Novel Transforming Growth Factor Beta Receptor I Kinase Inhibitor Galunisertib (LY2157299) in Advanced Hepatocellular Carcinoma
Author list: Sandrine Faivre,1 Armando Santoro,2 Robin K. Kelley,3 Ed Gane,4 Charlotte E. Costentin,5 Ivelina Gueorguieva6, Claire Smith6, Ann Cleverly,6 Michael M. Lahn,7 Eric Raymond,8 Karim A. Benhadji,7 Gianluigi Giannelli9
Affiliation list:
1Hopital Beaujon, Clichy, France
2Istituto Clinico Humanitas, Humanitas University, Rozzano, Italy 3University of California, San Francisco, California, USA 4Auckland City Hospital, Auckland, New Zealand
5Hospital Henri Mondor, Creteil, France
6Lilly Research Centre Erl Wood Manor, Windlesham, Surrey, UK
7Eli Lilly and Company, Indianapolis, Indiana, USA
8Centre Hospitalier Paris Saint-Joseph, Paris, France
9National Institute of Gastroenterology, Research Instituts “S. De Bellis” Research Hospital, Castellana Grotte, Bari, Italy

Corresponding author:
Prof Sandrine Faivre
Oncologie Médicale/Hopital Beaujon 100 Boulevard du Général Leclerc 92110 Clichy, France
+33 140 875614
[email protected]

Word count: Abstract 249; Introduction-Discussion 3608

Figures and tables: 4 in main manuscript; 10 in online supplement
Abbreviations: AFP, alpha fetoprotein; EMT, epithelial-mesenchymal transition; HCC, hepatocellular carcinoma; OS, overall survival; TGF-β1, transforming growth factor beta 1; TTP, time-to-tumor
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1111/liv.14113

Accepted Article
progression
Competing interests: IG and KB are employees and stockholders of Eli Lilly and Company. AC, ML, and CS are former employees and stockholders of Eli Lilly and Company. AS received fees as speaker or for Advisory boards from: Eli Lilly, Novartis, Roche, Servier, BMS, Gilead, Pfizer, Astra Zeneca, Bayer, Sandoz, Eisai, Takeda, MSD, and as consultant for Arqule. SF received consulting and research funding from BeiGene, Blueprint Medicine, Bristol-Myers Squibb, Bayer Pharma, Eli Lilly, Incyte, Ipsen, Merck Serono, MSD, and Novartis. RKK received research support to institution from Adaptimmune, Agios, AstraZeneca, Bayer, BMS, Celgene, Exelixis, Eli Lilly and Company, Novartis, QED, Taiho, Merck, and Medimmune, and is an advisor to Genentech/Roche.

Financial support: Funded by Eli Lilly and Company

Trial registration number: NCT01246986 at ClinicalTrials.gov

Abstract
Background and aims: We assessed the activity of galunisertib, a small molecule inhibitor of the transforming growth factor beta (TGF-β1) receptor I, in second line patients with hepatocellular carcinoma (HCC) in two cohorts of baseline serum alpha fetoprotein (AFP).

Methods: Patients with advanced HCC who progressed on or were ineligible to receive sorafenib, Child-Pugh A/B7, and ECOG PS ≤1 were enrolled into Part A (AFP≥1.5x ULN) or Part B (AFP<1.5x ULN). Patients were treated with 80 or 150 mg galunisertib BID for 14 days per 28-day cycle.
Endpoints were time-to-progression (TTP) and changes in circulating AFP and TGF-β1 levels, as well as. safety, pharmacokinetics, progression-free survival, and overall survival (OS).
Results: Patients (n=149) were enrolled with median age 65 years. Median TTP was 2.7 months (95% CI: 1.5-2.9) in Part A (n=109) and 4.2 months (95% CI: 1.7-5.5) in Part B (n=40). Median OS was 7.3 months (95% CI: 4.9-10.5) in Part A and 16.8 months (95% CI: 10.5-24.4) in Part B. OS was longer in AFP responders (>20% decrease from baseline, Part A) compared to non-responders (21.5 months versus 6.8 months). OS was longer in TGF-β1 responders (>20% decrease from baseline, all patients) compared to non-responders. The most common Grade 3/4 treatment-related adverse events were neutropenia (n=4) and fatigue, anaemia, increased bilirubin, hypoalbuminemia, and embolism (each, n=2).

Conclusions: Galunisertib treatment had a manageable safety profile in patients with HCC. Lower baseline AFP and a response in AFP or TGF-β1 levels (versus no response) correlated with longer survival.

Key words: hepatocellular carcinoma, TGF-β1 receptor I inhibitor, galunisertib, alpha fetoprotein, TGF-β1, liver cancer

Lay Summary:
Galunisertib, a novel targeted anti-cancer treatment, was tested in patients with hepatocellular carcinoma who progressed on or were ineligible to receive sorafenib. Patients were enrolled into one of two parts based on low or high blood levels of the biomarker alpha fetoprotein (AFP). Overall

Accepted Article
survival was longer for patients with lower levels of AFP at study entry, and for those with lower blood levels of another biomarker, TGF-β1, at study entry. Those who had a reduction of >20% of AFP or TGF-β1 in the first 6 cycles of treatment also had longer survival.

