Phase II Studies with Refametinib or Refametinib plus Sorafenib in Patients with RAS-Mutated Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death worldwide (1, 2), and the prognosis for HCC remains extremely poor (2, 3). The recommended standard of care in advanced HCC is treatment with the multikinase inhibitor sorafenib (3–5). Lenvatinib has been shown to be noninferior to sorafenib in a recent phase III trial (6). In second line, the multikinase inhibitor regorafenib has been approved after showing significantly improved survival versus placebo in patients who had disease progression on sorafenib (7). The kinase inhibitor cabozantinib has also demonstrated promising survival improvements versus placebo as second-line therapy in a phase III trial (8). Immunotherapy has shown promise in HCC, with the immune checkpoint inhibitor nivolumab recently approved for the second-line treatment of advanced HCC based on durable responses observed in a phase I/II trial (9). Treatment with the monoclonal antibody ramucirumab has shown survival improvement versus placebo in patients progressing to sorafenib with alpha-fetoprotein >400 ng/mL (10). Poor prognosis and a lack of treatment options highlight a need for additional viable treatment regimens in the advanced setting.

Refametinib (BAY 86-9766; Bayer AG) is an oral, potent, nonadenosine triphosphate competitive inhibitor targeting MEK 1 and 2 (11), which play a central role in the RAS signal transduction cascade. RAS–MAPK signaling has been implicated in tumor progression and dissemination in HCC (2). A phase I study of the combination of refametinib with sorafenib in patients with advanced malignancies including HCC demonstrated a favorable safety profile and pharmacokinetic profile at a maximum tolerated dose of refametinib 50 mg twice daily in combination with sorafenib 400 mg twice daily (12).

The analysis of cell-free circulating tumor DNA (ctDNA) using beads, emulsion, amplification, and magnetics technology (BEAMing; Sysmex Inostics GmbH) enables tumor genotyping at the time of treatment and offers a viable, noninvasive approach to identifying clinically relevant mutations (13, 14). BEAMing may therefore be a feasible tool to support the need for the identification of predictive biomarkers in HCC (3), through proof-of-concept studies. Previous proof-of-concept studies of kinase inhibitors in other cancer types have successfully detected predictive mutations, such as vemurafenib in patients with inoperable melanoma with a BRAF^V600 mutation (15) and crizotinib in patients with non–small-cell lung cancer with EML4–ALK fusion (16).

A retrospective analysis in a phase II study evaluating refametinib plus sorafenib in Asian patients with HCC found that patients with RAS mutations exhibited a robust clinical response compared with patients with wild-type RAS [objective tumor response rate (ORR): 3/4 patients (75.0%) compared with 1/65 patients (1.5%), respectively; ref. 17]. Here, we describe the first proof-of-concept studies based on mutations conducted in patients with HCC. Two phase II studies prospectively evaluated the efficacy of refametinib monotherapy (NCT01915589) and refametinib plus sorafenib (NCT01915602) in patients with unresectable or metastatic HCC with mutated RAS, as determined by BEAMing of ctDNA.

These were two phase II, prospective, single-arm, multicenter, uncontrolled, open-label studies. The primary objective was to evaluate the efficacy of refametinib alone or in combination with sorafenib in patients with RAS- (KRAS- or NRAS-) mutated unresectable or metastatic HCC. The primary efficacy variable was the central radiologic assessment of ORR [complete response (CR) plus partial response (PR)] according to modified Response Evaluation Criteria in Solid Tumors (mRECIST; ref. 18). The secondary objective was safety, and additional objectives included evaluation of biomarkers aiming to identify biomarkers or biomarker signatures which could correspond to therapy response. Secondary efficacy variables included centrally assessed ORR according to RECIST version 1.1, investigator-assessed ORR according to mRECIST and RECIST version 1.1, overall survival, disease control rate, time to radiographic tumor progression, duration of response, and progression-free survival.

Fifteen patients with RAS mutations were planned to be included in the first stage of each study. The second stage was to be initiated if five or more of these patients had a confirmed objective response according to mRECIST.

In the refametinib monotherapy study, the protocol was amended once, with changes implemented globally. Prior cytotoxic chemotherapy was added as an exclusion criterion to omit a population of overtreated patients who may have been different from patients conventionally treated with sorafenib; this change affected eligibility criteria for RAS mutation testing and treatment exclusion criteria. An exclusion criterion was also added regarding women of childbearing potential to reduce the time gap between the pregnancy evaluation and the beginning of treatment.

