Malnutrition criteria and computed tomography-measured psoas muscle indices in adult liver transplant recipients

Abstract

Background:

Malnutrition is prevalent in liver transplant candidates and associated with poor outcomes. Computed tomography (CT)-derived skeletal muscle measures predict liver transplant outcomes, but their relationship with malnutrition diagnoses remains unclear. We evaluated associations between The American Society for Parenteral and Enteral Nutrition and the Academy of Nutrition and Dietetics Indicators for Malnutrition (AAIM) and the Global Leadership Initiative on Malnutrition (GLIM) defined malnutrition and CT-derived muscle features. Additionally, we investigated whether incorporating muscle area or density measures alongside malnutrition assessments improves prediction of transplant outcomes.

Methods:

Ninety-one patients who underwent CT within 90 days before liver transplant were assessed for malnutrition within 48 hours of transplantation. Psoas muscle area indexed to height (psoas area index) and muscle density were captured. Stepwise regression evaluated associations between malnutrition, muscle measures, anthropometric measures, and whether muscle measures improved outcome predictions.

Results:

Malnutrition was associated with lower psoas area index for both criteria (AAIM: p<0.001; GLIM: p=0.02), and psoas density with GLIM criteria (p<0.01). Malnutrition weakly predicted length of stay (p=0.02, R²=0.06; p=0.14, R²=0.04) and discharge disposition (p<0.01, R²=0.07; p=0.17, R²=0.01). Adding muscle density substantially improved length of stay prediction (p<0.0001, R²=0.19; p<0.001, R²=0.17).

Conclusion:

Malnutrition criteria seem to better reflect CT-derived measures of psoas muscle area compared to density. Including muscle density measures alongside malnutrition assessments may improve clinical outcome predictions in liver transplant patients and allow for development of more targeted nutrition interventions.

1. Introduction

Malnutrition is highly prevalent among liver transplant candidates and impacts the prognosis of end-stage liver disease.^{1, 2} Up to 80% of patients with cirrhosis, the most common indication for liver transplantation, are malnourished, and approximately 5–40% of other etiologies facing liver transplantation are accompanied by malnutrition.^{2, 3} To standardize malnutrition diagnosis, two prominent frameworks are utilized: The American Society for Parenteral and Enteral Nutrition and the Academy of Nutrition and Dietetics Indicators for Malnutrition (AAIM) and the Global Leadership Initiative on Malnutrition (GLIM). AAIM defines malnutrition as the presence of at least 2 of 6 criteria: weight loss, reduced energy intake, loss of muscle mass, loss of subcutaneous fat, fluid accumulation and loss of grip strength.⁴ GLIM defines malnutrition as the presence of one phenotypic criterion (non-volitional weight loss, low body mass index, or reduced muscle mass) and one etiologic criterion (reduced food intake or assimilation, or disease burden/inflammation).⁵ While both AAIM and GLIM criteria provide a structured approach to diagnose malnutrition, it remains unknown how effectively these criteria capture objective computed tomography (CT)-derived measures of muscle mass and quality in a liver transplant population.

Computed tomography (CT)-derived skeletal muscle measures provide an objective and quantitative measurement of muscle area and density. This quantitative loss of muscle area has been associated with poor outcomes post-liver transplantation, such as longer stay in the intensive care unit (ICU) and greater hospital mortality.^6–8 Similarly, lower muscle densities, a measure of reduced muscle quality, have been linked to a greater risk of complications and mortality, longer ICU and hospital length of stay, and increased hospital cost compared to other measures in liver transplant recipients.^9–11 This study utilized CT-derived measures of psoas muscle area and density as a surrogate for total skeletal muscle. This choice is methodologically advantageous because psoas muscle can be easily and reliably quantified by clinicians, such as dietitians, using standard radiologic imaging software, and its measures are known to be comparable to total skeletal muscle measures at the third lumbar vertebra.^{12, 13} Understanding the degree of concordance between the established AAIM and GLIM criteria and these validated, objective CT-derived measures is essential for refining clinical malnutrition criteria and developing more targeted nutritional interventions.^{4, 5, 14}