Introduction
Hepatocellular carcinoma (HCC) represents the sixth most common cancer worldwide, with increasing incidence due to the current epidemics of HCV and non-alcoholic fatty liver disease (1). Most patients with HCC have established cirrhosis, and their therapeutic options and survival are determined by the severity of underlying liver disease (1). Patients with advanced HCC are not suitable for curative therapeutic options such as liver transplantation, resection, tumor chemoembolization, or percutaneous or open ablation (2). Patients with preserved performance status and liver function can be offered sorafenib as first line treatment (3) or lenvatinib (4); however, median overall survival (OS) with sorafenib is approximately 12 months (5, 6). Most patients experience disease progression after sorafenib treatment and need second-line treatment, options for which include regorafenib, cabozantinib, PD-1-based therapies, or participation in a clinical trial (7-12). Ramucirumab has demonstrated survival benefit in second-line patients with alpha fetoprotein (AFP) >400 ng/mL (13). The median overall survival for patients receiving second line therapy for advanced HCC ranges from 7 to 13 months in phase 2/3 studies (10, 14, 15).

The transforming growth factor beta (TGF-β) signaling pathway has been shown to be active in hepatocarcinogenesis. TGF-β signaling contributed to the epithelial-mesenchymal transition (EMT) in HCC models (16, 17), and TGF-β inhibitors showed promise in several preclinical in vitro and in vivo studies (16-18). TGF-β1 was elevated in plasma and tumors of patients with HCC (19), and a subgroup of HCC tumors demonstrated the presence of TGF-β-associated gene expression signatures (20).

The presence of elevated circulating AFP levels in patients with HCC may be important, as this patient group is thought to be enriched for TGF-β-associated EMT (21), cancer-initiating stem cells activated by TGF-β signaling (22), and high stem cell proliferation. Elevated AFP in this population is also associated with poor prognosis (23). Although there is no direct correlation between AFP and EMT markers such as E-cadherin, recent reports suggest that indirect regulation of EMT, such as Cripto-1 expression, may be correlated with AFP (24).
Galunisertib (LY2157299) is a small molecule that selectively inhibits the serine/threonine kinase of the TGF-β receptor type I (TGF-β RI/ALK5) (25-27). Galunisertib had an acceptable safety profile in phase 1 studies and may be well tolerated in patients with HCC (28). The present study was designed to test the safety and activity of galunisertib in patients with HCC, and to assess the prognostic value of baseline circulating AFP. Pharmacokinetics (PK) and clinical parameters for antitumor activity and biomarkers related to TGF-β signaling were investigated.

Accepted Article
Methods
Patients and study design
This was a Phase 2, open-label, multicenter study of galunisertib in patients with HCC and Child-Pugh A 5/6 or B7. The study design consisted of four parts (A, B, C and D); Parts A and B are reported here (Fig. S1). Patients aged ≥18 years with histological evidence of a diagnosis of HCC (not amenable to curative surgery), Child-Pugh Class A or B7, measurable disease as defined by the
RECIST 1.1, adequate organ function, ECOG PS ≤1, and who received sorafenib and had progressed, or were ineligible for sorafenib treatment (who received other first line treatments or with co- morbidities including but not limited to high blood pressure, skin or mouth problems, risk for bowl perforation, or QT elongation), were included in the study. Patients who had received >1 line of previous systemic treatment were excluded. The study was conducted according to the principles of good clinical practice, applicable laws and regulations, and the Declaration of Helsinki. Each
institution’s review board approved the study (Table S5) and all patients signed an informed consent document before study participation.
The study had 2 parts (A and B) based on baseline serum AFP levels. Patients with AFP ≥1.5x upper limit of normal (ULN) were enrolled in Part A and randomly assigned to receive one of two doses of galunisertib (80 or 150 mg) given twice daily on 14 consecutive days followed by 14 days rest in a 28-day cycle (intermittent dosing). Dose cohorts in Part A were balanced based on AFP
levels (≤400 ng/mL and >400 ng/mL), liver disease etiology, and whether sorafenib-naïve or not.

The anticipated optimal dose of galunisertib was 150 mg BID based on phase 1 studies (28). Because this study evaluated galunisertib for the first time in patients with HCC, the first dose level was lowered to 80 mg BID to allow a greater margin of safety but possibly reduced effectiveness (26). Based on higher PK exposure, comparison of clinical data, review of responses, and acceptable safety, the decision was made to drop the 80 mg dose and to enroll future patients at 150 mg BID until 70 patients had been treated. Therefore, patients enrolled into Part B (AFP <1.5 x ULN) were treated with 150 mg BID with intermittent dosing.
Objectives/procedures
The primary objective was to characterize the time-to-tumor progression (TTP) and the effect of treatment on plasma TGF-β1 and serum AFP. Secondary objectives were to evaluate safety and population PK, and to determine the recommended dose of galunisertib for future trials. Other secondary objectives were progression-free survival (PFS), OS, and response rates.
Pharmacokinetic/biomarker methods
Plasma samples were analyzed for galunisertib using validated LC-MS/MS methods (BPLY215A and BPLY215B). The lower and upper limits of quantification were 0.050 ng/mL and 10.0 ng/mL for BPLY215A and 5.00 ng/mL and 1000 ng/mL for BPLY215B.

Plasma samples were analyzed for AFP and TGF-β1 levels by enzyme-linked immunosorbent assay (ELISA) (R&D Systems, DB100B, Minneapolis, MN). AFP was initially assessed at the institutional laboratory for the purpose of including patients but only measurements from the central laboratory were used in statistical analyses. The normal value for AFP provided by the central laboratory was 0.0–8.3 ng/mL. For TGF-β1, we established a level of approximately 2.0 ng/mL as

Accepted Article
normal for people without cancer (29).