In the refametinib plus sorafenib study, the protocol was amended twice. Changes were implemented globally and included the following amendments: patients with a corrected QT interval >480 ms at the time of screening were excluded from the study because of the potential for QT prolongation with sorafenib; the exclusion criterion regarding women of childbearing potential was amended to reduce the time gap between the pregnancy evaluation and the beginning of treatment; the exclusion criterion regarding systemic anticancer therapy was clarified, as patients with prior systemic anticancer therapy were not eligible for this study. In addition, a dose-modification scheme for hepatotoxic events was included, because hepatotoxicity is an “identified risk” for the refametinib–sorafenib combination.

Amendments to the statistical analysis plan for the refametinib plus sorafenib study included the collection of survival data to be continued until the last patient's last visit instead of until 12 weeks after the last patient's first treatment, or earlier if all patients had withdrawn from the study. A data rule was also added regarding tumor assessment by centralized blinded reading; for cases with missing adjudication for patients who had completed or withdrawn from treatment at the time of primary analysis, the worst-case approach was to be applied.

Written, informed consent was obtained from all patients. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institutions' human research committees.

Eligibility criteria for RAS mutational testing included age ≥18 years with unresectable or metastatic HCC, confirmed histologically (mandatory for noncirrhotic patients) or by noninvasive radiologic criteria, Eastern Cooperative Oncology Group performance status of 0 or 1, and life expectancy ≥12 weeks. Prior use of targeted agents, experimental therapy, or systemic anticancer treatment was not allowed, although prior sorafenib treatment was permitted in patients who received refametinib only.

Treatment eligibility criteria included KRAS or NRAS mutation based on BEAMing plasma test; Child–Pugh class A liver function status; at least one unidimensional measurable lesion by computed tomography or magnetic resonance imaging; and Eastern Cooperative Oncology Group performance status of 0 or 1. Treatment exclusion criteria included any cancer curatively treated less than 3 years before study entry (except cervical carcinoma in situ, treated basal cell carcinoma, and superficial bladder tumors); eligibility for surgery, liver transplantation, ablation, or transarterial chemoembolization for HCC; renal failure requiring hemo- or peritoneal dialysis; a history of cardiac disease; or uncontrolled hypertension.

In both studies, eligible patients harboring RAS mutations received refametinib 50 mg twice daily in 21-day cycles. In the refametinib plus sorafenib study, patients also received standard sorafenib (400 mg twice daily), starting with a dose of 600 mg daily (200 mg in the morning plus 400 mg in the evening) in cycle 1, escalating to the standard sorafenib dose in cycle 2 if no hand–foot skin reaction, fatigue, or gastrointestinal toxicities of grade ≥2 occurred. Patients received treatment on a continuous basis until radiologic disease progression, clinical progression, or other criteria for discontinuation of treatment were met.

ctDNA from plasma samples collected in the pretreatment period was centrally evaluated for RAS mutational status using BEAMing technology (13), with a limit of detection at 0.02% mutant allele. Tumor assessments were performed at screening and every 6 weeks. Treatment response was centrally assessed according to mRECIST for the primary endpoint and was also investigator-assessed according to mRECIST (18). Safety, including adverse events (AEs) and concomitant medications, was monitored throughout the studies. Creatine phosphokinase (CPK) increase of grade ≥3 was considered an AE of special interest and was to be reported as a serious AE (SAE). Plasma samples for biomarker analysis were collected at screening, at cycle 1, days 1 and 15, and at cycle 2, day 15. Peripheral whole-blood samples from patients with mutated RAS were analyzed for detection of genomic alterations using FoundationACT (Foundation Medicine), a targeted next-generation sequencing (NGS) based ctDNA assay (19). The detection limit of FoundationACT is specified at 0.1% mutant allele frequency, i.e., a lower sensitivity than BEAMing. FoundationACT is a hybrid-capture-based assay that is designed to interrogate 62 genes, identifying all classes of alterations including base substitutions, insertions and deletions, copy-number variations, and rearrangements/fusions through computational algorithms (20).

In each study, it was estimated that approximately 350 patients were needed to be tested via BEAMing to identify 15 patients with mutated RAS in stage 1, and that approximately 2,300 patients would need to be tested via BEAMing to identify a sufficient number of patients with mutated RAS to be treated in stage 2.

Descriptive statistics were calculated for the presented endpoints.

In the refametinib monotherapy study, 498 patients were enrolled at 58 study centers in 17 countries across Asia, Europe, and the USA from September 2013 to June 2014. RAS mutational testing was performed in 493 patients (Fig. 1A); 32 (6.5%) had a RAS mutation. In the refametinib plus sorafenib study, 820 patients were enrolled at 80 study centers in 21 countries across Asia, Europe, and the USA from September 2013 to April 2015. RAS mutational testing was performed in 815 patients (Fig. 1B); 27 (3.3%) had a RAS mutation. Overall, 4.4% of patients with HCC screened (59/1,318) had a RAS mutation determined by BEAMing. Of those, 32 of 59 patients received treatment, either refametinib monotherapy (n = 16) or refametinib plus sorafenib (n = 16). Reasons for patients not receiving treatment are summarized in Fig. 1.

In the refametinib monotherapy study, the median age was 69 years and the median time since initial HCC diagnosis was 72.1 weeks (Table 1). Nine patients (56.3%) had received prior first-line sorafenib treatment. In the refametinib plus sorafenib study, the median age was 67 years and the median time since initial HCC diagnosis was 32.2 weeks (Table 1). Demographics and baseline characteristics were similar between patients irrespective of RAS mutational status in both studies.

Of the 16 patients treated with refametinib monotherapy, no patient had a CR or PR when centrally assessed according to mRECIST, and the ORR was 0% (Table 2). One patient (6.3%) achieved an unconfirmed PR, and eight (50.0%) achieved stable disease; the disease control rate was 56.3%. ORR was 0% by independent assessment according to RECIST version 1.1: no patients had a CR or PR, 10 (62.5%) had stable disease, two (12.5%) had disease progression, and four (25.0%) were not evaluable. The investigator-assessed ORR was 0% according to mRECIST (Supplementary Table S1) and RECIST version 1.1: no patients had a confirmed or unconfirmed CR or PR, 10 (62.5%) had stable disease, two (12.5%) had disease progression, and four (25.0%) were not evaluable.

Of the 16 patients treated with refametinib plus sorafenib, one patient (6.3%) achieved a confirmed PR when centrally assessed according to mRECIST, and the ORR was 6.3% (Table 2). Two patients (12.5%) achieved unconfirmed PRs (confirmatory computed tomography scan showed progression) and four (25.0%) had stable disease; the disease control rate was 43.8%. Independent assessment according to RECIST version 1.1 reported an ORR of 6.3%: one patient (6.3%) had a confirmed PR, six (37.5%) had stable disease, five (31.3%) had disease progression, and four (25.0%) were not evaluable or had data missing. Investigator-assessed ORR was 6.3% according to mRECIST [one patient (6.3%) had a confirmed PR; Supplementary Table S1]. The investigator-assessed ORR was also 6.3% according to RECIST version 1.1: one patient (6.3%) had a PR, one (6.3%) had an unconfirmed PR, five (31.3%) had stable disease, six (37.5%) had disease progression, and three (18.8%) had missing data.

In the refametinib monotherapy study, four patients (25.0%) had radiologic progression, and the median time to progression was 2.8 months (Fig. 2A). Seven patients (43.8%) in the refametinib plus sorafenib study had radiologic progression, and the median time to progression was 2.8 months (Fig. 2B).

Duration of response could not be calculated in the refametinib monotherapy study because no patient achieved a CR or PR. Duration of response based on central assessment in the refametinib plus sorafenib study was 1.4 months for the one patient who achieved a confirmed PR; this patient had a KRAS^G35A point mutation. Duration of response was 2.7 months for the one patient who was investigator assessed as achieving a confirmed PR; this patient had a KRAS^G38A mutation.

In the refametinib monotherapy study, nine patients (56.3%) had disease progression or died and median progression-free survival was 1.9 months (Fig. 2C). Eight patients (50.0%) died, and median overall survival was 5.8 months (Fig. 2E). In the refametinib plus sorafenib study, 10 patients (62.5%) had disease progression or died, and median progression-free survival was 1.5 months (Fig. 2D). Nine patients (56.3%) died, and median overall survival was 12.7 months (Fig. 2F).

With refametinib monotherapy, the median duration of treatment (including interruptions) was 7.14 weeks. The mean (±standard deviation) daily dose of refametinib (excluding interruptions) was 90.01 ± 13.88 mg.

With refametinib plus sorafenib, the median durations of treatment (including interruptions) for refametinib and sorafenib were 8.21 and 6.43 weeks, respectively. The mean (±standard deviation) daily doses (excluding interruptions) of refametinib and sorafenib were 85.06 ± 16.07 mg and 514.24 ± 124.23 mg, respectively. The majority of patients experienced treatment-emergent AEs (TEAE) of hand–foot skin reaction, fatigue, or gastrointestinal toxicities of grade ≥2 in cycle 1, so only three patients (18.8%) received the full dose of sorafenib (800 mg/day) following cycle 1. One patient remained on treatment at the time of data cutoff and has been ongoing for approximately 2 years.

TEAEs led to dose modification (interruption or reduction) in 14 patients (87.5%) receiving refametinib monotherapy (Table 3) and were considered drug-related in 13 patients (81.3%). Treatment was permanently discontinued because of TEAEs in four patients (25.0%) and were considered drug-related in three patients (18.8%).

With refametinib plus sorafenib, dose modifications were reported in 11 patients (68.8%) with refametinib and 11 patients (68.8%) with sorafenib. TEAEs led to dose modification in 15 patients (93.8%; Table 3); events were considered refametinib-related in 13 patients (81.3%) and sorafenib-related in 14 patients (87.5%).

At least one TEAE was reported in all 16 patients (100%) receiving refametinib monotherapy (Table 3). The most common TEAEs of worst grade 3 were fatigue and increased CPK (three patients each; 18.8%). Grade 5 TEAEs occurred in five patients (31.3%): sepsis, death not otherwise specified, multiorgan failure, lung infection, and heart failure. The causes of death were progressive disease (one patient) and AEs associated with clinical disease progression and AEs not associated with clinical disease progression (two patients each). Drug-related TEAEs occurred in 14 patients (87.5%; Supplementary Table S2). In most patients (75.0%), the worst grade of drug-related TEAEs was grade 3, and one patient (6.3%) had a drug-related TEAE of grade 4 (increased serum amylase). Twelve patients (75.0%) experienced SAEs (Supplementary Table S3), of which the most common worst grade was grade 3 (43.8%). SAEs were refametinib-related in seven patients (43.8%), most commonly increased CPK (three patients; 18.8%). All other refametinib-related SAEs were reported in one patient each (6.3%; Supplementary Table S3).

TEAEs occurred in all 16 patients (100%) receiving refametinib plus sorafenib (Table 3). Hand–foot skin reaction was reported in two patients (12.5%). The most common TEAEs of worst grade 3 were hypertension (10/16; 62.5%) and increased aspartate aminotransferase and increased CPK in five patients each (31.3%). Seven TEAEs of worst grade 4 were reported in three patients (18.8%): increased aspartate aminotransferase, increased CPK, decreased platelet count, investigations—other, hypophosphatemia, and hyperuricemia. Grade 5 TEAEs included general disorders and administration site conditions—other and dyspnea (one patient each); the cause of death was AE associated with clinical disease progression and progressive disease (one patient each). Refametinib- and sorafenib-related TEAEs were reported for all 16 patients (100%; Supplementary Table S2). Nine patients (56.3%) had refametinib-related TEAEs of grade 3, and three patients (18.8%) had refametinib-related TEAEs of grade 4 (Supplementary Table S2). Twelve patients (75.0%) had sorafenib-related TEAEs of worst grade 3, and three patients (18.8%) had sorafenib-related TEAEs of worst grade 4. One patient (6.3%) had a grade 5 TEAE considered related to both refametinib and sorafenib (general disorders and administration site conditions—other). SAEs were reported in 13 patients (81.3%; Supplementary Table S3), most commonly increased CPK in six patients (37.5%; five grade 3, one grade 4). Refametinib-related SAEs were experienced by 12 patients (75.0%), most frequently worst grade 3 (7/16; 43.8%). Increased CPK was the most commonly reported refametinib-related SAE (5/16; 31.3%; four grade 3, one grade 4). Sorafenib-related SAEs occurred in 10 patients (62.5%); 7 patients (43.8%) had events of worst grade 3, and one patient (6.3%) experienced worst grade 4 (increased CPK). One SAE of grade 5 was considered refametinib-related and sorafenib-related (general disorders and administration site conditions—other).

To identify potential genomic biomarkers that might be associated with resistance to refametinib monotherapy or combination therapy, NGS (FoundationACT) was performed on available ctDNA from 27 patients (refametinib monotherapy, n = 15; refametinib plus sorafenib, n = 12). RAS mutations were not called by NGS in over 60% of the samples with a mutant allele frequency of between 0.02% and 0.1% as determined by BEAMing. RAS mutational status was confirmed by NGS in 12 patients (44.4%), all with a mutant allele frequency above 0.1%. The RAS somatic aberration detected was concordant with BEAMing results in 11 patients (91.7%). Excluding RAS, the most frequently detected mutation was in the promoter region of telomerase reverse transcriptase (TERT; 17/27; 63.0%), followed by TP53 (13/27; 48.1%), and β-catenin (CTNNB1; 10/27; 37.0%; Fig. 3). Actionable mutations were rare (<10%) and included oncogenes such as EGFR, JAK2, BRAF, FLT3, PIK3CA, and cKIT.

These two phase II proof-of-concept studies prospectively evaluated the efficacy and safety of refametinib monotherapy or refametinib plus sorafenib in patients prospectively screened for RAS-mutant unresectable or metastatic HCC based on evaluation of mutational status in ctDNA. The previous phase II BASIL trial in a separate population of Asian patients with HCC receiving refametinib plus sorafenib demonstrated that the majority of patients who responded to this regimen had mutant RAS tumors, with an ORR of 75% in patients with RAS-mutant HCC compared with 1.5% in patients with HCC with no RAS mutation (17).

In these studies, prospective testing for RAS mutations using ctDNA isolated from plasma was a feasible, noninvasive approach for large-scale mutational testing in patients with HCC. The current findings support a previous report of the use of ctDNA to detect KRAS mutations via BEAMing in a small study of patients with refractory colorectal carcinoma treated with regorafenib (21), although KRAS mutational frequency was notably higher in the colorectal carcinoma population (∼40%) compared with that reported here (∼5%). Overall, 59 of 1,318 (4.4%) of the HCC patients screened had a RAS mutation. The RAS mutation rates reported here are consistent with previous reports in this patient population (∼5%; refs. 22–25). It should therefore be noted that the low RAS mutational frequency in this population suggests that identifying RAS-mutant patients may be challenging in practice.

The primary efficacy variable was not met in the refametinib monotherapy study, with no patient with mutated RAS achieving a CR or PR. In the refametinib plus sorafenib study, one patient with mutated RAS achieved a PR, resulting in an ORR of 6.3%, which is broadly similar to the ORR of 6.9% reported in the BASIL trial (17).

The target for the first stage of the trials (≥5/15 patients with a CR or PR) was not reached; therefore, these studies did not proceed to the planned evaluation of refametinib monotherapy or combination therapy in a larger number of patients. Further exploration would be required to understand the lower ORR with refametinib plus sorafenib in this study compared with previous reports (17). These results suggest that the use of RAS mutational status as a prognostic biomarker for treatment response to refametinib monotherapy or in combination with sorafenib was unsuccessful, and targeting MEK with refametinib in this RAS-mutant patient population did not lead to a significant proportion of objective responses. However, the low number of patients treated should be taken into account, and the low proportion of responses observed may reflect random error—these results should therefore be interpreted with caution. Additional molecular events may explain the limited responses seen using mutated RAS as a prognostic biomarker for targeted MEK inhibition in these studies. It is possible that with intratumor heterogeneity, mutations occurring in low-frequency subclonal tumor cell populations may have acquired mutations that conferred resistance to refametinib, which was targeted to progenitor cells expressing truncal driver mutations in RAS, negatively affecting clinical outcomes (26). Evaluation of nontruncal mutations, together with longer-term evaluations of changes in allele frequency, was not planned in this study, although may provide useful insights into the development of resistance to refametinib in patients with HCC.

Median overall survival was 5.8 months with refametinib monotherapy and 12.7 months with refametinib plus sorafenib, with over half of events occurring during the study period. KRAS mutation is generally associated with poorer outcomes in most cancers, although there are no established data in HCC due to the lack of robust testing in large studies of advanced disease (23). In our study, the effect of refametinib monotherapy on overall survival can be considered insignificant, because the expected outcome of placebo at first or second line is 7 to 8 months (27, 28). In fact, this survival of under 6 months might indicate that advanced RAS-positive HCC tumors have a poor natural history. It should also be noted that 56% of patients in the monotherapy arm had received prior sorafenib, possibly contributing to the poor survival seen. However, the approximately 13-month survival outcome with refametinib plus sorafenib treatment is more intriguing, considering the expected median survival with first-line sorafenib monotherapy alone is 11 months (29). This result may indicate a synergistic effect between sorafenib and refametinib, which is relevant as tumors harboring RAS mutations remain some of the most challenging to treat because of the paucity of successful drugs targeting the RAS pathway (30). However, this finding should be interpreted with caution because of the heterogeneity in baseline liver function and tumor factors, which could affect response to treatment. Also, patients in the refametinib plus sorafenib study had a much shorter median time from initial diagnosis to study treatment compared with patients in the monotherapy study (32.1 weeks vs. 72.1 weeks, respectively). The overall survival findings in the combination study support those described for patients receiving refametinib plus sorafenib in the BASIL trial (17), although median overall survival was increased in our study (12.7 vs. 9.5 months, respectively).

Median time to progression was the same across the refametinib monotherapy and refametinib plus sorafenib studies (2.8 months), with similar median progression-free survival observed between the studies (1.9 and 1.5 months, respectively). However, progression-free survival times were lower than those in previous reports (17).

Drug exposure was similar between both studies and similar to the median refametinib dose observed in the BASIL study (17). The majority of patients in the refametinib plus sorafenib study experienced AEs during cycle 1 that prevented sorafenib dose escalation to 800 mg per day, which was also observed in the BASIL study (17). The majority of patients across both studies experienced AEs leading to dose modifications, which may have caused insufficient drug exposure, potentially leading to reduced efficacy of both refametinib regimens. Overall, median duration of treatment was relatively short in both trials (7 and 8 weeks, respectively), similar to that reported for the BASIL study (8 weeks; ref. 17).

Overall, refametinib was tolerated as monotherapy and combination therapy, and the majority of TEAEs were manageable in both studies. In patients receiving refametinib monotherapy, the most common TEAEs of limb edema, fatigue, nausea, and vomiting were consistent with the safety profile previously reported in a phase I study of refametinib (31). The high overall incidence of grade 3 TEAEs irrespective of causality in both studies (68.8%) was similar to that reported in the BASIL trial (60.0%; ref. 17). Generally, the observed incidence and severity of refametinib-related TEAEs observed with refametinib monotherapy were comparable with data from the previous refametinib phase I study (31). Refametinib-related SAEs were less frequent with refametinib monotherapy than with refametinib plus sorafenib (43.8% vs. 75.0%, respectively). Increased CPK grade ≥3 was the most common refametinib-related SAE reported in both studies, consistent with reports of increased CPK as a class effect of MEK inhibitors (32–34).

Compared with the known safety profile of sorafenib monotherapy (29, 35), a higher incidence of liver and gastrointestinal toxicities and rash was observed in patients who received refametinib plus sorafenib. However, alopecia and hand–foot skin reaction were less common compared with those reported for sorafenib monotherapy (14% vs. 12.5% and 21% vs. 6.3%, respectively; ref. 29), possibly due to the reduced exposure to sorafenib in the majority of patients in our study.

