The Characteristics and Mortality of Chinese Herbal Medicine Users among Newly Diagnosed Inoperable Huge Hepatocellular Carcinoma (≥10 cm) Patients: A Retrospective Cohort Study with Exploration of Core Herbs

For patients with inoperable huge hepatocellular carcinoma (H-HCC, tumor size ≥10 cm), treatment options are limited. This study aimed to evaluate the characteristics and outcomes of patients with H-HCC who use Chinese herbal medicine (CHM). Multi-institutional cohort data were obtained from the Chang Gung Research Database (CGRD) between 1 January 2002 and 31 December 2018. All patients were followed up for 3 years or until the occurrence of death. Characteristics of CHM users and risk of all-cause mortality were assessed, and core CHMs with potential pharmacologic pathways were explored. Among 1618 patients, clinical features of CHM users (88) and nonusers (1530) were similar except for lower serum α-fetoprotein (AFP) and higher serum albumin levels in CHM users. CHM users had significantly higher 3 year overall survival rates (15.0% vs. 9.7%) and 3 year liver-specific survival rates (13.4% vs. 10.7%), about 3 months longer median survival time, and lower risk of all-cause mortality. Core CHMs were discovered from the prescriptions, including Hedyotis diffusa Willd combined with Scutellaria barbata D.Don, Salvia miltiorrhiza Bunge., Curcuma longa L., Rheum palmatum L., and Astragalus mongholicus Bunge. CHM use appears safe and is possibly beneficial for inoperable H-HCC patients; however, further clinical trials are still required.


Introduction
Liver cancer is the second-most common cause of cancer-related death worldwide and is one of the few neoplasms with a steadily increasing incidence and mortality. Hepatocellular carcinoma (HCC) makes up approximately 90% of all cases of primary liver cancer and is the fourth most common cause of cancer-related death worldwide [1][2][3][4]. The illness is usually associated with chronic liver immune disorders caused by excessive alcohol consumption, hepatitis B virus (HBV) infection, and hepatitis C virus (HCV) infection [5]. The international guidelines for treatment remain controversial for patients with huge HCC marketing, HFs are manufactured and mixed in proportions regulated by classical TCM theory. Accordingly, TCM doctors may prescribe HFs or SHs on the basis of each patient's condition. All CHMs are made by pharmacies following "good manufacturing practices" and have zero-tolerance regulations for potential renal or liver toxicities and pollution by pesticide or heavy metals. Figure 1 shows a flow diagram of this study. First, we retrieved patients newly diagnosed with HCC (ICD-10-CM code: C22.0, ICD-9-CM code: 155) between 1 January 2002 and 31 December 2018 from the CGRD. Next, we excluded patients with errors in diagnosis or death date, patients with missing gender, age, tumor size records, or TNM staging, and patients who received no treatment after diagnosis. We also excluded patients with tumor size <10 cm, those who received surgery for HCC, those who were aged ≥18 and <80 years, and those who died within 80 days of diagnosis. Then, we categorized the remaining eligible subjects into two groups as a function of whether CHM treatment was used after diagnosis. CHM users were defined as the population receiving at least two CHM treatments during the study period, while those who did not receive CHM treatment were classified as CHM nonusers. The intention-to-treat design was applied to define CHM users and nonusers; therefore, patients were not reallocated between groups during follow-up. The entire study protocol was approved by the Institutional Review Board of the Chang Gung Medical Foundation in Taiwan (IRB No.: 201900800B0). Written informed consent was exempted because the identification number of each patient was encrypted, and it was impossible to deduce the identity of the patients.

Outcome Assessment
All eligible subjects were followed up until the occurrence of the primary which was set at a maximum of 3 years after diagnosis, or until the end of 2019

Outcome Assessment
All eligible subjects were followed up until the occurrence of the primary endpoint, which was set at a maximum of 3 years after diagnosis, or until the end of 2019 ( Figure 1). The primary outcome of this study was the overall survival rate (OS), in which death could be caused by any causes, and the survival time was calculated from the date of diagnosis to the date of death. The secondary outcome of this study was the disease-specific survival rate, which was referred to as liver-specific survival rate and was only applied to patients who died due to HCC-related causes.