Statistical analyses
A sample size of 147 (107 for Part A and 40 for Part B) was sufficient to characterize the TTP distributions and the effect on TGF-β1 associated biomarkers of the 2 doses of galunisertib in patients with HCC with elevated AFP levels. The number for each cohort of approximately 40 patients was based on prior single-arm phase 2 studies in 2nd line patients with HCC to allow comparison to historic data and also to describe the PK profile of galunisertib in hepatic-impaired patients. In addition, the sample size was adequate to assess anti-tumor activity in patients with normal AFP levels dosed at 150 mg BID.

All patients who received at least 1 dose of study drug were evaluated for safety, efficacy, and PK/PD endpoints. Time-to-tumor progression (TTP) was measured from the date of first dose to the first date of objective progression of disease. Patients who had not progressed, or died from any cause as of the data cut-off date (May 15, 2015), were censored at the last tumor assessment prior to the cut-off date or at the date of first dose if no post-baseline tumor assessment had occurred.
PFS was measured from the date of first dose to the first date of progression of disease or of death from any cause. Duration of tumor response was measured from the date of tumor response to the date of objective progression or death prior to receiving any post-discontinuation anticancer treatment. Tumors were assessed using RECIST v1.1. Time-to-event endpoints were summarized descriptively using the Kaplan-Meier method and compared using the log rank test. A Cox regression model including baseline biomarker levels (split at the median for TGF-β1 and at 400 µg/L for AFP) was used for correlations of biomarker status with OS. All hazard ratios (HR) were estimated using Cox regression models.

Analyses of AFP and TGF-β1 kinetics included patients who had at least one post dose measurement. A patient was considered a biomarker responder if they had more than a 20% reduction from baseline within the first 6 cycles of treatment. Only patients in Part A were evaluated for AFP response. Biomarker responses were summarized descriptively and their relationship with OS was estimated using the Kaplan-Meier method.
Comparison to REACH 1 trial placebo group
In order to better estimate the probability of success for subsequent trials with galunisertib, we were able to access patient level data for patients on the placebo arm from a contemporary, randomized, placebo-controlled trial of the antiangiogenic agent, ramucirumab. REACH 1 enrolled a similar population of patients with advanced HCC following first-line therapy with sorafenib (12).
Using patient level data and excluding most patients from East Asia from the ramucirumab study (as our study had few Asians) allowed a better comparison of OS against placebo than using summary statistics from historical data from the literature. To be consistent with the current study, patients from the REACH 1 placebo arm were grouped based on their baseline AFP levels and matched to Part A (baseline AFP≥1.5x ULN) or Part B (baseline AFP<1.5x ULN) (Tables S2, S3).

Accepted Article
Results
Patients
A total of 149 patients received study treatment, 109 in Part A (37 at 80 mg BID and 72 at 150 mg BID) and 40 in Part B (150 mg BID); 144 patients discontinued study treatment (108 in Part A and 36 in Part B), and five patients (1 in Part A and 4 in Part B) were continuing to receive study treatment at the time of the data cut-off for this report. The most common reasons for treatment discontinuation were progressive disease (98 patients: 73 in Part A and 25 in Part B), death (20 patients: 18 in Part A and 2 in Part B), and adverse events (AE, 11 patients: 7 in Part A and 4 in Part B). Complete patient disposition is shown in Figure S2. Baseline demographic and clinical characteristics are given in Table 1. In Part B and after an amendment, patients with Child-Pugh B7 were allowed into the study.

Primary and secondary objectives
Median TTP was longer in Part B (4.2 months [95%CI: 1.7-5.5]) compared to either dose cohort of Part A (overall 2.7 months [95%CI: 1.5-2.9]) (Table 2, Fig. 1A-1B). The hazard ratio (HR) for TTP for Part A 150 mg BID versus Part A 80 mg BID was 1.2 (95% CI: 0.7-2.0) and the HR for TTP for Part A 150 mg BID versus Part B 150 mg BID was 1.3 (95% CI: 0.9-2.1).

Median PFS was 2.77 months (95% CI: 1.5-2.9) in the Part A 80 mg BID cohort, 1.50 months (95%CI: 1.4-2.6) in the Part A 150 mg BID cohort and 3.1 months (95% CI: 1.5-5.5) in the Part B 150 mg BID cohort. The HR for PFS for Part A 150 mg BID versus Part A 80 mg BID was 1.1 (95% CI: 0.7- 1.7) and the HR for PFS for Part A 150 mg BID versus Part B 150 mg BID was 1.5 (95% CI: 1.0-2.2).

Median OS in Part B (16.8 months [95%CI: 10.4-24.1]) was longer compared to either dose cohort of Part A (overall 7.3 months [95%CI: 4.9-10.5]) (Table 2, Fig. 2A-2B). The HR for OS for Part A 150 mg BID versus Part A 80 mg BID was 1.1 (95% CI: 0.7-1.7) and the HR for OS for Part A 150 mg BID versus Part B 150 mg BID was 2.1 (95% CI: 1.3-3.3).

Three patients (2.0%) had an objective tumor response based on investigator review using RECIST v1.1, no patients in the Part A of the study, and 3 (7.5%) patients in the Part B (Fig. S3).
Pharmacokinetics (PK) and pharmacodynamics (PD)
The combined PK data from Parts A and B comprised 448 observations from 143 patients. Galunisertib plasma concentration profiles at steady state (Days 14, 15 and 16) were similar for the two doses (80 mg BID and 150 mg BID) in Part A, with a general trend for higher concentrations with the 150 mg BID dose (Fig. S4).

