Radiation-induced hepatic toxicity from radiotherapy plus immune checkpoint inhibitors with targeted therapy in HCC patients: a propensity score-matched study.

Introduction
Programmed cell death receptor ligand 1/Programmed cell death protein 1 (PD-L1/PD1) combined with targeted drugs is an effective treatment method for patients with unresectable hepatocellular carcinoma (HCC). Statistically, about 50% of HCC patients in the course of the disease are treated with systemic therapies [1]. Atezolizumab plus bevacizumab has recently been included as a first-line HCC therapeutic option in the American Association for the Study of Liver Diseases (AASLD) HCC guidance [2]. Although atezolizumab plus bevacizumab has been associated with improved median overall survival (mOS) in the IMbrave150, its objective response rate (ORR) remains relatively low [3].
Radiotherapy (RT), a localized treatment, is safe and effective for HCC patients [4, 5]. Nonetheless, the therapeutic effect of RT alone is limited in HCC patients, and the addition of other therapeutic approaches that can improve the effect on HCC has become a new research direction. RT can favorably modulate immunological pathways, prime tumor-specific cytotoxic T cells, and enhance T cell homing, engraftment, and function in tumors [6]. A phase II trial of SBRT combined with camrelizumab revealed an ORR of 52.4% and an mOS of 14.2 months in unresectable HCC patients [7]. Moreover, antiangiogenic therapy enhances T lymphocyte infiltration [8]. Ning et al. [9] showed that patients receiving RT with combination therapy of immune checkpoint inhibitors (ICIs) and anti-angiogenic agents achieved better disease control rate (100% vs. 75.9%, p = 0.005) and ORR (75.9% vs. 24.1%, p < 0.001) than those of patients receiving ICIs plus antiangiogenic agents after propensity score matching (PSM).
Although favorable results have been reported, radiation-induced hepatic toxicity (RIHT) following RT remains a problem and radiation-induced liver disease (RILD), a severe RIHT, is a major problem of RT. Additionally, multiple trials have reported the development of hepatotoxicity in patients receiving ICIs and targeted therapy for unresectable HCC [10]. The ORIENT32 study demonstrated that grade 3 or 4 treatment-related adverse events in 7 (2.0%) patients with increased aspartate aminotransferase, 5 (1.0%) patients with increased alanine aminotransferase, and 19 (5.0%) patients with increased blood bilirubin, even one patient died of hepatic failure in the sintilimab-bevacizumab biosimilar group [11]. Our previous study found that the risk of RIHT may not be increased in HCC patients treated with RT plus anti-PD1 compared to that associated with RT alone in HCC patients [12]. However, evidence of the RIHT effect of RT and ICIs combined with targeted therapy in HCC is lacking. Therefore, the purpose of this study was to evaluate the RIHT associated with RT combined with ICIs and targeted therapy.

Materials and methods

Study design and patients
Data were collected from January 2017 to May 2023 for 146 HCC patients: 50 treated with a combination of RT with ICIs and targeted therapy (triple therapy), and 96 treated with RT alone or RT plus ICIs (RT/RT+ICIs) (Fig. 1). Eligible patients had been pathologically or clinically diagnosed with HCC [2]. The patients had Child–Pugh (CP) class A or B and Barcelona Clinic Liver Cancer (BCLC) stage A, B, or C classification, with an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1. Included were patients who had not undergone prior ablation, surgery, or interventional therapy within three months after RT, and had not undergone ablation or surgery therapy one month preceding of RT. All hepatotoxicity conditions returned to grade 1 or lower in patients who had undergone interventional therapy one month preceding the first fraction. Patients with intrahepatic cholangiocellular carcinoma, incomplete RT, or the loss of RIHT-relevant data were excluded. The institutional medical Ethics committee of the Guangxi Medical University Cancer Hospital approved the study (LW2023186).