Study Covariates
Demographic covariates, including gender, age, comorbidities, and lifestyles, were obtained from the CGRD. HCC-related covariates were also retrieved, such as the Child-Pugh score classification, cirrhosis, cancer staging according to both TNM staging and Barcelona Clinic Liver Cancer (BCLC) classification, tumor size, and initial treatment modalities (including TACE, RFA/PEI, target therapy, and chemotherapy). In addition, biochemical profiles were recorded to establish patients' baseline physical condition, including α-fetoprotein, serum albumin, hemoglobin, platelet, international normalized ratio (INR), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and total bilirubin. The laboratory values for these biochemical profiles were obtained within 1 year before diagnosis, and the value closest to the diagnosis date was used. For the regression models, the laboratory values were reclassified to binary covariates as 'normal' and 'abnormal' according to the reference values. In addition, two outpatient visits or one inpatient record of hypertension, type 2 diabetes mellitus (DM), chronic hepatitis (HBV, HCV, or HBV + HCV), or chronic kidney disease (CKD) within 1 year before the diagnosis of HCC were considered comorbidities. The diagnosis codes used in this study are listed in Supplementary Material S1. The use of metformin, aspirin, anti-HBC/HCV therapy, and diuretics more than 30 days was also recorded because they may influence the prognosis of HCC [31][32][33][34].

Bias Assessment
To minimize the confounding bias, we matched CHM users and nonusers through different propensity score (PS)-based models for the sensitivity tests in this study. Furthermore, the use of the CGRD prevents possible recall bias because it is an electronic medical record that is updated simultaneously during daily clinical practice in the CGMH. Since there was no exact time of CHM treatment initiation for HCC patients given in the clinical guideline, immortal time bias may be present; hence, we excluded patients who died within 80 days during the follow-up period, which was determined by calculating the median time between the date of diagnosis and receiving CHM treatment. In addition, all CGRD entries were linked to the national death registry database supported by the National Health Informatics Project, which allowed us to trace patients' outcomes afterward, even if the patient died outside the CGMH. As a result, we could rule out the possible biases in registration and detection of the cause of death in the CGRD.

Statistical Analysis
Outcome assessment and Chinese herbal medicine network (CMN) analysis on prescriptions made for inoperable H-HCC were performed in this study. Baseline demographic features were presented as the median with interquartile range (IQR) for continuous covariates or as counts with percentages for categorical covariates. The differences in demographics between CHM users and nonusers were examined with Student's t-tests and chi-square statistics. Moreover, the characteristics of CHM users were assessed by logistic regression when considering all possible covariates. The OS and liver-specific survival were estimated using the Kaplan-Meier method at 1, 2, and 3 year timepoints. The risk of all-cause mortality was assessed using the Cox regression model. Adjusted hazard ratios (aHR) were calculated by considering the imbalanced demographic covariates between CHM users and nonusers. Furthermore, we used several PS-based models to correct for the imbalanced case numbers and estimated the risk of mortality under different model settings, including overlap weighting, average treatment effect on the treated (ATT), and inverse probability treatment weighting (IPTW). For example, the overlap weighting model added weights to subjects on the basis of demographic features between CHM users and nonusers, including age, gender, lifestyles, comorbidities, initial treatments, and medications [35]. These covariates were used to generate the probability of using CHM as PS, and the PS was assigned as a weight to CHM nonusers while 1 − PS was assigned to CHM users [36]. In addition, covariates other than those used to generate PS were used in the Cox regression to assess the risk of all-cause mortality in different PS-based models. As a different sampling method, all subjects without a landmark setting and those with a 120 day landmark setting were also used to examine the risk of all-cause mortality. We also performed multivariate Cox regression stratified by demographic covariates to confirm the associations between using CHM and the OS in different subgroups of subjects.
Furthermore, with CMN analysis, we graphically demonstrate the treatment principle and the core CHM used for H-HCC. The process of CMN build-up was reported in detail in our previous studies [37]. Association rule mining (ARM) was used to determine the common CHM combinations and social network analysis (SNA) was used to graph and analyze the CMN. We collected commonly used CHM into clusters according to their relationships and identified core CHMs that had high prevalence and were highly connected to other CHMs (i.e., that were used when the core CHM was prescribed). According to TCM theory, the CHM was prescribed on the basis of the individual syndrome, and we identified each CHM cluster and its core CHM by integrating SNA and CHM indications from the CHM pharmacopoeia. We also analyzed possible molecular mechanisms of the core CHM from the CMN. To summarize the molecular mechanisms of CHMs, the target proteins of each cluster were used to propose the possible molecular mechanisms using overrepresentation tests in REACTOME, which is a freely accessible web resource to estimate, interpret, and visualize the molecular mechanisms of a given group of genes or proteins [38]. To analyze the CMN in this study, Stata statistical software (release 16; StataCorp, College Station, TX, USA) and the network analysis software NodeXL were used to determine the core CHM in complex CMNs, with a significance threshold of p < 0.05.

Features of CHM Users
The demographic and clinical data are presented in Table 1   A comparison of the above clinical variables between CHM users (n = 88) and CHM nonusers (n = 1530) is also given in Table 1. There were no significant differences in the variables observed between the two groups (all p > 0.05) except for albumin (p = 0.003); however, the value of albumin was still within the normal range in both groups (4.0 [3.5-4.2] for CHM users and 3.7 [3.2-4.1] for CHM nonusers). When considering all covariates, cigarette smoking (adjusted odds ratio [aOR] = 0.36; 95% CI = 0.14-0.90; p = 0.029) and high α-fetoprotein (> 400 ng/mL) (aOR = 0.41; 95% CI = 0.21-0.81; p = 0.010) were less associated with CHM use, while the occurrence of cirrhosis was positively associated with CHM use (aOR = 2.50; 95% CI = 1.14-5.47; p = 0.022) ( Table 2).  Table 3 presents the outcome of inoperable H-HCC among all subjects. The current study found significant differences between CHM users and CHM nonusers at any of the 1, 2, or 3 year timepoints for OS and liver-specific survival rate (p < 0.05). CHM use was associated with higher OS at each timepoint. The favorable tendency for liver-specific survival could be seen when excluding subjects unrelated to H-HCC at each time point (Table 3). For CHM users, the risk of all-cause mortality was reduced by 31% compared with CHM nonusers (HR = 0.69; 95% CI = 0.56-0.85; p = 0.001, Table 4). When considering all accessible covariates, CHM users even had a 38% lower risk of all-cause mortality (aHR = 0.62; 95% CI = 0.44-0.87; p = 0.006). Further, CHM use had a duration-dependent tendency to reduce all-cause mortality. Table 4 demonstrates that the risk of all-cause mortality was significantly reduced among CHM users with accumulative duration >28 days (HR = 0.54; 95% CI = 0.40-0.74; p < 0.001; and aHR = 0.44; 95% CI = 0.26-0.74; p = 0.002; Table 4), but no significant difference was found between CHM users with duration <28 days (HR = 0.84; 95% CI = 0.63-1.10; p = 0.204, and aHR = 0.85; 95% CI = 0.58-1.25; p = 0.416). Figure 2A shows the Kaplan-Meier estimation of OS and liver-specific survival of all patients with inoperable H-HCC. At the 3 year follow-up, the OS was 9.7% (95% CI = 8.3-11.4) for CHM nonusers and 15.0% (95% CI = 8.2-23.7) for CHM users. The median survival for CHM nonusers and CHM users was 200 days and 303 days, respectively. The outcomes of CHM users were significantly superior to those of CHM nonusers (log-rank test, p = 0.002). The same tendency was found with the liver-specific survival ( Figure 2B). Sensitivity tests with different PS-based matching methods and different populations were performed to remove potential confounding and selection biases ( Table 5). The ATT, IPTW, and overlap weighting tests all presented consistent and significantly favorable HR. Significance differences between populations were observed on all subjects with a 120 day landmark design (HR = 0.70; 95% CI = 0.54-0.91; p =0.008, and aHR = 0.67; 95% CI 0.49-0.93; p = 0.016) and without landmark design (HR = 0.55; 95% CI = 0.44-0.69; p < 0.001; and aHR = 0.58; 95% CI = 0.44-0.76; p < 0.001). As for the subgroup analysis, patients with initial treatment (HR = 0.73; 95% CI = 0.54-0.98; p = 0.037, and aHR = 0.59; 95% CI = 0.40-0.88; p = 0.009) were revealed to have better HR compared to CHM nonusers. Additionally, higher disease severity, including low albumin, coexistence of cirrhosis, and high α-fetoprotein, led to a lower risk of all-cause mortality.

Prescription Analysis on CHM Used for H-HCC
A total of 417 prescriptions were made from 293 CHMs, and 7.7 ± 4.4 (mean ± standard deviation) kinds of CHMs were used in each prescription. Table 6 lists the top 10 single CHMs. Hedyotis diffusa Willd. (43.2%) was the most used CHM for H-HCC, followed by Salvia miltiorrhiza Bunge. (31.7%). We also determined the 100 most prevalent CHM-CHM combinations used to construct CMN (Supplementary File S2). Figure 3 demonstrates the CHM network by clustering these combinations. This network presents a comprehensive overview of CHM for H-HCC, in which larger circles denote a higher prevalence of single CHM, and thicker and darker edges between CHMs represent more prevalent and stronger connections. By integrating SNA and CHM indications from the CHM pharmacopeia, five CHM clusters were identified, each with its own tendency to treat a specific TCM syndrome. The clusters are distinguished by different colors in Figure 3. The core CHMs among the five clusters could be identified from their relatively high prevalence and high connectivity to other CHM within clusters, which were calculated using SNA on the CMN. The composition of all herbal formulas (HFs) and the list of CHM and ingredients categorized by different CHM clusters in the CHM network are provided in Supplementary Materials S3 and S4. Cluster 1 reinforces healthy qi and eliminates the pathogenic factors, and has Hedyotis diffusa Willd and Scutellaria barbata D.Don as its core CHMs. HFs including Gui-Lu-Er-Xian Jiao, Xiang-Sha-Liu-Jun-Zi-Tang, and Zhen-Ren-Huo-Ming-Yin shared strong connections with the core CHMs in Cluster 1. Cluster 2 activates blood and resolves stasis, and has Salvia miltiorrhiza Bunge as its core CHM. Cluster 3 soothes the liver and harmonizes the stomach, and has Curcuma longa L as its core CHM. Cluster 4 focuses on bile-draining, as well as anti-icteric management, and has Rheum palmatum L. as its core CHM. Lastly, cluster 5 supplements qi, and has Astragalus mongholicus Bunge as its core CHM.
Res. Public Health 2022, 18, x FOR PEER REVIEW pathogenic factors, and has Hedyotis diffusa Willd and Scutellaria barbata D.Don as CHMs. HFs including Gui-Lu-Er-Xian Jiao, Xiang-Sha-Liu-Jun-Zi-Tang, and Zhe Huo-Ming-Yin shared strong connections with the core CHMs in Cluster 1. Clust tivates blood and resolves stasis, and has Salvia miltiorrhiza Bunge as its core CHM. 3 soothes the liver and harmonizes the stomach, and has Curcuma longa L as its core Cluster 4 focuses on bile-draining, as well as anti-icteric management, and has Rhe matum L as its core CHM. Lastly, cluster 5 supplements qi, and has Astragalus mong Bunge as its core CHM.

Proposed Molecular Pathways of Core CHMs for H-HCC
Furthermore, we investigated possible actions of core CHMs by consulting the RE-ACTOME pathway database to explore the mixed pharmacological effects of core CHMs. A total of 37 kinds of CHM (with 124 ingredients) were used to find their target proteins. The list of total possible binding proteins of each CHM cluster is provided in Supplementary Materials S3. Figure 4 shows the molecular pathways covered by core CHMs in the CMN. These pathways included the cell cycle, DNA repair and replication, the immune system, metabolism of lipids and proteins, and signal transduction. Overall, we found 109 molecular pathways covered by five core CHMs in the network. Among the five CHM clusters, only Cluster 2 affected binding proteins across all five different pathways, and all five core CHMs affected pathways in the metabolism of lipids and proteins.

Discussion
Our study demonstrates the potential of using CHM for patients with inoperable H-HCC, with CHM users having better 1 and 2 year OS and liver-specific survival rates (Table 3), with a duration-dependent tendency (Table 4). Additionally, CHM use may signif-

Discussion
Our study demonstrates the potential of using CHM for patients with inoperable H-HCC, with CHM users having better 1 and 2 year OS and liver-specific survival rates (Table 3), with a duration-dependent tendency (Table 4). Additionally, CHM use may significantly reduce all-cause mortality with or without considering all accessible covariates in the regression model (Table 4). We also performed sensitivity tests with different PS-based matching methods and other sampling populations (with 120 day landmark analysis, age, gender, and initial treatments) to reduce potential confounding and selection biases (Table 5). Since CHM is not yet a part of the standard management of HCC, the consistent results of sensitivity tests show the feasibility of using CHM among patients with inoperable H-HCC. Moreover, five core CHMs with different CHM indications and possible pharmacologic pathways were explored.
According to the BCLC staging system, liver resection is the best treatment for patients with BCLC stage 0 and stage A diseases; however, treatment options are often limited for inoperable patients, especially when the tumor burden is heavy. For these patients, individualized treatments are often demanded, and we propose that CHM may be used at this stage. Patients with asymptomatic large HCCs (>5 cm) without major vascular invasion or extrahepatic spread are categorized as having intermediate stage disease (BCLC stage B) and should receive TACE [39,40]. According to Lee et al., surgery is superior to TACE for patients with H-HCC (≥10 cm) in terms of disease control and survival, and should be performed for single H-HCC or H-HCC with limited daughter nodules confined in the same lobule, for which the median survival time may reach nearly 52.6 months in surgery group [9]. Unfortunately, the prognosis was much worse for patients with inoperable H-HCC. For inoperable large HCC (≥5 cm, media tumor size 8.6 [7-12.3] cm), Yoon et al. reported that the median survival time dropped to 15.8 months with modified cisplatinbased TACE [41]. According to the Sorafenib Hepatocellular Carcinoma Assessment Randomized Protocol (SHARP) trial [42][43][44][45], target therapy with sorafenib extends the median survival of patients with advanced HCC to 10.7 months. However, the study by Cheng et al. [46] reported that the median survival of patients with advanced HCC dropped to 6.5 months when treated with sorafenib, which could be due to the large degree of variation in disease severity in the study population. Tang et al. reported that combined CHM use with TACE and RFA for inoperable patients (main mean tumor sizes of 5.6 cm and 6.8 cm, respectively) may significantly improve 3 year overall survival (37.74% and 38.3% respectively) [21,22]. Our subgroup analysis also supported that, even with a much larger primary tumor size (mean = 12.7 cm), CHM users with initial treatments including TACE, RFA/PEI, chemotherapy, and target therapy may have much improved overall survival (Table 5). For other prognostic covariates, such as hypoalbuminemia, coexistence of cirrhosis, and high α-fetoprotein, use of CHM tended to be associated with lower risk of 3 year all-cause mortality. As a result, we believe that combined CHM use may be a possible option for better overall survival for patients with inoperable H-HCC.
Previous studies showed that HCC patients with tumor size larger than 10 cm were generally younger and had less HCV infection when compared to HCC patients with smaller tumor sizes [9,47], which is consistent with our current study. Several studies noted that, compared with older HCC patients, younger HCC patients have lower positive rates of hepatitis C virus, lower proportion of cirrhosis, and higher frequency of increased AFP levels [48,49], suggesting possible differences in the carcinogenesis of HCC between young and elderly patients. Although the hepatitis B virus may be important in HCC development without associated liver cirrhosis, nonviral causes such as nonalcoholic fatty liver disease accounted for nearly 40% of H-HCC patients in the current study, which implies the importance of metabolic disorder in HCC. Likewise, REACTOME pathway analysis revealed that all five core CHMs affected pathways in the metabolism of lipids and proteins, which may also suggest possible mechanisms for the carcinogenesis of H-HCC, as well as the value of targeting these mechanisms to manage H-HCC. Further studies are warranted to confirm these associations.
