Pembrolizumab and stereotactic body radiotherapy combined in advanced hepatocellular carcinoma post sorafenib – A phase II trial (PEMRAD)

Abstract

Background & Aims

There is strong rationale for integrated approaches combining immune checkpoint inhibitors with locoregional therapies in hepatocellular carcinoma (HCC).

Methods

The PEMRAD phase II trial investigated pembrolizumab in combination with stereotactic body radiation therapy (SBRT) in patients with advanced HCC following progression on sorafenib. Patients received pembrolizumab on Day 1, and SBRT commenced on Day 2. Up to 10 hepatic lesions (<20 cm) were treated with SBRT. Pembrolizumab was continued every 21 days until disease progression, unacceptable toxicity, or withdrawal. The primary endpoint was objective response rate (ORR) by RECIST v1.1, with a hypothesized improvement to 40% for the combination. Tissue and liquid biopsies were collected for correlative analyses, including immunohistochemistry, longitudinal blood cytometry by time-of-flight, and serum cytokine assessment.

Results

Between March 2018 and March 2023, 18 of a planned 22 patients were enrolled; 11 (61%) had macrovascular invasion (MVI) and 15 (83%) had extrahepatic metastases. Viral hepatitis was the underlying etiology in 50%. The ORR was 41% (95% CI 18–67%). Median progression-free survival was 5.4 months (2.8–9.9) and median overall survival was 12.6 months (5.7–25.8). Of 34 liver lesions treated with SBRT, only three progressed, in two patients. Among the 11 patients with MVI, 5 (45%) achieved a response, including one complete response. Treatment-related adverse events of grade ≥3 occurred in four patients (22%). One treatment-related death due to myocarditis was observed. Peripheral immunophenotyping showed that higher abundance of CD8⁺CD103⁺ T cells correlated with improved survival, whereas higher natural killer cell abundance was associated with inferior outcomes.

Conclusion

The combination of SBRT and pembrolizumab demonstrated a high response rate as second-line therapy for advanced HCC and warrants further evaluation, particularly in patients with MVI.

Impact and implications

This study demonstrates that combining stereotactic body radiotherapy with pembrolizumab is feasible and yields promising response rates in patients with advanced, previously treated hepatocellular carcinoma, including those with macrovascular invasion – a group typically associated with poor outcomes. These findings support further evaluation of this integrated approach and provide important guidance for the design of future clinical trials aimed at improving outcomes in advanced hepatocellular carcinoma.

Graphical abstract

Highlights

• •Stereotactic body radiotherapy plus pembrolizumab induces high response rates in advanced HCC previously treated with sorafenib.
• •The combination was well tolerated, with no treatment-related hepatic decompensation.
• •Patients with HCC and macrovascular invasion may particularly benefit from stereotactic body radiotherapy.
• •Higher peripheral CD8+CD103+ T cell levels were associated with improved overall survival.

Introduction

Hepatocellular carcinoma (HCC) is a leading cause of cancer related mortality world-wide occurring on a background of chronic liver disease in the majority. Although there has been a reduction in the incidence of viral hepatitis,¹ the simultaneous rise in metabolic dysfunction-associated steatotic liver disease has resulted in a projected increase in the prevalence of HCC of over 50% by 2040.² In addition, viral hepatitis remains a major challenge in endemic areas. Despite surveillance guidelines in high-risk patients, the majority present with incurable disease where locoregional and/or systemic therapies are recommended.³ Until 2018, the only systemic option in advanced disease was sorafenib, a multi-receptor tyrosine kinase inhibitor, which functioned mainly as a cytostatic agent, providing a median overall survival (OS) of between 6.5-10.7 months.^4,5 Approved treatments following progression or intolerance to sorafenib include the multi-receptor tyrosine kinase inhibitors cabozantanib⁶ and regorafenib,⁷ with objective response rates (ORRs) of 4% and 7%, respectively. In this setting, the median OS documented for cabozantinib was 10.2 months and for regorafenib was 10.6 months. In patients with an alpha-fetoprotein (AFP) level of >400 ng/ml, ramucirumab, a VEGF2 inhibitor demonstrated an ORR of 5% and median OS of 8.5 months.⁸ Notably in these three second-line trials, the percentage of patients with macrovascular invasion (MVI) was ≤35%. MVI is present or develops in over 50% of patients with HCC during the course of their illness and is associated with poor prognosis. In the context of portal vein thrombus (PVT), MVI can be defined according to the Japanese guidelines, where VP1 represents the presence of a thrombus distal to second order branches and VP4 equates to a thrombus in the main trunk of the portal vein. The optimal management of patients with advanced HCC and MVI remains unknown.

