In this multicenter study, we constructed a combined model (ModelMA−CI) to predict prognosis of HCC. In this model, the addition of sarcopenia and visceral adiposity improved the performance for both discrimination and calibration.
In clinical, preoperative prognostic evaluation is mainly based on patients' clinical factors, such as tumor stage 22 and potential liver function. When it comes to imaging indicators, we always focused on the exploration of tumor lesions and peritumoral zones but paid relatively little attention to sarcopenia and visceral adiposity which may reflect patients' nutritional status. In previous research, obesity has been shown to be a risk factor for various cancers, mainly in the digestive system, especially pancreatic 23 and liver cancers 24. Simultaneously, sarcopenia and obesity increase the mortality rate of cirrhosis25. Similar results were observed in patients with liver cancer who underwent liver transplantation26 or liver resection 27.
In this study, the AUC of the combined model was better than the clinical–imaging model in the external validation dataset. At the same time, the calibration of the combined model was also better than the metabolic model in the external validation dataset. Based on the two points of appeal, the combined model which incorporates clinical, imaging indicators, sarcopenia and visceral adiposity has a comprehensive and promising capacity in predicting prognosis. Meanwhile, the performance of the combined model was not influenced by different treatments or disease stages, which further proved its robustness under different conditions.
In our study, comparing the combined model and the clinical–imaging model using AUCs, we found that the addition of metabolic indicators improved the discrimination of the model. The related metabolic indicators are explained as follows: visceral adipose and subcutaneous adipose tissues are the two main types of adipose tissue. Inadequate subcutaneous fat is an independent risk factor for poor cancer prognosis in studies of relevant oncological microenvironment 28. Subcutaneous adipocytes may play a beneficial role in metabolism 29, which is similar to the results of our study. Adipose tissue is considered a secretory organ that produces pro-inflammatory and anti-inflammatory cytokines and adipokines. A high VSR value indicates that the fat distribution tends to be observed in the visceral area, which is often related to a poor prognosis 30. By analyzing CT images of the patient before treatment, it was possible to determine the condition of the tumor zone and evaluate the patient’s nutritional metabolism. From our study, the condition of muscle and adipose tissue is correlated with the prognosis of HCC, so provided nutritional support may be beneficial to the prognosis 31. Although it is not clear whether preoperative and postoperative interventions, such as nutritional therapy and rehabilitation, can improve postoperative results by changing obesity or muscle reduction, they are still worthy of attention.
In addition, comparing the combined model and the metabolic model by calibration showed that the addition of clinical–imaging indicators improved calibration. A higher BCLC stage, higher TBIL level 32, and more tumor nodules are associated with a poorer prognosis for HCC. A high TBIL level often indicates liver dysfunction. The capsule of liver cancer is often formed by the compression of the surrounding normal liver tissue. Intact capsules are often present in tumors with a low degree of malignancy, indicating that the tumor and other tissues are well demarcated and less aggressive 33.
Our study has some limitations. First, due to regional reason, almost all of the patients included in study had a history of hepatitis B. Whether our results were suitable for HCC related to hepatitis C still needed to be tested. Second, BCLC stages C patients were not included in our study, because according to the guidelines them cannot undergo TACE or liver resection. Third, treatments other than TACE and liver resection were not included in the study to control bias, but they deserved further exploration in follow-up studies. Four, the resolution of MRI for soft tissues, such as skeletal muscle and adipose, is higher than that of CT. In future studies, using MRI to evaluate may improve the accuracy and the predictive power of the model. Finally, in our study, skeletal muscle and adipose tissue were evaluated by two-dimensional imaging. Stereoscopic three-dimensional measurements will certainly provide more prognostic information.