PD response for all biomarkers was defined as a reduction of more than 20% from baseline within the first 6 cycles of treatment. An AFP response was observed in 22/103 patients (21.4%) in Part A, and the response rates were similar between the two dose cohorts (Fig. S3A-B). Patients with an AFP response (4.3 months [95%CI: 2.9-15.2]) had a longer TTP compared to AFP non-responders (1.5 months [95%CI: 1.4-2.8]; log rank P<0.0001) (Table 2, Fig. 1C). Similarly, OS was longer in AFP responders (21.5 months [95%CI: 6.8-25.1]) compared to non-responders (6.8 months [95%CI: 4.5- 8.9]; log rank P=0.0012) (Table 2, Fig. 2C).

Accepted Article
A TGF-β1 response was observed in 78/142 (54.9%) of patients overall; this included 48.5% (50/103) of patients Part A and 71% (28/39) of patients in Part B (Table 2). In Part A, the median TTP was 3.2 months (95%CI: 1.5-4.3) for TGF-β1 responders and 1.5 months (95%CI: 1.4-2.8) for non- responders. In Part B, the median TTP was 4.2 months for both TGF-β1 responders and non- responders. In Part A, the median OS was 11.2 months (95%CI: 6.8-14.5) for TGF-β1 responders and
5.3 months (95%CI: 3.0-8.9) for non-responders. In Part B, the median OS was 21.9 months (95%CI: 12.4-NE) for TGF-β1 responders, and 10.5 months (95%CI: 1.5-16.5) for non-responders. These data represent about a 2-fold difference in OS for TGF-β1 responders in both study parts, allowing the effect of TGF-β1 response to be assessed in all patients combined (Fig. 1D-1E, Fig. 2D-2E).

The number of patients receiving post-discontinuation anticancer therapies was 41 (29%) and the proportion was similar for patients with or without TGF-β1 response. The most common post-discontinuation drug used was sorafenib (10%) followed by oxaliplatin (8%); proportions were similar in Part A and Part B. On average, time on either agent was no longer than 1 cycle. Thus, the relationship of TGF-β1 response to OS was unlikely to be impacted by post-discontinuation therapies.

Comparison to REACH 1 trial placebo group

A comparison of the OS data for Parts A and B of this study, and the matched placebo groups from the REACH 1 study, is shown in Table S4. For patients in Part A, median OS was 7.3 months, and those from the placebo arm of REACH 1 had a median OS of 5.4 months (95% CI: 4.5-7.3) (log rank p=0.16; HR 0.81: 95%CI (0.61-1.08). For patients in Part B, median OS was 16.8 months, compared to the placebo cohort of REACH 1 with a median OS of 18.4 months (log rank p=0.77) (Fig. S5A-S5B).

A similar percentage of Part A patients were classified as AFP responders in both the galunisertib-treated (22/103; 21%) and the matched REACH 1 placebo (14/85; 16%) groups. We recognize that the low numbers of patients in this comparison, together with the high censor rate for the placebo group in particular, makes any interpretation difficult. However, for completeness, the observed results of the Kaplan-Meier analysis are presented (Table S4, Fig. S5C-S5D).
Safety
For Parts A and B combined, the median number of cycles per patient was 3.0 (range: 1–40). The median number of cycles initiated was higher in Part B (5 cycles) than in Part A (3 and 2 cycles for the 80 mg BID and 150 mg BID cohorts, respectively). At the time of data cut-off, 122 patients (81.9%) had died, 96 (88.1%) in Part A and 26 (65.0%) in Part B. Most deaths were due to disease progression (110 in total; 84 in Part A and 26 in Part B) and the remaining 12 deaths (11 on study treatment and one post study discontinuation) in Part A were due to an adverse event (AE, 5 in the 80 mg BID cohort and 7 in the 150 mg BID cohort). Two of the AEs leading to death in Part A were considered by the investigator to be related to study treatment: acute myocardial infarction in 1 patient in the 80 mg BID cohort and lobar pneumonia in 1 patient in the 150 mg BID cohort.

Accepted Article
Across the study, 65 (43.6%) patients had at least one serious AE (SAE). There was a slightly higher percentage SAEs in the Part A 150 mg BID cohort (48.6%) compared to the Part A 80 mg BID cohort (40.5%), or Part B (150 mg BID, 37.5%). The most frequently occurring SAEs (4% of patients) were (n, %) anemia (12, 8.1%), ascites (6, 4.0%), and encephalopathy (6, 4.0%). There were no notable differences between dose cohorts in the incidence of any individual AE, and the only classification for adverse events (by the Medical Dictionary for Regulatory Activities [MedDRA]) by system order class reported in >10% of patients was, as expected, gastrointestinal disorders (17%).

The majority of patients (134, 90%) reported at least one treatment-emergent AE (TEAE).
The most frequently occurring TEAEs (>20% of patients) overall were fatigue (33.6%), anemia (25.5%), edema peripheral (22.8%) and abdominal pain (21.5%) (Table S1). There were no notable differences between dose cohorts.

A total of 11 patients discontinued study treatment due to an AE, 7 (6.4%) in Part A (1 [2.7%] in the 80 mg BID cohort, 6 [8.3%] in the 150 mg BID cohort), and 4 (10.0%) in Part B. Of these 11, 8 were considered study drug related (asthenia, blood bilirubin increased, gallbladder perforation, hematemesis, left ventricular dysfunction, peripheral edema, pancreatitis and peritoneal hemorrhage).