Therefore, this observational cohort study investigated the concordance between established malnutrition criteria (AAIM and GLIM) and CT-derived measures of psoas muscle area and density. We hypothesize that AAIM and GLIM incompletely reflect these quantitative CT measures, and consequently, that the integrated use of CT-derived measures alongside the AAIM and GLIM criteria improves the prediction of patient outcomes. This work builds on our previous work demonstrating the prognostic value of CT-derived psoas muscle measures for clinical outcomes in a subset of this population.⁸ This study uses a 90 vs 180 days before transplant window for CT-derived measures and expands upon the previous cohort by including contrast and non-contrast CTs to evaluate the diagnostic capacity of both the GLIM and AAIM malnutrition criteria. In addition, and of particular clinical relevance, we assess whether combining objective CT-based skeletal muscle measures with GLIM and AAIM-based diagnoses improves predictive value for post-transplant outcomes.

2. Materials and Methods

2.1. Study design

We performed a descriptive cohort study in 91 adult liver transplant recipients, who represent a sub cohort of 368 patients, who underwent computed tomography derived muscle measurements at the third lumbar vertebra at the University of Minnesota Medical Center-Fairview from January 1, 2016, to May 30, 2021. The study was approved by the University of Minnesota Institutional Review Board of the (Protocol # STUDY00016254), adhering to the Declaration of Helsinki principles. Written informed consent was waived as the research poses no more than minimal risks to participants. Those who received combined liver-kidney transplantation were excluded from the study. Patients with non-contrast and contrast-enhanced abdominal CT scans within 90 days before transplant were included for analysis (n = 91). The use of CT-scans obtained within 90 days before transplant was chosen to align with previous studies, which typically use pre-transplant CT windows of 90 days to 1-year.^{14, 15}

2.2. CT analysis

Images were analyzed according to previously described methods.^{8, 12} All measurements were performed by the same individual J.T. who was trained by L.T. Briefly, the area of the psoas muscle was measured at the axial mid-L3 level using a freehand region-of-interest method. The total psoas area was the sum of the left and right psoas areas. The psoas area index was calculated as the total psoas area divided by height in square meters (cm²/m²). Mean Hounsfield units (i.e. muscle density) were generated using the following equation: [(right psoas HU*right psoas area) + (left psoas HU*left psoas area)] / (right psoas area + left psoas area). For contrast-enhanced abdominal CTs, a correction factor of −7.5 was applied to all mean Hounsfield unit measures.¹⁶ The correction factor was applied to contrast CT images captured during the arterial or venous phases.¹⁶

2.3. Data collection

Patient demographic information and disease characteristics were obtained from a liver transplant database maintained by the primary healthcare facility. Malnutrition diagnosis was collected through manual electronic health record review. Malnutrition was assessed, diagnosed, and documented within 48 hours of transplantation by a registered dietitian using the AAIM or GLIM malnutrition diagnostic criteria.^{4, 5} Regarding the GLIM malnutrition diagnosis, while validated body composition imaging methods are suggested, phenotypic criteria in this study were assessed through physical examination, a recommended alternative.⁵ Additional clinical and anthropometric measures collected included age, Body Mass Index, Model for End-Stage Liver Disease (MELD)-Na score, sex, weight, and the number of days between CT imaging and liver transplantation.

2.4. Outcomes

Our primary outcome was AAIM or GLIM defined malnutrition. Our primary objective was to evaluate the cross-sectional association between CT-derived measures of psoas muscle area and density and AAIM or GLIM defined malnutrition.

Our secondary objective was to assess the incremental predictive value of incorporating CT-derived psoas muscle measures (area and density) alongside AAIM or GLIM malnutrition diagnoses in relation to the key clinical outcomes hospital length of stay and discharge disposition (home vs. facility). Outcomes were obtained from a liver transplant database maintained by University of Minnesota Medical Center-Fairview and manual chart review.