The treatment landscape in HCC has been revolutionized by the introduction of immune checkpoint inhibitors (ICIs) either alone or in combination. Based on results of Keynote-224⁹ and CheckMate 040,¹⁰ the FDA approved both pembrolizumab and ipilimumab plus nivolumab post sorafenib in the second-line setting. Despite the integration of ICIs across stages in HCC, biomarkers to predict responses are still lacking. The PEMRAD trial investigated the combination of stereotactic body radiotherapy (SBRT) delivered to intrahepatic HCC combined with pembrolizumab to target both intrahepatic and extrahepatic HCC as a second-line therapy following progression on sorafenib.

Patients and methods

Study design

PEMRAD is a single center, open-label investigator led phase II trial of pembrolizumab in combination with hypofractionated SBRT in patients with HCC previously treated with sorafenib. The trial was approved by the institutional research ethics board, conducted in accordance with the Declaration of Helsinki, complied with all applicable regulations, and was registered on ClinicalTrials.gov (NCT03316872). All patients provided written consent prior to study participation.

Patient population

Patients with histologically confirmed advanced HCC with a Child-Pugh scores of 5 or 6 and ECOG performance status of 0-1 were eligible. Patients must have progressed on sorafenib. Patients had to have intrahepatic HCC, treatable with personalized SBRT (27.5 Gy to 50 Gy in 5 fractions). A maximum of 10 hepatic lesions could be treated with a total sum of treated tumor lesion diameters <20 cm, with no limit on the extent of MVI (if it could be encompassed with SBRT). In addition, lesions to be treated with radiation could not invade bile ducts or nearby viscera. It was permitted to exclude small intrahepatic lesions from the irradiated volume. Patients were excluded if they had received prior upper abdominal radiation therapy or transarterial radioembolization.

Procedures

Patients received the first dose of pembrolizumab (200 mg) on Day 1 and continued every 21 days, until disease progression or intolerable toxicity. SBRT commenced on Day 2 and was delivered in five fractions over 5-15 days in accordance with the institutional protocol, with a preference for delivery of SBRT daily Monday to Friday. The prescribed dose to the planning target volume (PTV) was 50 Gy, 45 Gy, 40 Gy, 35 Gy, 30 Gy, or 27.5 Gy in five fractions, determined by normal tissue constraints according to local standards of care and consistent with the RTOG 1112 trial.¹¹ The goal was to encompass the HCC gross tumor, including MVI, at the highest feasible dose level. Treatment of all dominant HCC lesions with MVI was mandated, up to a maximum of 20 cm, whereas treatment of satellite HCCs was not required, but was permitted at the treating physician’s discretion. Doses to multiple PTVs could be different. The preferred inter-fraction interval was 24 h, except when luminal organs at risk were at their maximal dose, in which case every-other-day SBRT was preferred. All lesions were treated on the same day. At the time of study design, the combination was novel with unknown toxicities. As a safety precaution, the first 10 patients were required to meet the following additional liver dosimetric constraints, in addition to local standard dose limits: a mean liver dose (liver minus HCC gross tumor volume [GTV]) of <15 Gy in five fractions; a liver-minus-GTV D800cc ≤18 Gy in five fractions; and a biologically corrected liver normal tissue complication probability of <0.5%. SBRT planning was performed using IMRT (intensity-modulated radiation therapy) or VMAT (volumetric-modulated arc therapy), with individualized breathing-motion management and daily image-guided radiation therapy according to local standards of care. When the stomach or more than 50% of the uninvolved liver was due to be irradiated, ondansetron was administered 1 h before SBRT as an antiemetic.

Outcomes

The primary objective of this study was to assess the efficacy of combined SBRT and pembrolizumab, as measured by ORR (RECIST 1.1) in patients with locally advanced HCC with or without distant metastases. In addition to the ORR (including irradiated and unirradiated measurable and non-measurable HCC), the study prespecified evaluation of the response of MVI separately. MVI response measurements were detailed, consistent with RECIST1.1 for non-measurable disease (Appendix 1). In brief, a complete response was indicated by complete resolution of MVI, with recanalization of the vessel; a partial response (PR) if there was any of a) partial recanalization of the vessel (if prior complete blockage), b) unequivocal reduction in the maximal girth of MVI, c) unequivocal reduction in the volume of MVI, or d) elimination of arterial enhancing portion of MVI.

Secondary objectives included progression-free survival (PFS), safety and tolerability, and overall survival (OS). Exploratory objectives at the time of study design included tissue and blood translational endpoints to determine putative biomarkers of response.

Tumor imaging was performed every 12 weeks on study until confirmed progression or treatment discontinuation. Suspected but unconfirmed progression had to be confirmed within 4-8 weeks. Adverse events (AEs) and treatment-related adverse events (TRAEs) were recorded according to National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) v5.0.

Correlative analyses

A screening biopsy was mandatory for immunophenotyping using immunohistochemistry (IHC), and blood samples were obtained longitudinally for changes in immune cell populations (supplementary CTAT table). Blood samples were collected at up to five time points on Cycle 1: Day 1, Day 5, and Day 10, Cycle 3 Day 1, and at progression (Fig. 1).

Fig. 1.

Study schema of the phase II PEMRAD trial combining SBRT and pembrolizumab.

