Radiotherapy plus anti-PD1 versus radiotherapy for hepatic toxicity in patients with hepatocellular carcinoma

Purpose In this study, we aimed to compare the radiation-induced hepatic toxicity (RIHT) outcomes of radiotherapy (RT) plus antibodies against programmed cell death protein 1 (anti-PD1) versus RT alone in patients with hepatocellular carcinoma (HCC), evaluate prognostic factors of non-classic radiation-induced liver disease (ncRILD), and establish a nomogram for predicting the probability of ncRILD. Patients and methods Patients with unresectable HCC treated with RT and anti-PD1 (RT + PD1, n = 30) or RT alone (n = 66) were enrolled retrospectively. Patients (n = 30) in each group were placed in a matched cohort using propensity score matching (PSM). Treatment-related hepatotoxicity was evaluated and analyzed before and after PSM. The prognostic factors affecting ncRILD were identified by univariable logistic analysis and Spearman’s rank test in the matched cohort to generate a nomogram. Results There were no differences in RIHT except for increased aspartate aminotransferase (AST) ≥ grade 1 and increased total bilirubin ≥ grade 1 between the two groups before PSM. After PSM, AST ≥ grade 1 occurred more frequently in the RT + PD1 group (p = 0.020), and there were no significant differences in other hepatotoxicity metrics between the two groups. In the matched cohort, V25, tumor number, age, and prothrombin time (PT) were the optimal prognostic factors for ncRILD modeling. A nomogram revealed a good predictive performance (area under the curve = 0.82). Conclusions The incidence of RIHT in patients with HCC treated with RT + PD1 was acceptable and similar to that of RT treatment. The nomogram based on V25, tumor number, age, and PT robustly predicted the probability of ncRILD. Supplementary Information The online version contains supplementary material available at 10.1186/s13014-023-02309-1.


Introduction
Hepatocellular carcinoma (HCC) is a major global health problem, and its incidence is currently rising in most countries [1].The primary treatment options for earlystage HCC are surgery, radiofrequency ablation, and orthotopic liver transplantation [2][3][4].Unfortunately, most patients with HCC are diagnosed with advanced disease and a poor prognosis [5].Currently, the firstline molecular-targeted therapy for unresectable HCC includes treatment with sorafenib and lenvatinib, and the overall survival of patients with HCC is still unsatisfactory [6].Antibodies against programmed cell death protein 1 (anti-PD1) have yielded promising results for advanced HCC [7].In the IMbrave150 study, atezolizumab plus bevacizumab led to significant survival benefits; however, the combination incurs a high cost and results in 56.5% grade ≥ 3 TRAEs [8].
With advancements in radiotherapy (RT) technologies, including intensity-modulated radiation therapy (IMRT), an increasing number of patients with HCC have achieved good disease control after radiotherapy (RT) [9].Radiotherapy for patients with liver cancer has been recommended by the National Comprehensive Cancer Network as a standard treatment method [10].RT can potentiate tumor immunity and enhance antitumor effects in combination with immunotherapy [11].A case series involving five patients with unresectable HCC treated with stereotactic body radiotherapy (SBRT) followed by anti-PD1 reported a 100% response rate to treatment and a median PFS of 14.9 months [12].In a phase II trial, the combination of RT with camrelizumab (an anti-PD1) for patients with unresectable HCC showed promising efficacy and acceptable safety profile, with 52.4% of patients achieving an objective response [13].Combined SBRT and immunotherapy resulted in significantly superior survival and less toxicity compared with transcatheter arterial chemoembolization (TACE) [14].The combination of RT with anti-PD1 may, therefore, be a novel therapeutic strategy for HCC.
Radiation-induced hepatic toxicity (RIHT) is a common dose-limiting factor in the use of RT for HCC, in which the radiation-induced liver disease (RILD) is described as severe RIHT [15,16].Multiple studies have shown that the adverse reactions to anti-PD1 include abnormal hepatic function, including elevated transaminase and/or elevated total bilirubin [17][18][19].However, it is unclear whether the combination of radiotherapy with anti-PD1 increases the incidence of RIHT.This study aimed to compare the severity of RIHT between RT combined with anti-PD1 (RT + PD1) versus RT alone for HCC.In addition, prognostic factors for RILD were investigated.