2.5. Statistical analysis

All statistical analyses were performed using R version 4.2.3 (R Core Team, Vienna, Austria). Significance testing was performed with the use of a two-sided alpha level of 0.05. Missing values for discharge disposition (n=2; mortality) were excluded from the analyses including discharge disposition as an outcome. Baseline descriptive statistics (mean and standard deviation for continuous variables or frequency for categorical variables) were used to assess demographic and clinical characteristics. Pearson correlation coefficients were calculated to assess relationships between AAIM or GLIM malnutrition criteria and CT-derived psoas muscle features – psoas area index and density. Forward stepwise regression was performed to identify specific predictors of the relationship between CT-derived muscle features and malnutrition diagnosis (CT-measure as independent variable and malnutrition diagnosis as dependent variable). The clinical and anthropometric predictors that were considered for inclusion in the stepwise regression were age, sex, height, weight, body mass index, laboratory-derived MELD-Na, and time from CT imaging to transplant. For our secondary aim, a second set of models was constructed using forward stepwise regression to examine the relationship between malnutrition diagnosis and hospital length of stay and discharge status and to determine if incorporating psoas area index and muscle density alongside malnutrition diagnosis influenced prediction of these outcomes. For this set of models, malnutrition diagnosis and CT measures were treated as independent variables and length of stay and discharge status as dependent variables. A logistic regression model was used for discharge status due to its nature as a binary outcome. Pearson’s Chi-squared test was performed to assess the distribution of the AAIM compared to the GLIM malnutrition criteria.

3. Results

3.1. Baseline characteristics

A total of 91 patients met the inclusion criteria out of the 368 reviewed (Table 1). The patient population was predominantly male (62%) with a mean (SD) age of 55.5 (11.1). Alcohol-associated cirrhosis (33%) was the most common primary liver diagnosis The mean (SD) MELD-Na score for patients undergoing liver transplant was 32.7 (6.8), indicating critical condition, high risk of mortality, and the need for transplantation. CT scans were obtained, on average (SD), 22.4 (24.6) days before transplantation (range: 0–90). Severe malnutrition was the most common AAIM classification (52%) and moderate malnutrition for GLIM criteria (52%). Within the AAIM, 38% were classified as having chronic disease-related, 17% as acute disease-related, and 45% as acute-on-chronic disease-related malnutrition. In contrast, the GLIM utilized inflammation criteria within its diagnostic framework.^{4, 5}

Table 1.

Summary of baseline characteristics and outcome variables.

	Total cohort (n = 91)
	Variable No. (%)/Variable Mean (SD)
Sex
Female	35 (38%)
Male	56 (62%)
Age	55.5 (11.1)
BMI	28.6 (5.5)
MELD-Na	32.7 (6.8)
AAIM Malnutrition
None	19 (21%)
Moderate	25 (27%)
Severe	47 (52%)
GLIM Malnutrition (n = 88)
None	21 (24%)
Moderate	46 (52%)
Severe	21 (24%)
Primary liver disease
Alcoholic cirrhosis	30 (33%)
NASH	11 (12%)
Hepatocellular carcinoma	19 (21%)
Other	31 (34%)
CT scan days pre-transplant	22.4 (24.6)
LOS	26.4 (20.3)
Discharge disposition (n = 89)
Home	47 (53%)
Rehab or long-term care	42 (47%)

Abbreviations: AAIM, Association of Parenteral and Enteral Nutrition and Academy of Nutrition and Dietetics Indicators of Malnutrition; BMI, Body Mass Index; CT, Computed tomography; GLIM, Global Leadership Initiative on Malnutrition; LOS, Length of Stay; MELD-Na, Model for End-Stage Liver Disease – Sodium; NASH, Non-Alcoholic Steatohepatitis.