HCC, hepatocellular carcinoma; SBRT, stereotactic body radiotherapy.

Cytokine analysis

Cytokine analysis was performed on longitudinal plasma samples using a custom ProcartaPlex™ (Luminex) panel (Thermo Fisher, MA, USA) following the manufacturer’s protocol. This panel allows for the simultaneous measurement of cytokines, chemokines, and growth factors (including GM-CSF, IFN-A, IFN-G, IL-1A, IL-1B, IL-10, IL-13, IL-15, IL-1RA, IL-2, IL-22, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-17A, CXCL10, MCP-1, MIP-1A, MIP-1B, RANTES, TNF-A, GRO-A, CXCL12, TNF-B, IL-21, IL23, IL27, IL-12p70, EOTAXIN, IL-21, IL-18 and HGF) in a single sample. Data were acquired on a Bio-Plex 200 (Bio-Rad, CA, USA).

Cytometry by time-of-flight analysis

Full methods are provided in Appendix 2. In brief, peripheral blood mononuclear cells (PBMCs) were isolated and viably cryopreserved from patients prior to treatment initiation and at several timepoints during therapy (Fig. 1). PBMCs were thawed and up to 20 samples were barcoded (Cell-ID 20-plex Pd Barcoding Kit, Standard BioTools) and pooled into a single batch, which was then labelled with cocktails of metal-tagged antibodies targeting either cell surface or intracellular markers. Data were pre-processed using FlowJo Software (BD). High-dimensional analysis and downstream statistical analyses were performed in R (v4.2.1).

Immunohistochemical staining

Available formalin-fixed paraffin-embedded slides were stained for PD-L1 using the Merck 22C3 antibody (membrane expression) and for tumor-infiltrating lymphocytes (TILs) using a morphological assessment of the presence/absence of TILs within tumor nests (with an overall average value for the whole sample, scale 0-3). For both PD-L1 and TIL, scoring was performed at a central site using QualTek Molecular Laboratories, Pennsylvania. In addition, formalin-fixed paraffin-embedded samples and cryosections were co-stained for NKp46 (#MAB1850, R&D Systems, MN, USA) and granzyme B (#AF1865, R&D Systems) (full methods in Appendix 2 and the supplementary CTAT table). Comprehensive image analysis was performed using HALO software v.3.4 (Indica Labs, NM, USA). Tumor areas annotated by a study pathologist were imported. NKp46, Granzyme B and hematoxylin intensities were detected using multiplex IHC v.3.1.4 algorithms to identify positive cells according to DAB and Warp Red staining status. The double-positive cells and tumor areas were calculated.

Statistical analyses

The trial was designed to detect an increase in ORR from 15% (with immune checkpoint inhibitor therapy alone, based on prior studies) to 40% with the combination, using a two-stage design to allow for 22 evaluable patients. The primary endpoint of ORR with corresponding 95% CIs was determined using the Clopper-Pearson method. Median PFS and OS were determined using the Kaplan-Meier method. Safety outcomes are summarized using descriptive statistics.

Results

Between March 2018 and July 2023, 19 patients were enrolled and 18 treated according to the study protocol (Fig. S1). One patient withdrew post enrollment. The study was terminated prematurely, primarily due to changes in the standard-of-care first-line systemic therapy for HCC, which prevented further enrollment. At database lock on February 15th, 2024, the median follow-up (from enrollment to clinical data cut-off) was 25.3 months. The median age was 66.5 (range: 45-78) years and 17 (94%) patients were male (Table 1). Of patients enrolled, 50% had non-viral-related HCC. All patients had advanced BCLC (Barcelona Clinic Liver Cancer) C HCC and had progressed on prior sorafenib, including 83% with extrahepatic disease. The median diameter of baseline target lesions (RECIST) was 10.4 cm (2.6-25.0 cm). The median time on prior sorafenib was 7 months (2-36) (Table S1). Notably, 11 (61%) patients enrolled had evidence of MVI, including nine patients with PVT (VP4: n = 5; VP3: n = 3; VP2: n =1) and two patients with hepatic vein involvement. Of the nine with PVT, two had synchronous hepatic vein tumor thrombus.

Table 1.

Baseline characteristics of patients enrolled.

Characteristic
Age	66.5 (45-78)
Sex	Male = 17 (94%)
Background liver disease
Hep B	3 (17%)
Hep C	6 (33%)
MASLD/alcohol/other	9 (50%)
Baseline AFP median (range) ng/ml	3,226 (2-129,670)
BCLC
B	1 (5%)
C	17 (95%)
ECOG performance status
0	12 (67%)
1	6 (33%)
Vascular invasion
Yes	11 (61%)
No	8 (39%)
Extrahepatic disease
Yes	15 (83%)
No	3 (17%)
Prior curative surgery
Yes	2 (11%)
No	16 (89%)
Prior LRT
Yes	6 (33%)
No	12 (67%)
Child-Pugh
A5	13 (72%)
A6	5 (28%)
ALBI
Grade 1	6 (33%)
Grade 2	12 (67%)
Prior systemic tx
Sorafenib	18 (100%)

AFP, alpha-fetoprotein; ALBI, albumin-bilirubin; BCLC, Barcelona Clinic Liver Cancer; LRT, locoregional treatment; MASLD, metabolic dysfunction-associated steatotic liver disease.