Treatment plan

Intensity-modulated radiation therapy (IMRT)
Each patient accepted a serial 2.5–5 mm section plain or contrast-enhanced computed tomography (CT) (slice thickness, 2.5–5 mm) for RT planning, and a vacuum mold was used for immobilization. CT and Magnetic resonance were then combined to clearly visualize the intrahepatic lesions. All target areas and organs at risk (OARs) were mapped in the MIM 6.8 system (MIM, USA). The planning target volume (PTV) was the gross tumor volume (GTV) with an additional margin of 5–10 mm. The IMRT plans were designed using the Monaco treatment planning systems (version 5.1) and Pinnacle 3 (Philips, Netherlands). Hypofractionated IMRT was performed daily in five fractions per week with 6 MV X-ray implemented (ELEKTA Versa-HD). In our patient cohort, a median total dose of 51.0 (48.5–60.0) Gy was given in 17 (15–20) fractions. The OARs dose constraints were determined according to the following principles: a V15 of less than one-third of the volume for the kidneys, maximum dose (Dmax) limits of 40–45 Gy for the stomach, duodenum, and small bowel, and a Dmax of under 40 Gy for the spinal cord [13].

ICIs and targeted therapy
The standard protocols for ICIs and targeted drugs were determined empirically by the treating physicians responsible for each patient, and the method of duration, dose, and injection were used on the basis of the manufacturer’s instructions for the different drugs until intolerable toxic reactions or disease progression occurred. ICIs targeting PD-1 and PD-L1 include camrelizumab, toripalimab, sintilimab, tislelizumab, atirizumab, and cardunolizumab. Molecular targeted drugs include multikinase inhibitors and vascular endothelial growth factors, such as lenvatinib, apatinib, regorafenib, sorafenib, erlotinib, and bevacizumab. In the RT/RT+ICIs group treated with RT or RT combined ICIs, 10, 17, and 3 patients were treated with ICIs before RT, during RT, and within 3 months after RT, respectively. In the triple therapy group treated with RT combined ICIs and targeted therapy, 33, 15, and 2 patients were treated with ICIs and targeted therapy before RT, during RT, and within 3 months after RT, respectively. The median duration of targeted therapy was 4 months, and the median number of ICI administered was 5 cycles.

Hepatic toxicity assessment
The laboratory within one week and imaging tests within one month before RT were used as the baseline to compare CP scores. RIHT was evaluated according to the common toxicity criteria for adverse events (version 5.0) and the highest CP scores were subtracted from the baseline scores to obtain the increase in CP scores within three months after completion of radiotherapy. Classical RILD (cRILD) was defined as a serum alkaline phosphatase (ALP) level ≥ 2-fold the upper limit of normal (ULN) with anicteric hepatomegaly and ascites within three months of RT. Meanwhile, non-classical RILD (ncRILD) was defined as a serum aspartate aminotransferase (AST)or alanine aminotransferase (ALT) level ≥ 5-fold the ULN or a increased CP score of ≥ 2 occurring within three months of RT. Tumor progression and hepatitis B virus (HBV) reactivation were excluded from RIHT. A 10-fold or greater increase in HBV DNA levels from baseline levels is defined as HBV reactivation, and disease progression as evidenced by progressive lesions on imaging or elevated tumor markers based on the Response Evaluation Criteria in Solid Tumours, version 1.1 [14–16]. The tumor response and hepatic toxicity of the patients were assessed based on laboratory and imaging tests one month after the last fraction and every 2–3 months thereafter.

Statistical methods
Student’s t-test (Wilcoxon test) and Pearson’s chi-squared test (Fisher’s exact test) were used to analyze continuous and categorical variables, respectively, to compare the clinical, dosimetric, and hepatotoxic parameters between the RT/RT+ICIs and triple therapy groups. To reduce confounding effects and potential selection bias between the groups, we applied 1:1 PSM using the nearest-neighbor matching algorithm. Matching variables included age, sex, HBV infection, CP grade, alpha-fetoprotein level, extrahepatic metastases, interventional therapy, hepatectomy, ablation, mean dose to the normal liver (Dmean), and equivalent dose in 2‑Gy fractions using LQ model α/β = 2 Gy (EQD22).
Cumulative ncRILD was estimated using the Kaplan-Meier curve. To analyze factors associated with ncRILD, univariate and multivariate logistic regression analyses were used to determine the identified significant variables. The univariate variables included as follows: clinical parameters included age, sex, bodyweight, HBV, hepatitis C virus infection, cirrhosis, ECOG PS, total bilirubin, albumin, AST, ALT, ALP, prothrombin time (PT), CP grade, albumin-bilirubin (ALBI) scores, alpha-fetoprotein (AFP), macrovascular invasion, BCLC stage, max tumor size, tumor number, extrahepatic metastases, interventional therapy, hepatectomy, and ablation; dosimetric parameters included GTV, normal liver volume (Vliver), Dmean, EQD22, and the percentage (%) of normal liver volume receiving more than x Gy (Vx, x = 5, 10, 15, 18, 20, 25, 30, or 35). For additional analyses, the ROC analysis was applied to obtain cut-off points for Vliver. Statistical significance was set at p < 0.05. SPSS® version 25.0 software (SPSS, Inc., Chicago, IL, USA) and R version 4.0.5 (http://www.r-project.org/) were used for all statistical analyses.