3.2. CT-derived body composition measures captured by AAIM and GLIM malnutrition criteria

AAIM and GLIM malnutrition were correlated with psoas area index (R² = 0.14, p-value < 0.01; R² = 0.08, p = 0.02; Figure 1A and 2A, respectively). AAIM malnutrition and muscle density were not correlated (R² = 0.00, p = 0.11) (Figure 1B). GLIM malnutrition achieved statistical significance in predicting variability in muscle density, but with small effect size (R² = 0.04, p-value < 0.01) (Figure 2B).

Figure 1A/B. CT-derived measures of psoas area index (A) and muscle density (B) by AAIM malnutrition diagnosis in liver transplant recipients (n = 91).

Box plot demonstrating distribution of Psoas Area Index (A) and Muscle Density (B) by malnutrition severity category by AAIM malnutrition diagnosis. Statistical differences were assessed using one-way analysis of variance (ANOVA), followed by Tukey’s post-hoc pairwise comparisons. Asterisks indicate significance levels: * p < 0.05; ** p < 0.01; *** p < 0.001. Comparing across malnutrition severity categories, psoas area index was higher in patients with no malnutrition compared to moderate or severe (p<0.05 and p<0.001, respectively). No differences were observed between moderate and severe malnutrition categories.

Abbreviations: AAIM, Association of Parenteral and Enteral Nutrition and Academy of Nutrition and Dietetics Indicators of Malnutrition; mHU, Mean Hounsfield Units; PAI, Psoas Area Index.

Figure 2A/B. CT-derived measures of psoas area index (A) and muscle density (mHU; B) by GLIM malnutrition diagnosis in liver transplant recipients (n = 91).

Box plot demonstrating distribution of Psoas Area Index (A) and Muscle Density (B) by malnutrition severity category by GLIM malnutrition diagnosis. Statistical differences were assessed using one-way analysis of variance (ANOVA), followed by Tukey’s post-hoc pairwise comparisons. Asterisks indicate significance levels: * p < 0.05; ** p < 0.01; *** p < 0.001. Comparing across malnutrition severity categories, psoas area index was higher in patients with no malnutrition compared to moderate or severe (p<0.05 and p<0.05, respectively). No differences were observed between moderate and severe malnutrition categories.

Abbreviations: GLIM, Global Leadership Initiative on Malnutrition; mHU, Mean Hounsfield Units; PAI, Psoas Area Index.

In stepwise regression models, AAIM malnutrition was associated with psoas area index (R² = 0.14, β = – 1.06, RMSE = 1.31, p-value < 0.01) (Table S1). Sex (R² = 0.31, beta = 1.12, RMSE = 1.16, p-value < 0.01) and body mass index (R² = 0.38, β = 0.08, RMSE = 1.10, p-value < 0.01) were also included in the model, and both predicted psoas area index. Body mass index was the only significant predictor of muscle density, although weakly (R² = 0.07, β = – 0.45, RMSE = 8.67, p = 0.01) and malnutrition was not a significant predictor of muscle density (R² = 0.00, β = – 3.87, RMSE = 9.04, p = 0.11) (Table S2).

GLIM malnutrition was associated with psoas area index in stepwise regression models, however the effect size was relatively weak (R² = 0.08, β = – 0.77, RMSE = 1.36, p = 0.02) (Table S3). Sex (R² = 0.29, β = 1.23, RMSE = 1.18, p-value < 0. 01) and body mass index (R² = 0.37, β = 0.08, RMSE = 1.11, p-value < 0.01) both predicted psoas area index as they did for AAIM. Body mass index (R² = 0.13, β = – 0.53, RMSE = 8.47, p-value < 0.01) and MELD-Na predicted mHU (R² = 0.17, β = – 0.29, RMSE = 8.24, p = 0.03) (Table S4). GLIM malnutrition was also a predictor of muscle density with small effect size (R² = 0.04, β = – 6.83, RMSE = 8.97, p-value < 0.01) (Table S4).