Patient disposition and treatment exposure

At the time of data cut-off, all patients were off treatment. Disease progression was the reason for discontinuation in 15 (83%) patients. The median number of cycles received was 7.5 (range: 2-35). Dose delays or omissions occurred in eight patients (44%) which were attributed to AEs.

A summary of radiation treatment is provided in Table 2. Radiotherapy was delivered over a median of 6 days (range 5-10 days). The median volume of the total GTV treated was 192.2 cc (IQR 93.2-306.9). The corresponding median PTV was 356.9 cc (IQR 114.5-594.4). All patients received five fractions of SBRT. Minimum dose (D_99%) and mean dose to the dominant GTV was 33.1 Gy (IQR 28.4-44.9) and 39.3 Gy (IQR 35.4-48.8), respectively. Median maximum dose to any luminal organ was 15.3 Gy (IQR 7.2-22.5). Median liver volume (non-GTV) was 1,672 cc (IQR 1,432-1,921). The median mean liver (non-GTV) dose was 12.7 Gy (IQR 11.9-14.3), and the median liver (non-GTV) D_800cc was 7.0 Gy (IQR 5.1-10.0). Subsequent anticancer therapy was initiated in four patients (22%) including one patient receiving SBRT for disease progression in the spleen.

Table 2.

Radiotherapy parameters delivered.

	Median (IQR)
GTVD	184.2 cc (69.1-284.8)
Total GTV	192.2 cc (93.2-306.9)
Prescribed dose to dominant lesion	32.5 Gy (30.0-40.0)
GTVD D99%	33.1 Gy (28.4-44.9)
GTVD mean dose	39.3 Gy (35.4-48.8)
PTVD	261.7 cc (114.5-417.5)
Total PTV	356.9 cc (114.5-594.4)
PTVD D99%	28.1 Gy (25.3-35.2)
PTVD mean dose	37.0 Gy (33.6-46.4)
Luminal GI organ max dose	15.3 Gy (7.2-22.5)
Heart max dose	26.6 Gy (11.1-33.3)
Liver volume	1,672 cc (1,432-1,921)
Liver mean dose	12.7 cGy (11.9-14.3)
Liver D800cc	7.0 Gy (5.1-10.0)

GTV, gross tumor volume; GTV_D, dominant GTV; PTV, planning target volume; PTV_D, dominant PTV. Liver size and doses represent liver volume with tumor subtracted. Non-GTV liver D800 precise to 5 cGy.

Response and survival outcomes

Of the 18 patients included, 17 were evaluable for response (Table 3). The ORR was 41% (95% CI 18-67%) (Fig. 2A,B; Table 3); responses occurred regardless of etiology or baseline AFP. Of the seven patients achieving a PR, four had non-viral and three viral etiologies. Of five patients with progressive disease as best response (two viral, three non-viral), four had disease progression outside of irradiated HCC lesions and one patient had evidence of clinical progression with AFP rise without confirmed radiological progression. The median PFS (95% CI) was 5.4 (2.8-9.9) months and the 12-month PFS rate was 22% (95% CI 7-43%) (Fig. 3A). Median OS (95% CI) was 12.6 (5.7-25.8) months (Fig. 3B) and the 12-month OS rate was 50% (95% CI 26-70%). Among patients who achieved a PR, median OS was 30.2 months (95% CI 8.2-not reached) (Fig. 2B).

Table 3.

Summary of overall response and vascular response as per RECIST v1.1.

Responses
ORR using RECIST v.1.1	7 (41%)
Best overall response, n (%)
CR	0
PR	7 (41%)
SD	5 (29%)
PD	5 (29%)
Unable to determine	1
DCR	12 (71%)
DCR with SD ≥6 months	9 (53%)
DOR, months, median	2.8 (0.6-16.5)
Vascular response (n = 11)
CR	1 (9%)
PR	4 (36%)
SD	5 (45%)
PD	1 (9%)
DCR	10 (91%)

CR, complete response; DCR, disease control rate; DOR, duration of response; ORR, overall response rate; PD, progressive disease; PR, partial response; SD, stable disease.

Fig. 2.

Tumor responses.

(A) Waterfall plot demonstrating the overall response rate of patients on trial; (B) spider plot showing change in tumor diameter over time. PD, progressive disease; PR, partial response; SD, stable disease.

Fig. 3.

Kaplan-Meier survival curves.

Kaplan-Meier survival curves showing (A) progression-free and (B) overall survival estimates.