Results

Patients
Our previous study showed that clinical and dosimetric data were balanced between patients receiving RT and RT+ICIs after PSM, resulting in similar RIHT [12]. This study finally enrolled 146 HCC patients; 50 and 96 had received treatment in the triple therapy and RT/RT+ICIs groups, respectively. The median follow-up time was 12.5 months (7.4–22.0months) after RT; 67 (45.9%) patients who were included in the study died during the follow-up time. Before PSM, the groups differed in terms of HBV, hepatectomy, age, Dmean, V25, V30, and V35 (p < 0.05) at baseline (Table 1). After PSM, 50 patients per group were included in the matched cohort. There were no significant differences in baseline characteristics between groups of patients (Table 1).

Evaluation and incidence of RIHT before and after PSM
Two, and five patients with hepatic toxicity were excluded for tumor progression, and viral reactivation, respectively. Before PSM, the incidence of ncRILD, increase CP scores of ≥ 2, increase CP scores of ≥ 1, increased ALT grade 3, increased AST grade 3, and increased ALP grade 2 for 146 patients were 17.1%, 15.1%, 39.0%, 2.7%, 4.1%, and 0.7% within 3 months after the last RT fraction, respectively. The cumulative incidence of ncRILD in the triple therapy and RT/RT+ICIs groups before PSM is shown in Fig. 2a. Hepatotoxicity including ncRILD, increased CP scores ≥ 2, increased CP scores ≥ 1, increased ALT grade 3, increased AST grade 3, increased total bilirubin grade 3, increased ALP grade 2, increased ALT ≥ grade 2, increased AST ≥ grade 2, increased ALP ≥ grade 1, increased total bilirubin ≥ grade 2, decreased albumin grade 2, increased ALT ≥ grade 1, increased AST ≥ grade 1, increased total bilirubin ≥ grade 1, and decreased albumin ≥ grade 1 in the triple therapy group did not differ from that in the RT/RT+ICIs group (Table 2, all p > 0.05). After PSM, the ncRILD, increased CP scores of ≥ 2, increased CP scores of ≥ 1, increased ALT grade 3, increased AST grade 3, and increased ALP grade 2 for 100 patients occurred in 17.0%, 16.0%, 39.0%, 3.0%, 3.0%, and 1.0% within 3 months after the last RT fraction, respectively. The cumulative incidence of ncRILD in both groups after PSM is shown in Fig. 2b. The hepatotoxicity parameters were not significantly different between the two groups in the matched cohort (Table 3, p > 0.05).

NcRILD in different subgroups after matching
In the present study, cRILD was not observed. Thus, our study analyzed the differences in ncRILD incidence between the triple therapy and RT/RT+ICIs groups using a subgroup analysis. In the unmatched cohort, for the subgroups of patients with CP-A (17.1% vs. 18.4%, p = 0.887), CP-B (13.3% vs. 16.7%, p = 1.000), HBV infection (17.4% vs. 17.8%, p = 0.961), cirrhosis (18.8% vs. 17.9%, p = 0.929), AFP ≥ 400 ng/mL (26.7% vs. 8.3%, p = 0.471), BCLC C stage (18.2% vs. 19.1%, p = 0.918), macrovascular invasion (19.4% vs. 20.0%, p = 0.953), maximum tumor size ≥ 5 cm (17.5% vs. 20.5%, p = 0.733), tumor number ≥ 4 (17.9% vs. 32.0%, p = 0.232), extrahepatic metastases (14.3% vs. 23.1%, p = 0.927), prior interventional therapy (18.0% vs. 15.4%, p = 0.761), and prior hepatectomy (22.2% vs. 8.3%, p = 0.787), the triple therapy group did not differ significantly in ncRILD from the RT/RT+ICIs group.

