Title of Manuscript: Development and Internal-external Validation of a Comprehensive Model for Predicting Risk of Post-RFA Recurrence in HCC Patients

Overview

Background A predictive model for post radiofrequency ablation (RFA) recurrence in patients with Hepatocellular carcinoma (HCC) that incorporates variables like sleep quality and psychological factors can provide more time to prevent the recurrence. Our aim is to investigate the relationship between these factors and post-RFA recurrence, and to construct a predictive model includes these highly preventable factors. Methods We collected data from HCC patients who underwent RFA for the first time from January 1, 2015, to July 2023, assessing their sleep quality, anxiety, and depression levels. We employed Restricted cubic splines (RCS), mediation analysis, Cox proportional hazards model, Elastic network Cox proportional hazards, Competitive risk model to ascertain the relationship between these factors and post-RFA recurrence. We then constructed a predictive model incorporating these factors, and evaluated the model's performance through internal and external validation datasets partitioning by time period.

Liver cancer is one of the most common malignant tumors globally, ranking sixth in incidence among all types of malignancies and third in mortality1. There were 905,700 new cases and 830,200 deaths from liver cancer worldwide in 20201. HCC accounts for 70%-85% of the pathological types of liver cancer2. The number of HCC cases in China accounts for about half of the global HCC patients2, and the incidence is on the rise. HCC has become the second leading cause of cancer-related deaths in China2. It is evident that HCC imposes a significant social disease burden and has become a major public health issue urgently needing attention in China. In China, the curative treatment methods for liver cancer mainly include surgical resection, liver transplantation and ablation3. However, studies have found that the recurrence rate of HCC within 5 years after treatment is high regardless of the treatment method3, especially for ablation, which is widely used in clinical practice. Studies have found that the recurrence rate after ablation is higher than that of other two curative treatment methods (surgical resection and liver transplantation)3. RFA is one of the most widely used ablation methods. By inserting electrodes into the tumor, RFA generates heat to make the local temperature reach high, killing tumor cells and reducing the damage to surrounding normal liver tissue4. RFA is one of the most important treatment methods for small HCC and advanced HCC patients who cannot be resected surgically. However, there may be thorny problems such as incomplete ablation, insufficient ablation volume, and tumor metastasis along the needle path during RFA, all of which will increase the local recurrence rate after RFA5. Post-RFA recurrence of HCC patients not only reduce the quality of life, but also increases the hospitalization rate and fatality rate, which is the most critical factor hindering the long-term survival of patients after RFA. Therefore, how to prevent post-RFA in HCC patients at an early stage and prolong the survival time of HCC patients is a key link to improve the overall survival rate of HCC patients. Researches have shown that precise prediction models can effectively forecast the occurrence of future events and assist in clinical decision-making and the formulation of health policies6,7. Therefore, identifying the predictive factors for post-RFA recurrence in HCC patients, constructing accurate prediction models for recurrence risk, and effectively identifying high-risk individuals for post-RFA recurrence in HCC patients to implement corresponding recurrence prevention management strategies are of significant importance in prolonging post-RFA survival time for HCC patients. Previous prediction models have primarily focused on predictive factors for post-RFA recurrence risk, such as tumor serum markers, serum albumin, and imaging data8-11. However, abnormalities in these factors often indicate early recurrence, leaving minimal room for prevention. Thus, there is an urgent need to explore predictive factors for post-RFA recurrence in HCC patients that exhibit early preventable characteristics. Research has found that psychological factors such as anxiety and depression, as well as sleep quality, may impact outcomes in patients with cancer12-17. Some researchers have attempted to investigate the relationship between these factors and postoperative recurrence in HCC patients17-21, suggesting that depression may increase the risk of postoperative recurrence18. Since these factors can be addressed through early interventions such as psychological and sleep therapies, confirming their association with post-RFA recurrence in HCC patients and incorporating them as predictive factors in constructing a predictive model for post-RFA recurrence risk could better facilitate early prediction and prevention of post-RFA recurrence in HCC patients. However, there is currently no research confirming the relationship between anxiety, depression, sleep quality, and post-RFA recurrence, nor have these factors been included as predictive factors in constructing models for predicting post-RFA recurrence in HCC patients. The current method for selecting predictive factors primarily involves stepwise regression22, which relies entirely on data-driven. However, this method is susceptible to overfitting and data bias, which may result in inconsistencies in predictive models constructed by different centers and lack of generalizability. In this study, we integrated expert knowledge with the Lasso model to achieve a combination of subjective and objective predictive factor selection, a more precise, reliable, and scientific factor selection is achieved, thereby improving prediction accuracy, enhancing decision-making effectiveness, better exploring and utilizing potential information in the data, and making factor selection more objective, scientific, and systematic.