Objective To quantify the potential of fully automated CT-based body composition metrics and clinical frailty data in predicting liver transplant recipient postoperative outcomes. Methods AI-enabled body composition tools were applied to pre-transplant abdominal CT scans in a retrospective cohort of first-time deceased-donor liver transplant recipients. Clinical frailty data (Fried frailty score) was obtained from an established transplant database. Age- and sex-corrected hazard ratios (HRs) were analyzed according to highest-risk quartiles compared with the other three quartiles combined. Area under the receiver operating characteristic curve (ROC AUC) analysis in univariate and multivariate scenarios was also performed. Results 598 liver transplant recipients (median age, 56 years [IQR, 49–61]; 383 men/215 women) were included from 2005 to 2021. Mean clinical follow-up interval after transplant was 8.6 ± 4.5 years, with 224 deaths (mean interval, 5.3 ± 3.9 years post-transplant) and 246 graft failures (mean interval, 4.7 ± 4.0 years post-transplant) observed. Univariate HRs for post-transplant survival included 1.53 (95% CI, 1.14–2.06) for muscle attenuation, 1.66 (95% Cl, 1.24–2.22) for aortic Agatston score, 1.35 (1.02–1.80) for SAT area, and 1.82 (1.35–2.46) for liver volume. For those meeting the frailty criteria, HR was 2.14 (1.08–4.22). Multivariate 10-year AUC for predicting mortality was 0.675 using liver volume, aortic Agatston score, and muscle attenuation. 10-year univariate AUC for clinical frailty assessment was 0.601 but increased to 0.878 when combined with CT measures. Conclusion Automated CT measurements of muscle density (myosteatosis), aortic calcification, subcutaneous fat, and liver volume are predictive of mortality in liver transplant recipients. Frailty was likewise predictive. Combining CT and clinical frailty assessment was complementary. Key Points <Emphasis Type="BoldItalic">Question</Emphasis> What is the prognostic value of pre-transplant CT-based body composition measures for deceased-donor liver transplant outcomes, and how do they correlate with frailty assessment? <Emphasis Type="BoldItalic">Findings</Emphasis> Increased post-transplant mortality was associated with pre-transplant increased liver volume, increased abdominal aortic Agatston score, decreased skeletal muscle attenuation, and decreased subcutaneous adipose tissue area. <Emphasis Type="BoldItalic">Clinical relevance</Emphasis> Pre-transplant AI-enabled body composition measures have predictive value for post-transplant survival, offering a novel and objective diagnostic tool to identify high-risk transplant recipients that are complementary to clinical assessments. Graphical Abstract

Automated abdominal CT imaging biomarkers and clinical frailty measures associated with postoperative deceased-donor liver transplant outcomes

期刊论文
发表日期：

2025年3月23日

查看全文

Daniel Liu,
David Ji,
John W. Garrett,
Ryan Zea,
Adam Kuchnia,
Ronald M. Summers,
Joshua D. Mezrich,
Perry J. Pickhardt

摘要

Objective

To quantify the potential of fully automated CT-based body composition metrics and clinical frailty data in predicting liver transplant recipient postoperative outcomes.

Methods

AI-enabled body composition tools were applied to pre-transplant abdominal CT scans in a retrospective cohort of first-time deceased-donor liver transplant recipients. Clinical frailty data (Fried frailty score) was obtained from an established transplant database. Age- and sex-corrected hazard ratios (HRs) were analyzed according to highest-risk quartiles compared with the other three quartiles combined. Area under the receiver operating characteristic curve (ROC AUC) analysis in univariate and multivariate scenarios was also performed.

Results

598 liver transplant recipients (median age, 56 years [IQR, 49–61]; 383 men/215 women) were included from 2005 to 2021. Mean clinical follow-up interval after transplant was 8.6 ± 4.5 years, with 224 deaths (mean interval, 5.3 ± 3.9 years post-transplant) and 246 graft failures (mean interval, 4.7 ± 4.0 years post-transplant) observed. Univariate HRs for post-transplant survival included 1.53 (95% CI, 1.14–2.06) for muscle attenuation, 1.66 (95% Cl, 1.24–2.22) for aortic Agatston score, 1.35 (1.02–1.80) for SAT area, and 1.82 (1.35–2.46) for liver volume. For those meeting the frailty criteria, HR was 2.14 (1.08–4.22). Multivariate 10-year AUC for predicting mortality was 0.675 using liver volume, aortic Agatston score, and muscle attenuation. 10-year univariate AUC for clinical frailty assessment was 0.601 but increased to 0.878 when combined with CT measures.

Conclusion

Automated CT measurements of muscle density (myosteatosis), aortic calcification, subcutaneous fat, and liver volume are predictive of mortality in liver transplant recipients. Frailty was likewise predictive. Combining CT and clinical frailty assessment was complementary.

Key Points

Question What is the prognostic value of pre-transplant CT-based body composition measures for deceased-donor liver transplant outcomes, and how do they correlate with frailty assessment?

Findings Increased post-transplant mortality was associated with pre-transplant increased liver volume, increased abdominal aortic Agatston score, decreased skeletal muscle attenuation, and decreased subcutaneous adipose tissue area.

Clinical relevance Pre-transplant AI-enabled body composition measures have predictive value for post-transplant survival, offering a novel and objective diagnostic tool to identify high-risk transplant recipients that are complementary to clinical assessments.

Graphical Abstract