Factors associated with NcRILD after matching
The prognostic factors evaluated by univariate analysis consisted of all clinical and dosimetric parameters in the matched cohort, as shown in Table 4. Triple therapy was not significantly correlated with ncRILD in patients with HCC. In the univariate analyses, tumor number (p = 0.041) and Vliver (p = 0.010) could significantly predict ncRILD. Multivariate analysis showed that the Vliver was the only independent predictor affecting ncRILD (p = 0.013). Thus, we further performed ROC analysis by Vliver to obtain the optimal cutoff value. In ROC analyses of the matched cohort, the optimal cutoff of Vliver for predicting ncRILD was 952.7 cc (AUC = 0.708, Supplemental Fig. 1a). The cut-off points of Vliver can effectively stratify the patients into low- and high-risk groups of developing ncRILD (Supplemental Fig. 1b).

Discussion
In recent years, RT combined with ICIs and targeted therapy for HCC patients has achieved favorable outcomes [17, 18]. A randomized controlled trial in HCC patients with PVTT suggested that the combination of camrelizumab-apatinib with SBRT showed clinical benefits, with 47.5% of patients achieving an objective response [18]. RIHT, which is a complication of RT for HCC, primarily affects patient survival. Several studies have shown that ICIs combined with targeted therapy predispose patients to hepatotoxicity and even liver failure [3, 10, 11]. However, few studies have explored the effect of RT combined with ICIs and targeted therapy on RIHT in HCC patients. In the present study, we found that patients who received RT combined with ICIs and targeted therapy had a comparable incidence of RIHT as those who received RT alone or with RT plus ICIs before and after PSM. Additionally, our results showed that Vliver ≥ 952.7 cc for ncRILD may provide some clues for risk stratification.
Our previous study confirmed that the addition of anti-PD1 did not significantly increase RIHT after RT with similar baseline characteristics [12]. In addition, Liu CM et al. [19] conducted a PSM study and reported no excessive risk of RILD (p = 0.537) when patients with advanced HCC receive RT combined with sorafenib compared with the RT-alone. Therefore, our study explored the RIHT of RT combined with ICIs and antiangiogenic therapy compared with RT alone or RT combined with ICIs in HCC patients. In the RT/RT+ICIs group, ncRILD, an increase in CP scores ≥ 2, CP scores ≥ 1, ALT grade 3, and AST grade 3 was observed in 12%, 12%, 34%, 2%, and 4%, respectively. Similarly, 16%, 16%, 38%, 2%, 2%, and 2% of patients who received triple therapy developed ncRILD, increased CP scores ≥ 2, increased CP scores ≥ 1, increased ALT grade 3, increased AST grade 3, and increased total bilirubin grade 3, respectively, and there was no other grade 3–5 hepatotoxicity. No difference in hepatotoxicity between the triple therapy and RT/RT+ICIs group before and after PSM. These findings demonstrate that HCC patients may not have an increased risk of RIHT with RT plus ICIs and targeted therapy compared to the risk associated with RT alone or with RT plus ICIs.
Our findings revealed that the prevalence of liver toxicity in HCC patients treated with RT in combination with ICIs and targeted therapy was similar to that reported in previous studies [7, 20, 21]. A study of hypofractionated three-dimensional conformal radiation therapy (3DCRT) reported a RILD incidence of 15.6% (17/109) in patients with primary liver carcinomas [20]. Another study in which patients received RT reported hepatic toxicity in HCC patients. The study showed that 10.3% of patients had RILD within 3 months after the completion of RT; 4.1%, 6.1%, 3.4%, 6.8%, 2.1%, 5.5%, 0.7%, and 4.8% of patients had G2 elevation in ALT, G2 elevation in AST, G2 elevation in ALP, G2 elevation in total bilirubin, G3 elevation in ALT, G3 elevation in AST, G3 elevation in ALP, and G3 elevation in total bilirubin within 6 months after