Traditional herbal medicine as an adjuvant therapy for preventing the recurrence of hepatocellular carcinoma after radical resection: Development and validation of a machine learning prediction model

Abstract

Introduction

Hepatocellular carcinoma (HCC) is the most prevalent pathological subtype of primary liver cancer. Although numerous studies have proposed models to predict recurrence following radical resection of HCC, none have accounted for the potential impact of traditional Chinese medicine (TCM). This study aims to evaluate whether adjuvant therapy with traditional herbal medicine (THM) can extend recurrence-free survival (RFS) in patients with HCC and to develop clinical prognostic models to aid in assessing the risk of recurrence after hepatectomy.

Methods

Clinical data from 403 patients who underwent radical resection at Shanghai Changhai Hospital from 2002 to 2023 were collected. Patients were categorized into THM and non-THM groups on the basis of whether they received THM therapy in the early postoperative period following HCC resection, and survival analysis was conducted to assess the differences in RFS and overall survival (OS) between the two groups. To build the models, the data were split into training and testing sets. Twenty-one variables were selected using Lasso regression, and predictive models were subsequently constructed employing Cox proportional hazards regression and the random survival forest (RSF) method. A nomogram was then developed based on the Cox regression model. The models were validated and compared based on their discriminatory power, calibration performance, and clinical utility. A nomogram was then developed based on the Cox regression model.

Results

RFS was significantly better in the THM group than in the non-THM group (p < 0.001), but there was no difference in OS between the two groups. The median RFS was 48.13 (95 % CI, 38.63, 57.64) months vs 18.07 months (95 % CI, 13.09, 23.04), and the 1-, 3-, and 5-year RFS rates in the THM and non-THM groups were 82.26 % vs 62.58 %, 62.10 % vs 35.55 % and 45.56 % vs 29.03 %, respectively. The variables identified through Lasso regression—including THM intervention, age, postoperative AFP levels, cirrhosis, tumor diameter, and gender—were incorporated into the development of both Cox and RSF models. Both models exhibited comparable performance with respect to discrimination, calibration, and clinical utility.

Conclusion

THM adjuvant therapy following HCC resection can effectively reduce the risk of early recurrence; however, it does not significantly extend overall survival. Cox regression and RSF prediction models were successfully established, and their combined application may aid clinicians in estimating individual recurrence risk and formulating personalized treatment strategies.