Discussion
The present study evaluated the anti-tumor effect of the novel TGF-bR1/ALK5 inhibitor galunisertib in patients with HCC with high versus low levels of circulating AFP. Higher baseline AFP levels were associated with worse outcomes; patients enrolled in Part A had a substantially lower OS compared to patients enrolled in Part B (7.25 versus 16.8 months). Serum AFP alone was a strong prognostic marker, considering that demographic factors were not significantly different between the two parts. Whether AFP levels are the sole prognostic marker for OS remains unclear, but AFP remains a component of prognostic scoring systems, such as CLIP (30).

More recently, ramucirumab was evaluated in patients with HCC and AFP levels of ≥400 ng/mL at baseline, leading to an OS of 8.5 months (13). While in the present study the OS was with
7.3 months more similar to prior 2nd line study OS numbers or their placebo group, we cannot compare the 2 studies due to the unavailability of the demographic information for patients in the REACH-2 study. All other studies, including those utilizing PD-1 inhibitors or cabozantinib have not selected patients based on their baseline AFP or TGF-β1 levels (10). The subgroups in these trials (eg, by AFP baseline levels) appear to be small, or the study did not measure TGF-β1 levels.

We anticipated that patients benefiting from galunisertib would manifest a decrease in serum AFP, as previously observed with chemotherapy or sorafenib (31). In Part A, the 22 (21.4%) patients who showed an AFP response had a longer OS compared to those patients who did not show an AFP response. These results were similar using a different cut-off (>50% decrease from baseline).

Higher TGF-β1 levels at baseline were also correlated with shorter OS in this study, but patients who had a TGF-β1 response (>20% decrease from baseline) lived significantly longer. In Parts A and B patients together, more than 50% of patients had a TGF-β1 response, and those patients had an approximately two-fold longer OS compared to patients who did not have a TGF-β1 response. TGF-β1 levels have been previously reported to be prognostic for response to treatment in HCC (32, 33). Serum TGF-β1 levels were correlated with OS in a cohort of Asian HCC patients treated with sorafenib with similar demographics to our population (33). Hence, plasma TGF-β1 response appears to be a useful marker to assess clinical benefit in second line patients during treatment with a TGF-β inhibitor.

Other biomarker-based approaches were used in trials with patients with HCC. Recently the study with the kinase inhibitor cabozantinib (9) yielded median OS benefits similar to those reported here for patients who had a response in either AFP or TGF-β1 levels (34). Also, the study REACH-2 enrolled patients with elevated AFP levels to determine the effect of ramucirumab in patients with HCC (13). But comparison of our study with these studies was difficult; hence, we accessed internal data from the REACH-1 study to obtain a comparative placebo group. By selecting patients to match the demographic features of patients enrolled in this study, we were able to observe that a placebo group would have resulted in median OS of 5.4 months compared to 7.3 months for galunisertib.
While statistically not significant for the entire population, the patients taking galunisertib who responded by AFP reduction >20% had a significantly longer OS (21.5 months) compared to the matched placebo group with an AFP reduction >20% (12.1 months). Importantly, this effect was not evident in the population with normal baseline levels of AFP. Such a comparison with a well-defined and well-characterized study dataset is helpful for future designs, as recently demonstrated by the results with ramucirumab in patients with HCC (REACH-2) (13).

Safety

The safety profile of galunisertib in HCC was relatively benign compared to other treatment

Accepted Article
options, including sorafenib. We observed no hand foot syndrome or worsening of liver function. Most of the AEs were related to tumor progression or to underlying liver disease such as bleeding in patients with cirrhosis.

Conclusions
We demonstrated that the TGF-β RI/ALK5 inhibitor galunisertib was well tolerated at 150 mg BID in an intermittent dosing schedule. The selection of the 150 mg BID dose over the 80 mg BID dose was based on a review of the clinical data, higher exposure to galunisertib, and similar efficacy. Elevated AFP at baseline was associated with worse prognosis, while AFP response during treatment was associated with longer survival. These results encourage further investigation of TGF-β pathway inhibition in patients with HCC.

Accepted Article
Recent data suggests that patients with exhausted immune responses may benefit from a combined treatment of PD-1/PD-L1 and TGF-β RI blockade (35). Hence, a small safety study was started to assess the safety of the combination of galunisertib and nivolumab (NCT02423343).

Acknowledgments
We thank the patients and their caregivers for participating in this trial. We also thank the investigators and their support staff who generously participated in this work and the trial personnel. Writing assistance was provided by Caryl J. Antalis (Eli Lilly and Company) and Ananya Biswas (Eli Lilly Services India Pvt. Ltd., India). Parts of this data were previously presented at the following congresses: European Association for the Study of Liver, April 9-13, 2014; Korean Cancer Association, June 19-20, 2014; Asia-Pacific Primary Liver Cancer Expert 2014; AACR Annual Meeting, 2016; ASCO Annual Meeting 2016; ILCA 2016.

Contributions of the authors
KB designed the experiments, analyzed data, interpreted results, and wrote the manuscript. GG and SF designed the study, collected data, interpreted results, and were involved in the drafting and editing of the manuscript. AS, RKK, EG, and ER were involved in the acquisition, analysis and interpretation of the data, and critical revision of the manuscript. CS and AC analysed and interpreted the data, drafted and reviewed the manuscript. ML developed the concept of the clinical study, designed the experiment/study, designed the biomarker research plan, collected data, reviewed data, drafted, edited, and revised the manuscript. All authors approved the manuscript for publication.

References
⦁ Mittal S, El-Serag HB. Epidemiology of hepatocellular carcinoma: consider the population. J Clin Gastroenterol. 2013;47 Suppl:S2-6.