There were 16 patients (89%) with an elevated AFP at baseline (median AFP at baseline in cohort: 3,226 ng/ml); of these, 81% (n = 13) achieved a reduction in AFP, including 9 of 16 (56%) patients with a >25% AFP decline by cycle 2 of pembrolizumab.

Of the 11 patients with MVI, a partial response was documented in 5 (45%) and disease control in 91%. One patient with VP4 had a complete response and three additional patients with VP4 MVI documented PRs. One additional patient with a hepatic vein thrombus also had a PR to treatment. In total, 34 lesions were treated with SBRT in the 18 patients enrolled. The full extent of intrahepatic HCC within the liver was irradiated with the exception of two patients who had small satellites excluded to avoid excessive hepatic irradiation.

At data cut-off, one patient had discontinued the study due to AEs, one had withdrawn voluntarily, and one had completed two years of pembrolizumab and stopped without evidence of progression. Of the remaining 15 patients who progressed, 11 had documented RECIST progression events. Only two patients progressed in irradiated lesions, including one patient with intrahepatic progression of two lesions and one with progressive MVI who also progressed outside the liver. Three patients developed new liver lesions, including one with new MVI within the portal vein, and six patients developed both new intrahepatic lesions and progression of extrahepatic disease (Table S1).

Safety

TRAEs of any grade were reported in 16 (89%) patients (Table 4). The most commonly reported AEs were fatigue (61%), anorexia (39%) and nausea (33%), the majority of which were grade 1-2. Grade 3 or greater AEs occurred in three patients including grade 3 AST elevation (n = 1), and grade 3 maculopapular rash (n = 1). One patient developed grade 3 neutropenia together with grade 2 anemia and thrombocytopenia; however, the patient declined a bone marrow aspirate for further investigation. The etiology of the myelosuppression was thus unknown, but it occurred in the setting of progressing HCC with liver decompensation and infection. However aplastic anemia could not be ruled out and this event was reported as a possible TRAE. One patient developed a fatal myocarditis after two cycles of pembrolizumab which was thought possibly related (no autopsy). In total, 5 (28%) patients were treated with steroids during the study for AEs attributed to pembrolizumab. Of note, there was no synthetic liver function decline or luminal gastrointestinal toxicity attributable to SBRT or combined modality therapy. Of the 18 patients included, liver decompensation was documented in 12; all events were attributed to disease progression.

Table 4.

Table of treatment-related adverse events.

	Any grade	Grade 1-2	Grade 3	Grade 4	Grade 5
Adverse events
Fatigue	11 (61%)	11(61%)
Anorexia	7 (39%)	7 (39%)
Nausea	6 (33%)	6 (33%)
AST	4 (22%)	3 (17%)	1 (6%)
ALT	4 (22%)	4 (22%)
Pruritis	4 (24%)	4 (22%)
Rash	3 (22%)	11 (17%)	1 (6%)
Arthralgia	2 (11%)	2 (11%)
Diarrhea	2 (11%)	2 (11%)
Dysgeusia	2 (11%)	2 (11%)
Hypothyroidism	2 (11%)	2 (11%)
Fever	2 (11%)	2 (11%)
Pneumonitis	1 (6%)	1 (6%)
Myocarditis	1 (6%)				1 (6%)
TSH increased	1 (6%)	1 (6%)
Myalgia	1 (6%)
Aplastic anemia∗	1 (6%)		1 (6%)
Colitis	1 (6%)	1 (6%)
Hypothyroidism	1 (6%)	1 (6%)

ALT, alanine aminotransferase; AST, aspartate aminotransferase; TSH, thyroid-stimulating hormone.

Correlative analyses

Peripheral immunophenotyping and cytokine response

We used cytometry by time-of-flight to evaluate the temporal dynamics of peripheral immune cells in response to therapy and their potential association with clinical outcomes. A panel of 41 metal-tagged antibodies targeting proteins that define immune cell lineage, differentiation and activation was used to profile serial PBMCs. Unsupervised clustering revealed 21 clusters corresponding to 18 distinct immune populations (Fig. 4A). We first compared the abundances of immune subsets at 10 days (C1D10) and 9 weeks (C3D1) after treatment initiation to those at C1D1. At C1D10 we observed a reduction in the abundance of several lymphocyte populations (B cells, CD4 memory and naïve subsets, CD8 naïve, CD8 TEMRA, CD8+CD103+ and double-positive T cells). By contrast, an increased frequency of myeloid cells (CD11c+HLADR+) and myeloid-derived suppressor cell-like cells (CD11c+ HLA-DRlo) was observed at C1D10. These widespread changes in the composition of circulating immune cells could be due to homing to tissues or a systemic inflammatory response induced by the radiotherapy. By C3D1, most populations recovered to a similar abundance as baseline; however, a modest but significant increase in CD4+ regulatory T cells was also observed. These changes in immune profile occurred independently of patient outcomes (Fig. 4B).

Fig. 4.

Immune cell populations.