3.3. AAIM or GLIM malnutrition criteria and length of stay and discharge status liver transplant outcomes

In stepwise regression models, a diagnosis of AAIM malnutrition was weakly predictive of length of stay (R² = 0.06, β = 13.29, RMSE = 19.4, p = 0.01; Table 2) and discharge disposition (R² = 0.07, β = 1.53, RMSE = 0.48, p-value = 0.01; Table S5). The inclusion of psoas area index in the model did not impact the ability of AAIM malnutrition to predict length of stay (R² = 0.07, β = 2.18, RMSE = 19.2, p = 0.17) nor discharge disposition (R² = 0.06, β = 0.15, RMSE = 0.48, p = 0.38) (Table 2 and Table S5). When muscle density was selected for inclusion, the model was more predictive of length of stay (R² = 0.19, β = 0.81, RMSE = 18.0, p-value <0.01; Table 3) but not discharge disposition (R² = 0.06, β = 0.02, RMSE = 0.48, p = 0.54; Table S6).

Table 2.

Stepwise regression models for association between AAIM malnutrition and LOS with selected inclusion of PAI.

Step	Variables	Regression Equation	P-value	Adj. R2	RMSE
1	Malnutrition	13.29(Malnutrition) + 15.90	0.01*	0.06	19.4
2	Psoas Area Index (PAI)	10.42(Malnutrition) − 2.183(PAI) + 28.41	0.17	0.07	19.2

Significance determined from forward stepwise linear regression using a two-sided alpha level of 0.05.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Abbreviations: AAIM, Association of Parenteral and Enteral Nutrition and Academy of Nutrition and Dietetics Indicators of Malnutrition; LOS, Length of Stay; PAI, Psoas Area Index.

Table 3.

Stepwise regression models for association between AAIM malnutrition and LOS with selected inclusion of mHU.

Step	Variables	Regression Equation	P-value	Adj. R2	RMSE
1	Malnutrition	11.11(Malnutrition) + 43.35	0.02*	0.06	19.4
2	Psoas Density (mHU)	11.11(Malnutrition) − 0.814(mHU) + 43.35	<0.01***	0.19	18.0

Significance determined from forward stepwise linear regression using a two-sided alpha level of 0.05.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Abbreviations: AAIM, Association of Parenteral and Enteral Nutrition and Academy of Nutrition and Dietetics Indicators of Malnutrition; LOS, Length of Stay; mHU, Mean Hounsfield Units.

GLIM malnutrition was not predictive of length of stay (R² = 0.04, β = 7.92, RMSE = 19.9, p = 0.14), nor of discharge disposition (R² = 0.01, β = – 0.72, RMSE = 0.49, p = 0.17) (Table 4 and Table S7, respectively). The selected inclusion of psoas area index in the model did not impact the ability of GLIM malnutrition to predict length of stay (R² = 0.05, β = 2.49, RMSE = 19.7, p = 0.12; Table 4) nor discharge disposition (R² = 0.01, β = 0.17, RMSE = 0.49, p = 0.29; Table S7). When muscle density was selected for inclusion, the model was more predictive of length of stay (R² = 0.17, β = 0.87, RMSE = 18.4, p-value < 0.01; Table 5) but not predictive of discharge disposition (R² = 0.00, β = 0.01, RMSE = 0.49, p = 0.63; Table S8).

Table 4.

Stepwise regression models for association between GLIM malnutrition and LOS with selected inclusion of PAI.

Step	Variables	Regression Equation	P-value	Adj. R2	RMSE
1	Malnutrition	7.917(Malnutrition) + 31.87	0.14	0.04	19.9
2	Psoas Area Index (PAI)	7.917(Malnutrition) − 2.486(PAI) + 31.87	0.12	0.05	19.7

Significance determined from forward stepwise linear regression using a two-sided alpha level of 0.05.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Abbreviations: GLIM, Global Leadership Initiative on Malnutrition; LOS, Length of Stay; PAI, Psoas Area Index.

Table 5.

Stepwise regression models for association between GLIM malnutrition and LOS with selected inclusion of mHU.