⦁ Kudo M, Trevisani F, Abou-Alfa GK, Rimassa L. Hepatocellular Carcinoma: Therapeutic Guidelines and Medical Treatment. Liver Cancer. 2016;6(1):16-26.

⦁ Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med. 2008;359(4):378-390.

⦁ Kudo M, Finn RS, Qin S, Han KH, Ikeda K, Piscaglia F, et al. Lenvatinib versus sorafenib in first- line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial. Lancet. 2018;391(10126):1163-1173.

⦁ Marrero JA, Lencioni R, Ye S-L, Kudo M, Bronowicki J-P, Chen X-P, et al. Final analysis of GIDEON (Global Investigation of Therapeutic Decisions in Hepatocellular Carcinoma [HCC] and of Its Treatment with Sorafenib [Sor]) in >3000 Sor-treated patients (pts): Clinical

Accepted Article
findings in pts with liver dysfunction. 2013 ASCO Annual Meeting, 2013. Chicago. ASCOJ Clin Oncol 31, 2013 (suppl; abstr 4126).

⦁ Lencioni R, Kudo M, Ye SL, Bronowicki JP, Chen XP, Dagher L, et al. GIDEON (Global Investigation of therapeutic DEcisions in hepatocellular carcinoma and Of its treatment with sorafeNib): second interim analysis. Int J Clin Pract. 2014;68(5):609-617.

⦁ Bruix J, Qin S, Merle P, Granito A, Huang YH, Bodoky G, et al. Regorafenib for patients with hepatocellular carcinoma who progressed on sorafenib treatment (RESORCE): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet. 2017;389(10064):56-66.

⦁ Woo HY, Yoo SY, Heo J. New chemical treatment options in second-line hepatocellular carcinoma: what to do when sorafenib fails? Expert Opin Pharmacother. 2017;18(1):35-44.

⦁ Abou-Alfa GK, Meyer T, Cheng A-L, El-Khoueiry AB, Rimassa L, Ryoo B-Y, et al. Cabozantinib in Patients with Advanced and Progressing Hepatocellular Carcinoma. New England Journal of Medicine. 2018;379(1):54-63.

⦁ Vogel A, Cervantes A, Chau I, Daniele B, Llovet J, Meyer T, et al. Hepatocellular carcinoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2018;29(Supplement_4):iv238-iv255.

⦁ El-Khoueiry AB, Sangro B, Yau T, Crocenzi TS, Kudo M, Hsu C, et al. Nivolumab in patients with advanced hepatocellular carcinoma (CheckMate 040): an open-label, non-comparative, phase 1/2 dose escalation and expansion trial. The Lancet. 2017;389(10088):2492-2502.

⦁ Zhu AX, Park JO, Ryoo BY, Yen CJ, Poon R, Pastorelli D, et al. Ramucirumab versus placebo as second-line treatment in patients with advanced hepatocellular carcinoma following first- line therapy with sorafenib (REACH): a randomised, double-blind, multicentre, phase 3 trial. Lancet Oncol. 2015;16(7):859-870.

⦁ Zhu AX, Kang YK, Yen CJ, Finn RS, Galle PR, Llovet JM, et al. Ramucirumab after sorafenib in patients with advanced hepatocellular carcinoma and increased alpha-fetoprotein concentrations (REACH-2): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol. 2019;20(2):282-296.

⦁ Zhu AX, Kudo M, Assenat E, Cattan S, Kang YK, Lim HY, et al. Effect of everolimus on survival in advanced hepatocellular carcinoma after failure of sorafenib: the EVOLVE-1 randomized clinical trial. JAMA. 2014;312(1):57-67.

⦁ Llovet JM, Hernandez-Gea V. Hepatocellular carcinoma: reasons for phase III failure and novel perspectives on trial design. Clin Cancer Res. 2014;20(8):2072-2079.

⦁ Giannelli G, Bergamini C, Fransvea E, Sgarra C, Antonaci S. Laminin-5 with transforming growth factor-beta1 induces epithelial to mesenchymal transition in hepatocellular carcinoma. Gastroenterology. 2005;129(5):1375-1383.

⦁ Giannelli G, Mazzocca A, Fransvea E, Lahn M, Antonaci S. Inhibiting TGF-beta signaling in hepatocellular carcinoma. Biochim Biophys Acta. 2011;1815(2):214-223.

⦁ Serova M, Tijeras-Raballand A, Dos Santos C, Albuquerque M, Paradis V, Neuzillet C, et al. Effects of TGF-beta signalling inhibition with galunisertib (LY2157299) in hepatocellular carcinoma models and in ex vivo whole tumor tissue samples from patients. Oncotarget.

Accepted Article
2015;6(25):21614-21627.

⦁ Giannelli G, Fransvea E, Marinosci F, Bergamini C, Colucci S, Schiraldi O, et al. Transforming growth factor-beta1 triggers hepatocellular carcinoma invasiveness via alpha3beta1 integrin. Am J Pathol. 2002;161(1):183-193.

⦁ Hoshida Y, Nijman SM, Kobayashi M, Chan JA, Brunet JP, Chiang DY, et al. Integrative transcriptome analysis reveals common molecular subclasses of human hepatocellular carcinoma. Cancer Res. 2009;69(18):7385-7392.

⦁ Wendt MK, Allington TM, Schiemann WP. Mechanisms of the epithelial-mesenchymal transition by TGF-beta. Future Oncol. 2009;5(8):1145-1168.

⦁ Yao Z, Mishra L. Cancer stem cells and hepatocellular carcinoma. Cancer Biol Ther. 2009;8(18):1691-1698.