Patients
All patients with HCC undergoing radiotherapy were screened between January 2017 and November 2022.The patients were diagnosed with HCC histologically and/ or radiologically based on the guidelines of the American Association for the Study of Liver Diseases [20] and staged according to the Barcelona Clinic Liver Cancer (BCLC) system [21].The general inclusion criteria for the study were as follows: [1] Patients with Child-Pugh (CP) class A or B and an Eastern Cooperative Oncology Group performance score of 0-2; [2] were not combined with intrahepatic cholangiocarcinoma; [3] had not received concurrent targeted therapy; [4] had not received surgery and ablation therapy between one month before the first fraction of radiotherapy and three months after the last fraction; [5] recovery of all hepatotoxic conditions of patients to grade 1 or less before the first fraction of radiotherapy in those who received prior interventional therapy; [6] patients without interventional therapy during RT and three months after the last fraction; [7] availability of dose-volume histogram (DVH) dosimetric parameters and RIHT-relevant data.After applying these criteria, 135 patients were registered; of these, 39 were excluded as shown in Fig. 1.Total of 96 patients were ultimately enrolled in this study, including 30 patients treated with RT plus anti-PD1 (RT + PD1 group) and 66 patients treated with RT alone (RT group) (Fig. 1).Finally, 30 patients in each group were included in the matched cohort.Ethical approval was obtained from the Guangxi Medical University Cancer Hospital (LW2022112).

Radiotherapy protocol
Contrast-enhanced computed tomography (CT) scans for RT planning were performed at 2.5-5 mm slice thickness under spontaneous breathing in the supine position.The gross tumor volume (GTV) was defined as the size of the intrahepatic tumor that was enhanced in the arterial phase.The magnetic resonance and CT images were then fused to better sketch the GTV.To compensate for organ motion and setup error, the planning target volume (PTV) comprised the GTV plus a 5-10 mm margin in all directions.All target volumes and organs at risk were delineated using the MIM 6.8 system (MIM, USA).The plans were designed using IMRT or volumetric-modulated arc therapy (VMAT).Based on this plan, the Pinnacle 3 system (Philips, Netherlands) or Monaco treatment planning system (version 5.1) was generated.A 6 MV X-ray (ELEKTA Versa-HD or ELEKTA Synergy, Sweden) linear accelerator was used.
The fractionated radiation doses were chosen based on the principles of 2 to 6 Gy/fraction.The patients received a median total IMRT dose of 51.0 Gy (47.5-60.0Gy) with a median of 3.0 Gy (2.4-4.0Gy) per fraction for a median of 20 (15)(16)(17)(18)(19)(20) fractions administered five days a week.The organs at risk (OARs) were well protected when the DVH analysis was performed to evaluate the radiotherapy plan.For the liver, the mean dose to the normal liver (Dmean) was less than 21 Gy.For the kidneys, V15 was < 1/3 volume.For the spinal cord, Dmax < 40 Gy.Similarly, the Dmax for the stomach, small bowel, and duodenum were < 40-45 Gy each [22].

Evaluation of liver hepatic metrics and dosimetric parameters
All patients underwent a CT and/or magnetic resonance imaging (MRI) within one month before the initiation of RT and every 2-3 months after RT to evaluate the hepatic toxicity and tumor response.RIHT was assessed based on the CP scoring system and common toxicity criteria for adverse events (version 5.0) within three Anti-PD1, antibodies against programmed cell death protein 1; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; RIHT, radiation-induced hepatic toxicity; RT, radiotherapy months after completion of the RT.The CP score (CP ≥ 1 or CP ≥ 2) is recognized as an effective system for evaluating RIHT [23].RILD was categorized into two types: classic RILD (cRILD) and non-classic RILD (ncRILD), within three months after completion of the RT.RILD resulted in anicteric hepatomegaly and ascites, an alkaline phosphatase (ALP) level at least twice the upper normal or baseline value (cRILD), an increase in the CP score by two or more, or an increase in alanine aminotransferase (ALT) or aspartate aminotransferase (AST) levels ≥ five times the upper limit of the normal or baseline value (ncRILD) in the absence of tumor progression and/or HBV reactivation (a 10-fold or greater increase in HBV DNA levels) [15,24,25].Dosimetric parameters, including the GTV, normal liver volume (Vliver), mean dose to the normal liver (Dmean), and percentage of normal liver volume receiving > x Gy radiation (Vx, x = 5, 7.5, 10, 15, 20, 25, 30, or 35) were analyzed using DVH [23].The Vliver was calculated by subtracting the lesion volume from the total liver volume.

Statistics
For patients in the RT + PD1 and RT groups, we adopted a 1:1 propensity score matching (PSM) method to minimize between-group heterogeneity and selection bias using a logistic regression model.The propensity score for the study included the following: age, sex, hepatitis B virus infection (HBV), CP grade, alpha-fetoprotein, tumor number, max tumor size, interventional therapy, hepatectomy, ablation, prothrombin time (PT), and Dmean.The clinical and dosimetric parameters were estimated using continuous or categorical variables.The chi-squared test (Fisher's exact test), Student's t-test, and Wilcoxon test were performed to compare the clinical, dosimetric, and hepatotoxicity between patients with RT or RT + PD1.
This study included 96 patients with HCC as factors for ncRILD, which were analyzed using a logistic regression model for univariate (p < 0.1) analysis.The Spearman rank test was used to analyze the correlations between the clinical and dosimetric parameters and that among the various dosimetric parameters (p < 0.2).The nomogram model was generated using the risk factors affecting ncRILD by multivariable logistic regression and assessed using the area under the ROC (AUROC) curves and calibration curve (with 1000 bootstrap resamples).We used R version 4.0.5 (http://www.r-project.org/) and SPSS® version 25.0 software (SPSS, Inc., Chicago, IL, USA) to analyze the data.

Patient characteristics and follow-up data
Of the 96 patients, 30 treated with RT + PD1 were matched to 66 treated with RT using PSM.The baseline characteristics, including clinical data and dosimetric factors, were not significantly different between the two groups after PSM (Table 1).In the RT + PD1 group, 10, 17, and 3 patients received anti-PD1 before the first fraction of RT, during RT, and after the last RT fraction, respectively.The patients received a median of five (range: 1-22) cycles of anti-PD1; 20, 2, 2, and 4 patients received camrelizumab, toripalimab, sintilimab, and tislelizumab, respectively.

Evaluation and incidence of RIHT
Five patients with liver disease were excluded because of tumor progression and HBV reactivation.The incidence of RIHT in the two groups before and after PSM is summarized in Tables 2 and 3.Among the 96 evaluable patients, 17.7%, 39.6%, 14.6%, 5.2%, 3.1%, and 1.0% experienced ncRILD, CP score ≥ 1, CP score ≥ 2, increased AST grade 3, increased ALT grade 3, and increased ALP grade 2 within three months after completion of the RT, respectively.The incidence of ncRILD before PSM showed in Supplemental Fig. 1a.No grade 4/5 hepatotoxicity was observed in any metric, and no grade 3 hepatotoxicity was observed in the metrics of increased ALP, increased total bilirubin, or decreased albumin.None of the patients developed cRILD.Before PSM, increased AST ≥ grade 1 was more frequent in the RT + PD1 group than in the RT group (66.7% vs. 37.9%, p = 0.016), while increased total bilirubin ≥ grade 1 was more frequent in the RT group than in the RT + PD1 group (57.6% vs. 33.3%,p = 0.048).There were no differences in other hepatotoxicity parameters, including ncRILD, CP score ≥ 1, CP score ≥ 2, increased AST ≥ grade 2, increased AST grade 3, increased ALT ≥ grade 1, increased ALT ≥ grade 2, increased ALT grade 3, increased ALP ≥ grade 1, increased ALP grade 2, increased total bilirubin grade 2, decreased albumin ≥ grade 1, and decreased albumin grade 2 (Table 2).Among the 60 evaluable patients after PSM, 23.3%, 38.3%, 18.3%, 5.0%, 5.0%, and 1.7% experienced ncRILD, CP score ≥ 1, CP score ≥ 2, increased AST grade 3, increased ALT grade 3, and increased ALP grade 2 within three months after completion of the RT, respectively.The incidence of ncRILD after PSM showed in Supplemental Fig. 1b.Increased AST ≥ grade 1 occurred more frequently in the RT + PD1 group (p = 0.020) than in the RT group, while there were no significant differences in the other hepatotoxicity parameters after PSM between the two groups (Table 3).

Prognostic factors for ncRILD
Univariate analyses of all patients after PSM were performed for the clinical and dosimetric factors of ncRILD, as shown in Table 4.The absolute Spearman's Rho values close to 1 of the dosimetric parameters showed that the two parameters were highly correlated (Supplemental Fig. 2).To avoid overfitting, only a dosimetric risk factor of V25 was included in the model.Optimal predictors, including V25, tumor number, age, and PT, were significantly associated with ncRILD (Table 4).Univariate analyses before PSM were performed for the clinical and dosimetric factors of ncRILD, as shown in Supplemental Table 1.The tumor number and Vliver were significantly associated with ncRILD (Supplemental Table 1).

Nomogram model
A nomogram model in the matched cohort was integrated based on multivariable logistic regression (Fig. 2a).

Discussion
In recent years, the combination of RT with immunotherapy has received close attention for HCC.RT can enhance antigen presentation and tumor immunogenicity for tumor phenotype modulation, improving the efficacy of cancer immunotherapy [26].Our previous studies suggested that RT combined with immunotherapy as a novel treatment strategy in patients with HCC showed promising efficacy and acceptable safety and may, therefore, be a promising therapeutic strategy for patients with HCC [13,27].RIHT remains a major challenge in patients with HCC undergoing liver irradiation, particularly RILD, which is a serious treatment-related complication [28,29].HCC patients receiving anti-PD1 can experience hepatic injury, such as elevation of transaminase or blood bilirubin [18,19].To the best of our knowledge, few studies to date have compared the effect of RT plus anti-PD1 versus RT alone on RIHT in patients with HCC.The present study showed that patients who received RT combined with anti-PD1 had a comparable incidence of hepatotoxicity as those who received RT alone before and after PSM.Our findings demonstrated RT plus anti-PD1 may not increase the risk of RIHT over that of RT alone among patients with HCC.
The incidence of hepatotoxicity in the present RT group is similar to that in the literature [25,30].Chapman et al. [25] reported that 48%, 25%, 10%, 17%, 13%, 2%, 6%, and 2% of patients with primary liver malignancies who received 30-50 Gy in five fractions with SBRT had at least a CP score increase of 1, CP score increase of 2, total bilirubin of G2, AST of G2, ALT of G2, ALP of G2, AST of G3, and ALT of G3, respectively.In a prospective study using SBRT (39-50 Gy in 3-5 fractions), an increase in CP score ≥ 1 and CP score ≥ 2 was observed in 14.3% and 9.4%, respectively, of 85 patients at three months and in 19.0%, and 11.8%, respectively, of 85 patients at six months.There was no observed cRILD or ncRILD (elevated ALT or AST) [23].In addition, Jun et al. [31] reported that the incidence of RILD (elevated liver transaminases ≥ grade 3 or CP ≥ 2) was 24.7% among patients with HCC treated with SBRT using 40-60 Gy in 3-5 fractions.In summary, the hepatotoxicity when using RT to treat patients with HCC are acceptable.
A case series of five patients with unresectable HCC who were treated with SBRT followed by anti-PD1 showed that none of the patients developed classic RILD or a CP score ≥ 2. There were 1, 2, and 2 patients who had G1 elevation in AST, G1 elevation in ALT, and G2 elevation in AST/ALT, respectively [12].However, the number of patients treated with RT combined with anti-PD1 in the study was relatively small.Moreover, in a phase II trial of 21 patients with unresectable HCC treated with combined RT and camrelizumab (an anti-PD1), grade 1-2 adverse events comprised increased AST in 11 patients (52.4%), increased ALT in 10 (47.6%), increased blood bilirubin in 4 (19.1%), and decreased albumin in 11 (52.4%)[13].These studies showed that the treatment toxicities were manageable in patients with HCC treated with RT + PD1.Similarly, only one patient (3.3%) who received RT combined with anti-PD1 experienced increased AST grade 3, and no other grade 3-5 hepatotoxicity was observed in this study.The hepatotoxicity in the RT + PD1 group did not differ from that in the RT group except for increased AST ≥ grade 1 and increased total bilirubin before PSM and decreased albumin ≥ grade 1 after PSM; these toxicities were mild and manageable.Additionally, the rates of RILD did not differ between the RT and RT + PD1 groups (incidence of 23.3% for both, p = 1.000).Thus, our study showed that the combination of RT with anti-PD1 for patients with HCC was feasible and that its hepatotoxicity was acceptable, although prospective studies are required to improve its safety for further study.
Notably, accurate prediction of RT toxicity in patients with HCC will assist with achieving optimal RT planning, which may help physicians choose the best therapeutic regimen.However, the predictors of hepatotoxicity are not well established.In the present study, cRILD was not observed.Therefore, the relatively serious hepatic toxicity, described as ncRILD, was selected to analyze the prognostic factors for patients with HCC [32].The results showed that treatment with RT alone or combined with anti-PD1 was not correlated with ncRILD.Several dosevolumetric factors are significantly associated with RILD [15,33].In a study of patients who received three-dimensional conformal radiation therapy with a radiation dose of 38-68 Gy and a fraction size of 4-6 Gy, a V25 of 35% showed statistical significance as liver radiation tolerance for RILD.Age, tumor number, and PT were found to be optimal predictors for ncRILD to construct an effective model.Moreover, the tumor number and PT were the most significant factors associated with ncRILD for patients with Child-Pugh grade B with HCC after IMRT [34].According to the model, the probability of ncRILD was relatively low for patients with lower scores, which predicts the safety of RT.Therefore, in the era of precision oncology, our results may make an important contribution to RT treatment strategies for patients with HCC.
This study had several limitations.First, this study was retrospective, although PSM was used to balance the differences between the two groups.Second, this was a single-center study with a small sample size.Third, the types and schedules of anti-PD1 used for the treatment were heterogeneous, although the best RT / anti-PD-1 schedule, RT dose, fractionation scheme has not been specified yet [35].Fourth, large number of patients lost lab test (n = 21) may resulted in bias, yet the clinical data of enrolled patients is complete, and we observed that the incidence of RIHT in patients with HCC treated with RT plus anti-PD1 was acceptable and similar to that of patients treated with RT alone.In addition, our study lacks independent validation.Multi-center and prospective studies are required to confirm these findings.

Conclusions
results of this study indicate that the incidence of RIHT in patients with HCC treated with RT plus anti-PD1 was acceptable and similar to that of patients treated with RT alone.A nomogram based on V25, tumor number, age, and pre-PT, which are useful predictors of ncRILD, can help with delivering personalized therapy for patients with HCC.

Fig. 2
Fig. 2 Model prediction and evaluation for ncRILD.(a) Nomogram based on V25, tumor number, age, and PT for ncRILD prediction.The total score for each patient is used to predict the probability of ncRILD.(b) Receiver operating curve curves of the nomogram to predict ncRILD.(c) Calibration curves for ncRILD nomogram prediction.AUC, the area under the curve; ncRILD, non-classic radiation-induced liver disease; PT, prothrombin time; V25, the percentage of normal liver volume receiving > 25 Gy radiation

Table 2
Post-treatment Hepatotoxicity Metrics before PSM.

Table 3
Post-treatment Hepatotoxicity Metrics after PSM.

Table 4
Univariate analysis of parameters associated with the risk of ncRILD after PSM (n = 60)