Association between ASPEN/AND and GLIM malnutrition diagnostic criteria and CT-derived body composition features in pre-liver transplant patients.

Abstract

Background: Malnutrition is common in patients undergoing liver transplantation and is associated with poor outcomes. A diagnosis of malnutrition is primarily intended to capture those in need of nutrition support by identifying a loss of lean mass, which reflects a depleted pool of amino acids needed to support metabolic demands of disease. Computed tomography (CT)-derived measures of skeletal muscle have also been associated with liver transplant outcomes, but their relationship with malnutrition diagnoses in a liver transplant population is largely unexplored. Therefore, we evaluated the association between the American Society for Parenteral and Enteral Nutrition and the Academy of Nutrition and Dietetics (ASPEN/AND) and the Global Leadership Initiative on Malnutrition (GLIM) malnutrition diagnosis and CT-derived body composition features and identified factors that predict this relationship in a pre-liver transplant cohort. Additionally, we investigated whether incorporating CT measures alongside standard malnutrition assessments can better predict liver transplant outcomes. Method: Liver transplant patients (n=91) underwent CT scans within 90 days before transplant. Malnutrition assessments based on ASPEN/AND and GLIM criteria were conducted within 48 hours of transplant. A freehand region-of-interest was used to generate measures of psoas area and density (mean Hounsfield units; mHU) from CT scans. Psoas area index (PAI) was calculated by dividing psoas area (cm2) by height in meters squared (m2). Forward stepwise regression was performed to identify specific predictors of the relationship between malnutrition diagnosis and CT-derived body composition features. Predictors included in the stepwise regression were age, gender, Body Mass Index (BMI), laboratory-derived Model for End-Stage Liver Disease (MELD-Na) score, height, and weight. Forward stepwise regression was also performed to determine whether the inclusion of PAI and mHU improved the ability of malnutrition to predict outcomes such as predictive ability of length of stay (LOS) and discharge disposition. Results: Malnutrition diagnosis, based on both ASPEN/AND and GLIM criteria, was associated with lower PAI (ASPEN/AND: p < 0.001; GLIM: p = 0.02). ASPEN/AND malnutrition was not associated with psoas density (mHU: p = 0.11). In contrast, GLIM malnutrition showed a weak association with mHU, though with a negligible effect size (p < 0.01; R2 = 0.04). Other predictors selected into the model included gender, BMI and MELD-Na. Malnutrition, under both criteria, was weakly predictive of LOS (ASPEN/AND: p = 0.02; R2 = 0.06; GLIM: p = 0.14; R2 = 0.04) and discharge disposition (ASPEN/AND: p < 0.01; R2 = 0.07; GLIM: p = 0.17; R2 = 0.01). Inclusion of mHU significantly improved the prediction of LOS (p < 0.0001; R2 = 0.19 for ASPEN/AND; p < 0.001; R2 = 0.17 for GLIM), but did not improve discharge disposition prediction for either (ASPEN/AND: p = 0.54; R2 = 0.06; GLIM: p = 0.63; R2 = 0.00). Inclusion of PAI did not improve either prediction for either ASPEN/AND (LOS: p = 0.01; R2 = 0.07; discharge disposition: p = 0.38; R2 = 0.06) or GLIM (LOS: p = 0.12; R2 = 0.05; discharge disposition: p = 0.29; R2 = 0.01). Conclusion: The ASPEN/AND and GLIM malnutrition criteria appear to capture CT-derived muscle area but not density. The inclusion of mHU with malnutrition diagnosis strengthened the ability to predict LOS. This finding supports considering the inclusion of CT-derived mHU to complement malnutrition diagnosis in patients undergoing liver transplantation, which could improve how malnutrition criteria can identify those most in need of nutrition support.