⦁ Raoul JL, Bruix J, Greten TF, Sherman M, Mazzaferro V, Hilgard P, et al. Relationship between baseline hepatic status and outcome, and effect of sorafenib on liver function: SHARP trial subanalyses. J Hepatol. 2012;56(5):1080-1088.

⦁ Wang JH, Wei W, Xu J, Guo ZX, Xiao CZ, Zhang YF, et al. Elevated expression of Cripto-1 correlates with poor prognosis in hepatocellular carcinoma. Oncotarget. 2015;6(33):35116- 35128.

⦁ Sawyer JS, Beight DW, Britt KS, Anderson BD, Campbell RM, Goodson T, Jr., et al. Synthesis and activity of new aryl- and heteroaryl-substituted 5,6-dihydro-4H-pyrrolo[1,2-b]pyrazole inhibitors of the transforming growth factor-beta type I receptor kinase domain. Bioorg Med Chem Lett. 2004;14(13):3581-3584.

⦁ Herbertz S, Sawyer JS, Stauber AJ, Gueorguieva I, Driscoll KE, Estrem ST, et al. Clinical development of galunisertib (LY2157299 monohydrate), a small molecule inhibitor of transforming growth factor-beta signaling pathway. Drug Des Devel Ther. 2015;9:4479-4499.

⦁ Yingling JM, McMillen WT, Yan L, Huang H, Sawyer JS, Graff J, et al. Preclinical assessment of galunisertib (LY2157299 monohydrate), a first-in-class transforming growth factor-beta receptor type I inhibitor. Oncotarget. 2018;9(6):6659-6677.

⦁ Rodon J, Carducci MA, Sepulveda-Sanchez JM, Azaro A, Calvo E, Seoane J, et al. First-in- Human Dose Study of the Novel Transforming Growth Factor-beta Receptor I Kinase Inhibitor LY2157299 Monohydrate in Patients with Advanced Cancer and Glioma. Clin Cancer Res. 2015;21(3):553-560.

⦁ Giles BM, Underwood TT, Benhadji KA, Nelson DKS, Grobeck LM, Lin B, et al. Analytical Characterization of an Enzyme-Linked Immunosorbent Assay for the Measurement of Transforming Growth Factor β1 in Human Plasma. The Journal of Applied Laboratory Medicine. 2018;3(2):200-212.

⦁ Kudo M, Chung H, Osaki Y. Prognostic staging system for hepatocellular carcinoma (CLIP score): its value and limitations, and a proposal for a new staging system, the Japan Integrated Staging Score (JIS score). J Gastroenterol. 2003;38(3):207-215.

⦁ Shao YY, Hsu CH, Cheng AL. Predictive biomarkers of sorafenib efficacy in advanced hepatocellular carcinoma: Are we getting there? World J Gastroenterol. 2015;21(36):10336-

Accepted Article
10347.

⦁ Casadei Gardini A, Scarpi E, Faloppi L, Scartozzi M, Silvestris N, Santini D, et al. Immune inflammation indicators and implication for immune modulation strategies in advanced hepatocellular carcinoma patients receiving sorafenib. Oncotarget. 2016;7(41):67142-67149.

⦁ Lin TH, Shao YY, Chan SY, Huang CY, Hsu CH, Cheng AL. High Serum Transforming Growth Factor-beta1 Levels Predict Outcome in Hepatocellular Carcinoma Patients Treated with Sorafenib. Clin Cancer Res. 2015;21(16):3678-3684.

⦁ Kelley RK, Rimassa L, Ryoo BY, Park JW, Blanc JF, Chan SL, et al. Alpha fetoprotein (AFP) response and efficacy outcomes in the phase III CELESTIAL trial of cabozantinib (C) versus placebo (P) in advanced hepatocellular carcinoma (HCC). J Clin Oncol. 2019;37(suppl 4; abstr 423)

⦁ Sia D, Jiao Y, Martinez-Quetglas I, Kuchuk O, Villacorta-Martin C, Castro de Moura M, et al. Identification of an Immune-specific Class of Hepatocellular Carcinoma, Based on Molecular Features. Gastroenterology. 2017;153(3):812-826.
Table 1. Demographics and baseline characteristics

Parameter All (n=149) Part A Part B
80 mg BID
(n=37) 150 mg BID
(n=72) 150 mg BID
(n=40)
Male, n (%) 127 (85.2) 32 (86.5) 59 (81.9) 36(90.0)
Age, median (range), yr 65 (31–89) 61 (31–85) 63 (33–87) 67.5 (46–89)
Ethnicity, n (%)
White (non-Asian) 127 (85.2) 34 (91.9) 58 (80.6) 35 (87.5)
Asian 9 (6.0) 1 (2.7) 7 (9.7) 1 (2.5)
Other 0 0 0 2 (5.0)
Missing 5 (3.4) 0 2 (2.8) 3 (7.5)
Prior sorafenib, n (%)
Yes 123 (82.6) 29 (78.4) 60 (83.3) 34 (85.0)
No* 26 (17.4) 8 (21.6) 12 (16.7) 6 (15.0)
Eastern Cooperative Oncology Group Performance Status, n (%)
0 84 (56.4) 25 (67.6) 35 (48.6) 24 (60.0)
1 65 (43.6) 12 (32.4) 37 (51.4) 16 (40.0)
Child Pugh score, n (%)
5 67 (45.0) 13 (35.1) 31 (43.1) 23 (57.5)
6 61 (40.9) 23 (62.2) 31(43.1) 7 (17.5)
7 21 (14.1) 1 ( 2.7) 10 (13.9) 10 (25.0)
Pre-existing liver disease, n (%)
Hepatitis C 36 (24.2) 10 (27.0) 19 (26.4) 7 (17.5)
Hepatitis B 30 (20.1) 10 (27.0) 14 (19.4) 6 (15.0)
Alcohol use 30 (20.1) 8 (21.6) 13 (18.1) 9 (22.5)
Steatosis (non-
alcoholic fatty liver disease) 13 (8.7) 3 (8.1) 5 (6.9) 5 (12.5)
Haemochromatosis 6 (4.0) 1 (2.7) 4 (5.6) 1 (2.5)
Other (no cause
provided) 20 (13.4) 4 (10.8) 8 (11.1) 8 (20.0)
Multiple 14 (9.4) 1 (2.7) 9 (12.5) 4 (10.0)