RT, respectively [21]. In a phase 2 clinical trial, Li et al. suggested that 21 HCC patients who received SBRT plus camrelizumab experienced manageable toxicities, with no other grade 3–5 hepatotoxicity. A total of 9.5%, 4.8%, 4.8%, 9.5%, and 19.1% of patients developed increased ALT in grade 2, increased AST in Grade 2, increased ALP in grade 2, increased blood bilirubin in grade 2, and decreased albumin in grade 2, respectively [7]. Zhong et al. [22] evaluated advanced HCC patients treated with PD-1/PD-L1 Inhibitors combined with RT and anti-angiogenic therapy and found that 12.5%, 12.5%, 6.3%, 6.3%, and 6.3% of patients with grade 2 ALT, grade 2 AST, grade 2 blood bilirubin, grade 3 AST, and grade 3 ALT increases, respectively. These results suggest that the hepatotoxicity can be managed. Consequently, RT plus ICIs and targeted therapy may be feasible for HCC patients with acceptable hepatotoxicity and safety.
Furthermore, RILD has been reported to be a serious toxicity in HCC patients undergoing RT [23, 24]. In our study, subgroup analysis revealed that no difference in the rates of ncRILD was observed for patients with CP-A and CP-B between the triple therapy and RT/RT+ICIs groups (17.1% vs. 18.4%, p = 0.887; 13.3% vs. 16.7%, p = 1.000). Moreover, the rates of ncRILD were not significantly different based on HBV infection, cirrhosis, AFP ≥ 400 ng/mL, BCLC C stage, macrovascular invasion, max tumor size ≥ 5 cm, tumor number ≥ 4, extrahepatic metastases, prior interventional therapy, prior hepatectomy, and prior ablation. Therefore, RT combined with ICIs and targeted therapy in HCC patients may be particularly safe and acceptable.
The prediction of RILD remains challenging, and no criteria have been established for predicting RILD. Kim et al. administered 3DCRT at 54 Gy (range, 44–58.5 Gy) in 2-Gy fractions to 105 HCC patients, and found that V30 was a significant dosimetric predictor of radiation-induced hepatic toxicity [25]. Son et al. [26] conducted an investigation HCC patients treated with hypofractionated SBRT (30–39 Gy in three fractions) and showed that a total liver volume greater than 800 mL receiving less than 18 Gy significantly correlated with CP class progression. One prospective study analyzed HCC patient data of SBRT from phase 1 or 2 clinical trials, reporting that a dose of 800 cc of liver was strongly associated with liver toxicity in 114 patients, which was defined as an increase of at least 2 points in the CP score within 3 months after SBRT [27]. In the multivariate analysis, our study revealed Vliver as an independent prognostic factor for ncRILD after matching. Further, the lowest risk of RILD was seen when the Vliver ≥ 952.7 cc for using RT to treat HCC patients.
There were limitations of the research. First, this was a retrospective, small sample size, and single-center cohort study; thus, a potential for selection bias existed. PSM was used to balance the covariates between the two groups and to minimize bias. However, some biases were inevitable, such as differences in RT, ICI, and targeted drug regimens. Second, the study was conducted in an area where hepatitis B is endemic, and 91 (91.0%) patients in the matched group had an HBV infection. The subgroup analyses confirmed that the combination of RT with ICIs and targeted therapy was also feasible in the HBV population. The third limitation is the heterogeneity in ICIs and targeted agents as well as treatment cycles, and the inability to clearly determine treatment interruptions, although only a small number of patients experienced brief pauses.

Conclusions
The combination of RT with ICIs and targeted therapy does not markedly increase RIHT occurrence and may be a safe treatment option for HCC patients. As a dosimetric parameter, the Vliver is a strong predictor of ncRILD and can help identify effective and safe treatment strategies.

Supplementary Information
Below is the link to the electronic supplementary material.