(A) UMAP of peripheral cytometry by time-of-flight identifying 21 clusters corresponding to 18 distinct immune cell populations. (B) Abundance of peripheral immune cell populations at C1D1, C1D10 and C3D1 (cytometry by time-of-flight). UMAP, uniform manifold approximation and projection. Statistical analyses performed using wilcoxon signed-rank test.

We next investigated whether any immune subsets at baseline were correlated with the occurrence of a progression event. Notably, PFS (p = 0.02) and OS (p = 0.07) were longer in patients with a higher circulating frequency of CD8+CD103+ T cells. In contrast, a lower frequency of a cluster of natural killer (NK) cells (cluster 15) expressing CD57 and Granzyme B was associated with improved PFS (p = 0.004) and OS (p = 0.02) (Fig. 5A-D).

Fig. 5.

Kaplan-Meier curves demonstrating impact of immune cells on PFS and OS.

Kaplan-Meier curves demonstrating impact of (A) peripheral CD8+CD103+ T cells on PFS, (B) peripheral GzmB+CD57+NK cells on PFS, (C) peripheral CD8+Cd103+ T cells on OS, and (D) peripheral GzmB+Cd57+ NK cells on OS. NK, natural killer; OS, overall survival; PFS, progression-free survival.

We further investigated the systemic response to therapy by measuring a panel of inflammatory and immunosuppressive cytokines. We observed a high degree of inter-patient variability in cytokine profiles with no significant changes in profiles over time and no correlation with patient outcomes (Fig. S2). However, CXCL10, Eotaxin and HGF significantly increased in patients with MVI between C1D1 and C3D1, whereas no difference was seen for patients without MVI (Fig. 6).

Fig. 6.

Dynamic changes in peripheral cytokines CXCL10, EOTAXIN and HGF when MVI present.

MVI, macrovascular invasion. Statistical analyses performed using wilcoxon signed-rank test.

Tissue IHC

Of the 18 patients included, tumor tissue from 12 patients was available for NKp46 and granzyme B co-stain IHC. We observed low frequencies of NK cell infiltration. There were no correlations between the % or density of the double-positive cells with circulating double-positive NK cell abundance, or outcome.

Tumor PD-L1 expression and TIL score were available in 11 patients but did not appear to predict response to combination treatment (Table S1). Nine of these patients had PD-L1 modified proportion scores ≤5%. One patient did have a PD-L1 score of 35% and a PR, with a PFS of 25 months. Samples were not available in six patients due to absent tumor in biopsy or inadequate volume or quality of tissue for analysis.

Discussion

In the PEMRAD study, despite early closure, we document an overall ORR of 41% when combining pembrolizumab and SBRT in patients with advanced, treatment-refractory HCC. Patients enrolled in this study had definitively progressed on prior sorafenib, and 83% had extrahepatic disease while 61% had MVI. The median OS of 12.3 months in a second-line population with high disease burden is promising. In addition, of those patients with MVI, 45% had a response of tumor MVI, including one complete response. This control of HCC MVI could have meaningful impact on preserving liver synthetic function. Furthermore, local control rates were high with only 3 of 34 irradiated lesions progressing in two patients, during the study period. There were no episodes of liver decompensation related to treatment with the combined approach. The safety profile was as expected; however, one patient died early from suspected treatment-related myocarditis.

In the KEYNOTE-224 trial, in which only 17% of patients had MVI, pembrolizumab alone after sorafenib demonstrated an ORR of 17% and a median OS of 12.9 months.⁹ In the IMBrave150 study, patients with VP4 disease had significantly inferior outcomes. The median OS was only 7.6 months (95% CI 6.0–13.9) for those treated with atezolizumab and bevacizumab compared to 5.5 months (95% CI 3.4–6.7) with sorafenib.¹² Although the ORR (RECIST) was higher than Keynote-224, ORR is not a good surrogate for OS in HCC, particularly when most progression events in this study were outside the irradiated field.¹³ Increasingly, combination strategies with ICIs and locoregional therapies are being investigated in HCC, particularly in those with intermediate disease. The LEAP-012 trial¹⁴ and the EMERALD-1 trial,¹⁵ evaluating transarterial chemoembolization (TACE) and ICI combinations in early/intermediate stage HCC, both met their primary endpoints of improved PFS. Similarly, the NASIR HCC trial documented the efficacy of selective internal radiation therapy plus nivolumab as first-line therapy in patients with BCLC B2 HCC or unilobar disease with MVI.¹⁶