Liver transplant, n (%) 4 (2.7) 0 2 (2.8) 2 (5.0)
Portal vein thrombosis,
n (%)
Yes 39 (26.2) 8 (21.6) 21 (29.2) 10 (25.0)
No 109 (73.2) 29 (78.4) 51 (70.8) 29 (72.5)
Missing 1 (0.7) 0 0 1 (2.5)
Tumor morphology, n (%)
Unimodular 19 (12.8) 4 (10.8) 10 (13.9) 5 (12.5)
Mulitmodular 100 (67.1) 28 (75.7) 44 (61.1) 28 (70.0)
Massive 28 (18.8) 5 (13.5) 17 (23.6) 6 (15.0)
Missing 2 (1.3) 0 1 (1.4) 1 (2.5)
Baseline alpha fetoprotein
≥400 ng/mL 66 (44.3) 24 (64.9) 42 (58.3) 0
Missing 1 (0.7) 0 0 1 (2.5)
Accepted Article
*In patients who had not previously been treated with sorafenib, 5 did receive other first line treatments (3 patients received brivanib and 2 received investigational agents).

ccepted Article

Table 2. Time-to-tumor progression and overall survival by biomarker baseline status and response

n Part A (n=109)

Baseline AFP ≥1.5x ULN
n Part B (n=40)

Baseline AFP <1.5x ULN
Time-to-tumor progression (mo), median (95% CI)
109 All: 2.7 (1.5-2.9)
All 37 80 mg BID: 2.8 (1.4-4.2) 40 4.2 (1.7-5.5)
72 150 mg BID: 1.6 (1.5-3.2)
AFP responders
22 4.3 (2.9-15.2) NA
AFP non-responders 81 1.5 (1.4-2.8) NA
TGF-β1 responders 50 3.2 (1.5-4.3) 28 4.2 (1.5-6.5)
TGF-β1 non-responders 53 1.5 (1.4-2.8) 11 4.2 (1.4-6.9)
Overall survival (mo), median (95% CI)
109 All: 7.3 (4.9-10.5)
All 37 80 mg BID: 9.0 (5.4-12.1) 40 16.8 (10.4-24.1)
72 150 mg BID: 6.8 (4.2-10.6)
AFP responders 22 21.5 (6.8-25.1) NA

ccepted Article

AFP non-responders 81 6.8 (4.5-8.9) NA
TGF-β1 responders 50 11.2 (6.8-14.5) 28 21.9 (12.4-NE)
TGF-β1 non-responders 53 5.3 (3.0-8.9) 11 10.5 (1.5-16.5)

Biomarker response was defined as a decrease of >20% from baseline level.
AFP=alpha fetoprotein; mo=months; NE=not estimated; TGF-β1=transforming growth factor beta 1; ULN=upper limit of normal.

Accepted Article

Figure Legends
Figure 1. Time-to-tumor-progression (TTP). A. TTP for Part A. B. TTP for Part B. C. Part A, comparison of patients with (red) or without (blue) an AFP response. D. Part A, comparison of patients with (red) or without (blue) a TGF-β1 response. E. Part B, comparison of patients with (red) or without (blue) a TGF-β1 response.
Figure 2. Overall survival (OS). A. OS for Part A. B. OS for Part B. C. Part A,
comparison of OS for patients with (red) or without (blue) an AFP response. D. Part A, comparison of OS for patients with (red) or without (blue) a TGF-β1 response. E. Part B, comparison of OS for patients with (red) or without (blue) a TGF-β1 response.
Figure S1. Study design. AFP=alpha fetoprotein; BID=twice daily; ULN=upper limit of normal.
Figure S2. Patient disposition.
Figure S3. Change in tumor size by investigator determined radiographic responses (RECIST 1.1). A. Part A 80 mg BID dose cohort. B. Part A 150 mg BID dose cohort. C. Part B. PD=progressive disease; PR=partial response. Note: two patients with >30% reduction from baseline did not have PR on their confirmatory scan.
Figure S4. Biomarker changes from baseline. A. AFP maximum change from baseline for Part A. Patients on 80 mg BID are in red; those on 150 mg BID are in blue. B. TGF-β1 maximum change from baseline for Part A. Patients on 80 mg BID are in red; those on 150 mg BID are in blue. C. TGF-β1 maximum change from baseline for Part B.
Figure S5. Galunisertib plasma pharmacokinetics for patients in each dose cohort of Part A.
Figure S6. Overall survival comparison: Parts A and B and matched placebo arms from REACH 1. A. Galunisertib Part A versus REACH 1 matched placebo. B. Galunisertib Part B versus REACH 1 matched placebo. C. OS by AFP response: Galunisertib Part A versus REACH 1 matched cohort. D. OS by AFP response: Galunisertib Part B versus REACH 1
matched cohort.

Accepted Article

Accepted Article