Step	Variables	Regression Equation	P-value	Adj. R2	RMSE
1	Malnutrition	6.318(Malnutrition) + 48.89	0.19	0.04	19.9
2	Psoas Density (mHU)	6.318(Malnutrition) − 0.868(mHU) + 48.89	<0.01***	0.17	18.4

Significance determined from forward stepwise linear regression using a two-sided alpha level of 0.05.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Abbreviations: GLIM, Global Leadership Initiative on Malnutrition; LOS, Length of Stay; mHU, Mean Hounsfield Units.

3.4. Association between AAIM and GLIM malnutrition criteria

Pearson’s Chi-squared test was performed to evaluate the association between AAIM and GLIM malnutrition diagnostic criteria. No significant differences in distribution of malnutrition diagnoses between AAIM and GLIM malnutrition criteria were found (X² = 3.28, df = 4, p-value = 0.51; Table S9).

4. Discussion

In this retrospective analysis, we found that malnutrition diagnosis based on AAIM was associated with lower CT-derived psoas area index but not muscle density. GLIM malnutrition was associated with lower CT-derived psoas area index and showed a statistically significant, albeit limited, association with CT-derived psoas muscle density. Additional predictors, such as body mass index, sex, and MELD-Na, slightly strengthened the relationship between malnutrition and CT-derived psoas muscle measures. AAIM had limited associations with post-transplant outcomes, length of hospital stay and discharge disposition, while GLIM criteria was not predictive of these outcomes. However, when muscle density was added alongside both malnutrition criteria, it improved the prediction of length of stay. In contrast, the addition of psoas area index did not improve the model. This finding suggests a potential role for CT-derived muscle density measures as a complement to AAIM and GLIM malnutrition diagnoses in liver transplant populations.

The association of malnutrition diagnosis with lower psoas area index values demonstrates its ability to capture reduced skeletal muscle mass. This aligns with the AAIM and GLIM malnutrition criteria, which utilize loss of muscle mass as an indicator of malnutrition.^{4, 5} However, lower muscle density (considered a surrogate for muscle quality) can still be present with preserved muscle mass due to differences in metabolic processes and composition of the muscle.^{17, 18} The use of muscle mass to estimate available amino acid reserves assumes a homogenous composition. These findings highlight the importance of incorporating muscle density measures as a complement to traditional malnutrition criteria to produce a more comprehensive assessment of malnutrition and available protein stores.

In our study, the AAIM did not capture muscle density, while GLIM only did so weakly. GLIM can be mainly differentiated from AAIM by the inclusion of body mass index, the suggested use of quantitative lean mass or skeletal muscle measurements (depending on the methodology chosen^{5, 19}), and the use of inflammation as an indicator. Low body mass index served as an indicator for only 2 of our patients, and in each of those cases they met GLIM criteria for malnutrition based on other indicators. Given that nutrition focused physical assessment was used to identify reduced muscle mass, the ability of GLIM to seemingly better capture muscle density in this instance may be related to the inclusion of inflammation in the diagnostic criteria.^{4, 5} Notably, muscle density has been shown to be a better predictor of liver transplant outcomes compared to muscle mass.^{9, 18} Therefore, a malnutrition diagnosis, when used in isolation, may have limitations in predicting outcomes for liver transplant patients due to its emphasis on muscle mass.

Our study showed that including CT-derived tissue density measures alongside malnutrition diagnosis in prediction models for liver transplant outcomes such as length of stay significantly improved the model’s predictive ability. CT-derived measures of muscle density have been strongly associated with clinical outcomes in literature, such as length of stay, discharge disposition, transplant complications and 90-day mortality.^{14, 18}