However, few studies have evaluated locoregional therapy in combination with ICIs in advanced HCC. In particular, treatment options are limited in patients with MVI, and trials often exclude patients with VP3/4 PVT. Intra-arterial therapies can have a limited role when MVI is present. In a meta-analysis of 13 studies and 1,933 patients with HCC and MVI treated with TACE, the median OS was only 8 months.¹⁷ In contrast, SBRT has an emerging role in patients with MVI. Our group’s early experience with SBRT and MVI has documented a median OS of 18.3 months¹⁸ and a retrospective study has shown local control and stability in MVI in patients progressing on ICIs.¹⁹ Yet most patients with MVI progress outside the irradiated volume, providing rationale for combining SBRT with systemic and/or regional therapies. The combination of systemic therapy and SBRT is particularly attractive in patients with high-risk HCC. The phase III RTOG 1112 trial provides strong support for the role of SBRT in this patient population, as nearly 75% of enrolled patients had MVI.¹¹ Compared to sorafenib alone, the combination of SBRT and sorafenib improved median OS from 12.3 months to 15.8 months (HR 0.77, p = 0.06). When adjusting for performance status, stage, Child-Pugh score, and degree of MVI, this difference was statistically significant (HR 0.72, p = 0.04).¹³ In addition, the combination of radiation, anlotinib and tislelizumab has demonstrated an ORR of 57.8% and a median PFS of 7.8 months in patients with advanced HCC, not previously treated with systemic therapy.²⁰ Chiang et al. recently described a series of 63 patients with liver-confined unresectable HCC receiving SBRT and ICIs, with or without prior TACE. In this series, 60% of patients had MVI and 46% achieved a CR. The local control rate at 3 years was 90%.²¹ Similarly the disease control rate of irradiated lesions in our study was 89%.

Technological advances in radiation therapy, including MR-adaptive radiation therapy, which targets the tumor daily, and proton beam therapy, which delivers less low-dose radiation to normal tissues, have the potential to further improve the efficacy of radiation treatment by enabling more extensive HCC to be treated while reducing doses to the liver, blood, and other normal tissues.

Biomarkers to predict response to ICIs are lacking. The presence of CD8+ T cells or an effector T cell signature at baseline has been shown to associate with improved OS in early-stage disease when employing neoadjuvant strategies.^22,23 Pre-existing immunity has also correlated with response to atezolizumab and bevacizumab in advanced HCC.²⁴ We herein show that the abundance of peripheral CD8+CD103+ T cells was associated with improved outcomes. In tissue specimens, CD103 can identify tissue-resident T cells and has correlated with improved outcomes and response to ICIs in non-small cell lung cancer and melanoma, respectively.^25,26 We found a higher abundance of peripheral CD57+Granzyme B+ NK cells was associated with poorer PFS. Interestingly, data regarding the prognostic value of peripheral NK cells is conflicting in other solid tumors. For instance, higher NK populations were associated with improved OS in colorectal cancer²⁷ although CD57 was not utilized as a cell surface marker in that study. Conversely, a higher proportion of peripheral NK cells was associated with poorer pathologic CR in breast cancer.²⁸ Notably, the latter study demonstrated that patients with higher peripheral CD57+ NK cells had lower numbers of tumor-infiltrating NK cells, similar to our study. A meta-analysis of the prognostic value of NK cells in HCC, pooling data from 12 studies totaling 1,611 patients, suggested that high NK cell numbers predict improved disease-free survival; however, this finding was driven primarily by tumor-infiltrating NK cells.²⁹ Results need to be extrapolated carefully given data originate from different cancers, or from patients with HCC for whom treatment data are not available. The expression of PD-L1 and CD8 did not influence outcomes as previously shown in other studies,^9,30 although the impact of CD8+ cell burden remains controversial.

With regard to cytokine data, we did not identify a significant correlation of serum HGF, IL-6, and IL-10 with survival outcomes in our study, unlike earlier reports.^31,32 Differences between the cohorts in disease extent and treatment factors may contribute to the discrepancies. The present study included patients with a high burden of treatment-refractory HCC, in contrast to prior studies of earlier stage HCC. Despite a lack of correlation with survival, we found higher levels of HGF, IP-10 (CXCL10) and eotaxin in patients with MVI. In HCC, little is known about the role of IP-10, which is involved in angiogenesis and pro-inflammatory responses, but some early studies suggest that it may be elevated in patients with HCC over healthy controls, patients with hepatitis B or cirrhosis without HCC.³³ CCL24, an eotaxin family member, has been shown to promote angiogenesis via the VEGFR2 pathway, and was predictive of poorer OS in a prior study.³⁴ Further studies will be required to ascertain the role of these cytokines and investigate their utility as biomarkers in HCC.

This study has several limitations. Firstly, it was closed early due to slow accrual as sorafenib was no longer the favored first-line option after study initiation; thus, results are hypothesis generating. It does however add to the emerging literature evaluating the impact of SBRT and systemic therapy in HCC, particularly in patients with MVI and as such, SBRT remains a tool in the locoregional therapy armamentarium.³⁵ Second, the findings of our study are limited by the small number of patients enrolled. Third, we were unable to provide a deep characterization of tissue immunophenotyping. Continued biomarker development will be necessary in HCC to select patients who are most likely to benefit from ICI combinations.

The addition of SBRT to pembrolizumab in the second-line setting for advanced HCC demonstrated a high ORR and encouraging OS in a very high-risk population, with no signal of increased toxicity with the combination. This suggests that combinations with newer ICI-containing regimens in the advanced HCC setting should consider the addition of locoregional therapies, particularly SBRT, which may have a specific benefit in patients with MVI.