According to TCM theory, healthy qi not only promotes the metabolic function of each organ, but also defends the human body from pathogens. Solid cancers such as H-HCC may be viewed as a result of an internal pathogen caused by qi and blood stasis, such as dysregulation of the metabolic and immune systems [50,51]. This explains the core CHM of Cluster 2 (Salvia miltiorrhiza Bunge) and how activating blood and resolving stasis may a have beneficial effect on fatty liver, nonalcoholic fatty liver disease (NAFLD), and its progressive form, nonalcoholic steatohepatitis (NASH) [52]. The active constituent of Salvia miltiorrhiza Bunge, Tanshinone II-A (TIIA), was reported to have many beneficial effects on HCC, including arrest at G0/G1 phase, inducing apoptosis, as well as antiinvasion and anti-metastasis effects [53]. Tanshinone II-A can also protect against the hepatic lipid peroxidation process, and it may regulate intracellular molecular targets like PPARα, CYP1A2, and MMP2, to mediate lipid metabolism and promote antioxidant activity and anti-fibrogenesis [54,55]. Curcuma longa L (turmeric) is the only core CHM that affects only lipid and protein metabolism pathways, and its liver-protective and antioxidant features have been widely reported [56]. Curcumin is the main constituent reported to be effective against numerous types of oxidative-associated liver diseases [57][58][59]. Curcumin and curcumin-rich Curcuma longa L. extract have been reported to promote recovery in a representative model of carbon tetrachloride (CCl 4 )-induced hepatotoxicity for both acute and chronic hepatotoxicity and lipid dyslipidemia [60].
Pharmacological pathway analysis revealed that, in addition to affecting the metabolic pathways of lipids and proteins, the core CHMs of Cluster 1 (Hedyotis diffusa Willd and Scutellaria barbata D.Don) and Cluster 4 (Rheum palmatum L.) may also mediate interleukin-4related immune pathways and interleukin-13 signaling. According to the immunotherapy trials of HCC in recent years [61], the checkpoint inhibitors of programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) engage CD8 immunity [62]. However, the core CHMs have the opposite effect and mediate CD4 cytokine signaling. Studies following the dynamics of CD8 T-cells revealed that changes in the systemic PD-1(+) CD8 subpopulation in response to PD-L1/PD-1 blockade therapy might lead to significant antitumor activities in patients [63,64], but later studies suggested that these changes may depend on systemic CD4 immunity [65,66]. A very recent review from Zuazo et al. [67] showed that the presence of specific CD4 T-cell memory subsets in peripheral blood before the initiation of treatments is a strong predictor of responses in non-small-cell lung cancer patients. Furthermore, the polysaccharides extracted from the core CHM of cluster 1, Hedyotis diffusa Willd, were reported to enhance the antitumor activity of cytokine-induced killer cells and could be used for cancer immunotherapy when combined with cytokine-induced killer cell therapy [68]. The extract of another core CHM of cluster 1, Scutellaria barbata D. Don, was reported to have anticancer effects, mediated by its immunomodulatory activity in hepatoma H22-bearing mice by downregulating Treg cells and manipulating the Th1/Th17 immune response [69]. These results not only pave the way for further studies on the pharmacological mechanism anticancer CHMs, but also provide clinical guidance for TCM practitioners to treat patients undergoing immunotherapy.
Nevertheless, our study had some limitations. First, the CGRD covers only CHM prescribed through CGMH. The HCC patients who received CHM from local clinics or other medical facilities could not be identified; hence, the use of CHM may be underestimated. Additionally, CHM is not a standard treatment for HCC, and there is no suggested timing to initiate CHM management. In clinical practice, patients autonomously ask for and consensually receive CHM management from TCM doctors; hence, accidental exposure among CHM nonusers should be minimal, and the use of landmark analysis could reduce the possibility of immortal time bias. CHM prescriptions and compositions are complicated because these prescriptions are personalized and are composed of various kinds of CHMs that are concocted on the basis of each patient's condition. Therefore, we assessed the duration-dependent trend through exposure duration instead of a single CHM dose dependency. Furthermore, we found that the prevalence of CHM use for H-HCC was lower than CHM use among all HCC patients in Liao et al.'s report [23]. The insufficient evidence in using CHM among H-HCC patients and the inclusion of newly diagnosed H-HCC only may both contribute to the low prevalence of CHM use. Although strict inclusion criteria could help us to estimate the outcome for H-HCC more accurately, the low proportion of CHM use may lead to the issue of poor generalizability of this result. Overall, our study is still a retrospective observational study and is limited by the database used; thus, further large randomized controlled trials with specific SH or HF are still needed to confirm the actual causality of different CHM prescriptions.

Conclusions
This study indicates the feasibility of CHM use for patients with inoperable H-HCC. In addition, the evaluation of core CHMs and proposed molecular pathways disclosed the role of CHM in managing H-HCC. These results may warrant further clinical and bench studies for CHM use among patients with inoperable H-HCC.