While CT-derived measures can provide robust measures of body composition, their use for routine assessment is limited by cost and radiation exposure. Therefore, there is a need to identify surrogate measures of muscle density that can be used to enhance prediction models for liver transplant outcomes. Handgrip strength offers a potential surrogate measure of muscle function, and the AAIM does specifically include handgrip strength as a functional criterion, but fewer than 40% of dietitians include a measure of function in their nutritional-focused physical exam.²⁰ Because the AAIM only require two of the six criteria for diagnosis, there are inconsistencies in the application of functional measures such as handgrip strength.⁴ Underscoring the limited role handgrip strength plays in application of the AAIM diagnosis, investigators have found removal of handgrip strength and edema from the criteria does not meaningfully change malnutrition diagnosis.²¹ This is likely due to challenges in obtaining this specific measure relative to the other criteria. Similarly, while GLIM malnutrition criteria recommend the use of techniques for measuring lean or muscle mass, the criteria still overlooks direct measures of muscle quality.⁵

Recent work has demonstrated the use of the Liver Frailty Index^™, which combines handgrip strength with timed chair stands and balance testing, is able to predict liver transplant outcomes and inform pre-transplant interventions.^22–24 Liver Frailty Index was designed for end-stage liver disease patients awaiting liver transplant and uses three objective performance measures (handgrip strength, timed chair stands and balance) to create an overall frailty score.²² This score is well-studied, reproducible and reliable, specifically in the liver transplant patient population.²⁵ This tool could potentially provide a standardized approach to measuring muscle quality and serve as a robust complement, in place of CT-based skeletal muscle density measures, to standard malnutrition criteria resulting in more comprehensive nutrition assessment of liver transplant patients.²⁶

Limitations of this study include its retrospective design and the use of participant outcomes from a single site. Further, the outcome measures length of stay and discharge disposition are only two of many post-transplant outcomes measures and can be influenced by factors independent of nutritional status and muscle parameters. However, length of stay and discharge disposition reflect high impact outcome measures tied to healthcare burden and costs.²⁷ The use of a window of up to 90 days prior to transplant represents an important consideration for future research as patients awaiting liver transplantation represent a metabolically unstable cohort that may experience significant body composition changes over a 90-day period. Further, the requirement for CT imaging for inclusion in the study may have selected for more acutely ill or closely monitored patients. Additionally, inconsistent use of handgrip strength and prioritization of other criteria to meet the two of six thresholds for diagnosing malnutrition with AAIM could have influenced our findings. However, as noted earlier, because the AAIM only requires 2 of 6 criteria it is rare that handgrip strength serves as a determining factor in diagnosis. Our findings suggest that CT-derived muscle density, or muscle quality/function may be best assessed independent from mass. Further research is needed to investigate the utility of incorporating functional measures, such as the Liver Frailty Index, alongside malnutrition criteria to improve patient assessment by potentially capturing both muscle quality and mass, respectively.

5. Conclusion

The AAIM captured CT-derived psoas muscle area but did not capture psoas muscle density, a surrogate of muscle quality. The GLIM criteria captured psoas area index, and showed a statistically significant, albeit small, association with muscle density. The inclusion of muscle density but not psoas area index with both malnutrition diagnoses strengthened the prediction of length of hospital stay. This finding supports the inclusion of CT-derived muscle density to complement malnutrition diagnosis in patients undergoing liver transplantation. Further research should explore the utility of including a distinct functional assessment such as the Liver Frailty Index alongside malnutrition criteria in clinical settings to target muscle quality as a complement to muscle mass measures captured by the AAIM or GLIM criteria. Addressing these gaps in malnutrition criteria will enhance its predictive ability and better inform pre-transplant interventions in efforts to reduce adverse outcomes.

Clinical Relevancy Statement:

Malnutrition is common in liver transplant patients and contributes to clinical outcomes. This study shows that current malnutrition diagnostic criteria are associated with CT-derived muscle area but not density. Incorporating muscle density measures alongside malnutrition diagnosis improved the prediction of length of stay. These findings underscore the value of integrating CT-derived muscle density measures into nutrition assessments. By combining these objective measures with traditional assessment tools, healthcare providers can enhance the accuracy of malnutrition diagnoses and tailor pre- and post-transplant interventions more effectively to reduce adverse outcomes.