Abbreviations

AE, adverse events; AFP, alpha-fetoprotein; BCLC, Barcelona Clinic Liver Cancer; GTV, gross tumor volume; ICI, immune checkpoint inhibitors; NK, natural killer; ORR, objective response rate; OS, overall survival; PBMCs, peripheral blood mononuclear cells; PFS, progression-free survival; PR, partial response; PTV, planning target volume; PVT, portal vein thrombus; SBRT, stereotactic body radiation therapy; TACE, transarterial chemoembolization; TIL, tumor-infiltrating lymphocytes; TRAE, treatment-related adverse events.

Financial support

The primary funder of the study (Merck) did not participate in study design, data collection, data analysis, data interpretation, and writing of the report. The study was supported in part by a research grant from Investigator-Initiated Studies Program of Merck Canada Inc. The opinions expressed in this paper are those of the authors and do not necessarily represent those of Merck Canada Inc. Correlative analyses were supported by the Agnico Eagle Beyond Chemotherapy High Definition Therapeutics Pillar Grand Challenge, Princess Margaret Cancer Centre Foundation.

Authors’ contributions

Grainne M O’Kane: conceptualization, methodology, investigation, data curation, writing original draft, writing review and editing; Aruz Mesci: conceptualization, methodology, investigation, data curation, writing original draft, writing review and editing; Raymond Jang: investigation, data curation, writing original draft, writing review; Aisling Barry: investigation, data curation, writing original draft, writing review and editing; Cynthia M Bocaya: investigation, data curation, writing original draft, writing review and editing; Giselle M Boukhaled: methodology, investigation, data curation, writing original draft, writing review and editing; Holly Acton: investigation, data curation, writing original draft, writing review and editing; David Doddington: investigation, data curation, writing original draft, writing review and editing; Rowena Rodrigo: investigation, data curation, writing original draft, writing review and editing; Leo Chan: investigation, data curation, writing original draft, writing review and editing; Babak Noamani: investigation, data curation, writing original draft, writing review and editing; Harry Harvey: data curation, writing original draft, writing review and editing; Andrew Elia: investigation, data curation, writing original draft, writing review and editing; Rhoda Law: investigation, data curation, writing original draft, writing review and editing; Rebecca Prince: investigation, data curation, writing original draft, writing review and editing; Mark Doherty: conceptualization, investigation, writing original draft, writing review and editing; Anna Dodd: project management, conceptualization writing original draft, writing review and editing; David K. Wong: investigation, data curation, writing original draft, writing review and editing; Lisa Wang: investigation, data curation, formal analysis writing original draft, writing review and editing; Ben X Wang: methodology, investigation, resources, data curation, writing original draft, writing review and editing; Eric X Chen: Investigation, data curation, writing original draft, writing review and editing; Laura A Dawson: conceptualization, methodology, investigation, data curation, writing original draft, writing review and editing, supervision; Jennifer J Knox: conceptualization, methodology, investigation, data curation, writing original draft, writing review and editing, supervision.

Data availability

Clinical data supporting this manuscript are found in the supplementary data. Additional information is available from the corresponding author upon reasonable request.

Role of the funding source

The primary funder of the study (Merck) did not participate in study design, data collection, data analysis, data interpretation, and writing of the report. The study was supported in part by a research grant from Investigator-Initiated Studies Program of Merck Canada Inc. The opinions expressed in this paper are those of the authors and do not necessarily represent those of Merck Canada Inc. Correlative analyses were supported by the Agnico Eagle Beyond Chemotherapy High Definition Therapeutics Pillar Grand Challenge, Princess Margaret Cancer Centre Foundation.

Conflicts of interest

GOK reports research/grant support from AstraZeneca and Roche, Honoraria and consulting: Roche, AstraZeneca, Servier, Incyte, MSD, travel expenses from Novartis, Takeda, MSD, AstraZeneca. RJ declares advisory roles from BMS and Merck and has received research funding from Merck, Astra Zeneca and Camurus. MD has received grant/research support from AstraZeneca, honoraria and consulting fees from Roche, AstraZeneca, Takeda, Novartis, Ipsen and travel from AstraZeneca, Roche, Ipsen, Takeda, Janssen and MSD. BXW has received honoraria and/or consulting fees from AstraZeneca, Providence Therapeutics and Tessa therapeutics. EXC has participated in advisory boards of AstraZeneca, Eisai, Ipsen, Merck, Roche and BMS and in clinical trials sponsored by AstraZeneca, BMS, Roche, Merck and Iterion. LAD has received research grants from Merck Elekta Varian and consulting fees from Elekta and AstraZeneca. JK has received research grants from AstraZeneca, Roche, Merck, Ipsen, And consulting fees from Incyte AstraZeneca, Roche, Merck, Ipsen, Jazz Pharmaceutics and Nucana.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article: