AUTOMATED BODY COMPOSITION AND LIVER AND SPLEEN ANALYSIS ARE ASSOCIATED WITH MORTALITY, INCIDENT CIRRHOSIS, AND HEPATIC DECOMPENSATION IN METABOLIC DYSFUNCTION-ASSOCIATED STEATOTIC LIVER DISEASE

Background: Body composition measures including sarcopenia, visceral fat volume, and subcutaneous fat volume are associated with fibrosis stage in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). However, few studies have evaluated associations between anthropometrics and organ measurements have with risk of liver-related events (LREs) or incident cirrhosis. We assessed associations between body composition and hard outcomes in MASLD.

Methods: We included patients with MASLD diagnosed by computed tomography and metabolic risk factors at Michigan Medicine between 2010-2020. Scans were processed using fully automated algorithms within the analytic morphomics platform. The primary predictors were axial body composition measures derived from L3 and vessel and organ measures derived from segmented volumes. Axial measures included visceral fat, subcutaneous fat, and skeletal muscle area and attenuation. Organ measures included spleen size and liver size, attenuation, and eccentricity. All axial measures were normalized to standard deviation units stratified by sex and vertebra level. Outcomes were incident cirrhosis, LRE, or mortality. LRE and cirrhosis were modeled with a Fine-Grey competing risk model (with death without LRE/cirrhosis as a competing risk) and mortality with a Cox proportional hazards model, with a 6 month landmark for both. We assessed time-dependent area under the receiver operating characteristic curve (tAUC) at 5 years using clinical predictors (FIB4, BMI, age, sex, race), morphomic, and combined clinical/morphomic predictors.

Results: We included 4,419 individuals with MASLD and analyzable abdominal CT scans. Median age was 51 years, 54% were female, and 78% were white. Median body mass index was 32.6 kg/m². 25%, 51%, and 41% had diabetes, hypertension, and dyslipidemia, respectively. Median follow-up was 4.3 years. During follow-up, 75, 24, and 35 patients developed death, LRE, or incident cirrhosis. The strongest predictors of LREs and cirrhosis were spleen volume, liver volume and eccentricity, visceral and subcutaneous fat density, total body area, and muscle volume. Predictors of mortality included lower muscle area, higher subcutaneous fat density, lower bone mineral density, and higher aortic calcification. For incident cirrhosis prediction, addition of morphomic predictors yielded significantly higher tAUC (0.87 [95% CI 0.77-0.96]) than clinical predictors (0.76 [0.68-0.85]) alone; p=0.028 (Table). Combining clinical and morphomic predictors resulted in higher tAUC for mortality than clinical parameters alone: 0.85 (0.81-0.90) vs. 0.77 (0.71-0.83); p=0.00012.

Conclusions: Body composition obtained from clinical CT scans was associated with clinical outcomes in MASLD. Routinely incorporating high-dimensional morphomic data may improve risk stratification in MASLD.