Development and Validation of a Machine Learning-Based Model Used for Predicting Hepatocellular Carcinoma Risk in Patients with Hepatitis B-Related Cirrhosis: A Retrospective Study

Yixin Hou; Jianguo Yan; Ke Shi; Xiaoli Liu; Fangyuan Gao; Tong Wu; Peipei Meng; Min Zhang; Yuyong Jiang; Xianbo Wang

doi:10.2147/OTT.S444536

Development and Validation of a Machine Learning-Based Model Used for Predicting Hepatocellular Carcinoma Risk in Patients with Hepatitis B-Related Cirrhosis: A Retrospective Study

Onco Targets Ther. 2024 Mar 23:17:215-226. doi: 10.2147/OTT.S444536. eCollection 2024.

Authors

Yixin Hou^#¹, Jianguo Yan^#², Ke Shi^#^{1

3}, Xiaoli Liu¹, Fangyuan Gao¹, Tong Wu¹, Peipei Meng¹, Min Zhang², Yuyong Jiang¹, Xianbo Wang¹

Affiliations

¹ Center of Integrative Medicine, Beijing Ditan Hospital, Capital Medical University, Beijing, 100015, People's Republic of China.
² People's Liberation Army Fifth Medical Center, Beijing, 100039, People's Republic of China.
³ Department of Gastroenterology, Dongzhimen Hospital, Beijing University of Chinese Medicine, Beijing, 100700, People's Republic of China.

^# Contributed equally.

Abstract

Object: Our objective was to estimate the 5-year cumulative risk of HCC in patients with HBC by utilizing an artificial neural network (ANN).

Methods: We conducted this study with 1589 patients hospitalized at Beijing Ditan Hospital of Capital Medical University and People's Liberation Army Fifth Medical Center. The training cohort consisted of 913 subjects from Beijing Ditan Hospital of Capital Medical University, while the validation cohort comprised 676 subjects from People's Liberation Army Fifth Medical Center. Through univariate analysis, we identified factors that independently influenced the occurrence of HCC, which were then used to develop the ANN model. To evaluate the ANN model, we assessed its predictive accuracy, discriminative ability, and clinical net benefit using metrics such as the area under the receiver operating characteristic curve (AUC), concordance index (C-index), calibration curves.

Results: In total, we included nine independent risk factors in the development of the ANN model. Remarkably, the AUC of the ANN model was 0.880, significantly outperforming the AUC values of other existing models including mPAGE-B (0.719) (95% CI 0.670-0.768), PAGE-B (0. 710) (95% CI 0.660-0.759), FIB-4 (0.693) (95% CI 0.640-0.745), and Toronto hepatoma risk index (THRI) (0.705) (95% CI 0.654-0.756) (p<0.001 for all). The ANN model effectively stratified patients into low, medium, and high-risk groups based on their 5-year In the training cohort, the positive predictive value (PPV) for low-risk patients was 26.2% (95% CI 25.0-27.4), and the negative predictive value (NPV) was 98.7% (95% CI 95.2-99.7). For high-risk patients, the PPV was 54.7% (95% CI 48.6-60.7), and the NPV was 91.6% (95% CI 89.4-93.4). These findings were validated in the independent validation cohort.

Conclusion: The ANNs model has good individualized prediction performance and may be helpful to evaluate the probability of the 5-year risk of HCC in patients with HBC.

Keywords: hepatitis B-related cirrhosis; hepatocellular carcinoma; machine learning-based model; risk.

Grants and funding

This work was supported by Beijing Hospitals Authority Youth Programme (QMl220201802), the Beijing Traditional Chinese medicine science and Technology Development Fund Project (No. Qn-2020-25), Application of Clinical Features of Capital City of Science and Technology commission (z181100001718052),High-level public health technical personnel construction project (The backbone of the discipline-03-07), National Key R&D Program of China(20232023YFC230880).