Comparison of Models for Predicting Quality of Life After Surgical Resection of Hepatocellular Carcinoma: a Prospective Study.

BACKGROUND

The essential issue of internal validity has not been adequately addressed in prediction models such as artificial neural network (ANN), support vector machine (SVM), Gaussian process regression (GPR), and multiple linear regression (MLR) models.

METHODS

This prospective study compared the accuracy of these four models in predicting quality of life (QOL) after hepatic resection received by 332 patients with hepatocellular carcinoma (HCC) during 2012-2015. An estimation subset was used to train the models, and a validation subset was used to evaluate their performance. Sensitivity score approach was also used to assess the relative significance of input parameters in the system models.

RESULTS

The ANN model had significantly higher performance indicators compared to the SVM, GPR, and MLR models (P < 0.05). Additionally, the ANN prediction of QOL at 6 months after hepatic resection significantly correlated with age, gender, marital status, Charlson comorbidity index (CCI) score, chemotherapy, radiotherapy, hospital volume, surgeon volume, and preoperational functional status (P < 0.05). Preoperational functional status was the most influential (sensitive) variable affecting sixth-month QOL followed by surgeon volume, hospital volume, age, and CCI score.

CONCLUSIONS

The comparisons showed that, in preoperative and postoperative healthcare consultations with HCC surgery candidates, QOL at 6 months post-surgery should be estimated with an ANN model rather than with SVM, GPR, or MLR models. The best QOL predictors identified in this study can also be used to educate candidates for HCC surgery in the expected course of recovery and other surgical outcomes.