3D assessment of skeletal muscle and adipose tissue for prognosis of hepatocellular carcinoma: a multicenter cohort study.

Abstract

Background: Changes in protein and lipid metabolism could provide additional prognostic information for hepatocellular carcinoma (HCC).This study aimed to explore whether 3D automatic assessment of skeletal muscle and adipose tissue can contribute to the precise prognosis for HCC.

Methods: The data of 458 HCC patients from 6 hospitals were divided into training and external validation datasets. Preoperative CT Images were used for this study. First, we tested the stability of the 2D factors. Second, we tested whether standardization for volume assessment was necessary. Third, we compared the clinical (Model^C), skeletal muscle and adipose tissue (Model^NSA), and combined (Model^C-NSA) models by discrimination and calibration to identify the optimal model. Subgroup analysis was performed for the optimal model.

Results: For the 16 2D factors, 13 factors were statistically different among the three 2D slices. Standardization of the volume factors was necessary. Among the three models, Model^C-NSA had a higher area under the curve [AUC] than Model^C and Model^NSA, both in the training dataset (0.809 vs. 0.649 vs. 0.797) and the validation dataset (0.770 vs. 0.718 vs. 0.719). For calibration, the performance of Model^C-NSA was similar to those of Model^C and Model^NSA. The performance of Model^C-NSA was not influenced by age (P=0.753), sex (P=0.781), treatments (P=0.504), Barcelona Clinic Liver Cancer stage (P=0.913), or Child-Pugh class (P=0.580).

Conclusions: Compared to 2D evaluation, 3D assessment is more stable. 3D automatic assessment of skeletal muscle and adipose tissue can accurately predict progression in patients with HCC.