AUC of each of the other parameters have been performed by a bootstrap resampling analysis (lower panels).

FIGURE 4

Comparison of different multivariable models for the prediction of PHLF B+C, PHLF grade C (PHLF C), and 90-day mortality in 620 patients out of an international multicenter cohort. Models included in the comparison are an APRI+ALBI score-based multivariable model trained in the NSQIP cohort, as well as models trained in 620 patients from 10 international centers and based on ICG clearance retention 15 min after administration (ICG-R15) and plasma disappearance rate (ICG-PDR), ALICE grade, and FIB-4 index, respectively. All models include age, sex, tumor type, and extent of resection as variables, as well as one of the respective liver function tests (APRI+ALBI, ICG-R15, ICG-PDR, ALICE, and FIB-4). The performance of each model is indicated as the AUC from ROC. 95% CI and tests between the AUC of APRI+ALBI versus the AUC of each of the other parameters have been performed by a bootstrap resampling analysis (lower panels).