Biomarker analysis of ctDNA analyzed by NGS showed the observed mutational landscape to be consistent with published data for HCC (36). The most common mutation was in the promoter region of TERT, supporting previous observations in patients with HCC and combined HCC-cholangiocarcinoma (37, 38). Few actionable mutations were found, with none appearing to explain the resistance to refametinib alone or in combination with sorafenib, and few of the detected mutations are feasible for targeting with existing drugs. It therefore remains inconclusive from our results as to whether somatic mutations in oncogenes affect the efficacy of refametinib in monotherapy or combination therapy. Although analyses were planned to evaluate the role of biomarkers in the response to treatment, due to limited sample size and early study termination it was not possible to fully address the role of intratumor heterogeneity (26). In addition, although the two study populations included only Child–Pugh A patients, these patients were heterogeneous for various factors that may be prognostic for treatment response, such as a history of ascites (in four patients overall; 12.5%; ref. 39), alpha-fetoprotein (>400 μg/L in 12 patients; 37.5%; ref. 40), microvascular invasion (in 11 patients; 34%; ref. 41), extrahepatic spread (in 16 patients; 50%; ref. 41), and hepatitis C (in seven patients; 21.9%; ref. 41). However, no formal analysis of lung function status and tumor factors as prognostic markers for treatment response was planned in these studies.

In these studies, RAS mutational status as determined by BEAMing was confirmed in 44% of samples using NGS, all with mutant allele frequencies of 0.1% or higher. Although BEAMing technology is highly sensitive (42), the newly developed NGS from ctDNA approach has demonstrated high concordance, confirming nearly all mutations identified by BEAMing, and offers the additional advantage of providing the mutational landscape based on ctDNA. A comparison of sensitivity between both assays is difficult due to the different detection limit of each method (0.02% for BEAMing vs. 0.1% for NGS), which did not allow for the detection of RAS mutational status by NGS in over 60% of samples with mutant allele frequency between 0.02% and 0.1%. Nonetheless, our results demonstrated that NGS appears to be a promising noninvasive approach to determine the landscape of somatic mutations, particularly for patients in whom a biopsy is not an option (19).

Despite the poor ORR in patients with RAS mutations, a median overall survival of 13 months in the small population included in the refametinib plus sorafenib study may indicate a synergistic effect between refametinib and sorafenib that should be further explored in a larger patient population that is not stratified by RAS mutational status, taking into account other prognostic factors based on patient heterogeneity and intratumor heterogeneity. The analysis of mutational status using ctDNA isolated from plasma as a liquid biopsy was a feasible, noninvasive technique in patients with unresectable or metastatic HCC, although RAS mutational frequency was low. Further analysis of this technique is warranted for discovery of predictive biomarkers in HCC and other cancers.

K.H. Weiss reports grants from Novartis and MSD, personal fees from Bayer, GMPO, and Vivet Therapeutics, and grants and personal fees from Bristol-Myers Squibb, Univar, Wilson Therapeutics, and Alexion. P. Ross reports grants from Sanofi Aventis, fees for educational sessions from BMS, travel support from BMS, Sirtex, Bayer, Merck Serono, Servier, Amgen, and Celgene, advisory board fees from BMS, Sirtex, Bayer, Merck Serono, Servier, and Celgene, and speaking fees from Merck Serono, Sirtex, and Bayer. V. Mazzaferro reports personal fees from Bayer and BTG. J.-F. Blanc reports personal fees from Bayer SP, Lilly, and BMS. Y.T. Ma reports honoraria and travel support from Bayer and consultancy fees from Baxalta UK, Ltd. S.P. Choo reports grants from Bristol-Myers Squibb, travel support from Amgen, and honoraria from Bristol-Myers Squibb, Bayer, Novartis, and Sirtex. J.-F. Dufour reports participating in advisory committees for AbbVie, Bayer, Bristol-Myers Squibb, Falk, Genfit, Gilead Sciences, Intercept, Lilly, Merck, and Novartis, speaking and teaching for AbbVie, Bayer, Bristol-Myers Squibb, Genfit, Gilead Science, and Novartis, and an unrestricted research grant from Bayer. E.J. Gane reports participating in clinical advisory committees or speaker bureaus for Alios, AbbVie, Arrowhead, Janssen, Merck, Gilead Sciences, and Mylan. M. Peck-Radosavljevic reports being an investigator for AbbVie, ArQule-Daiichi, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Imclone, Lilly, MSD, Novartis, and Roche, grant support from AbbVie, ArQule-Daiichi, Bayer, MSD, and Roche, and acting as a speaker or advisor for Abbott, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Lilly, MSD, and Roche. T. Dao reports congress and travel support from Bayer, Gilead Sciences, AbbVie, and MSD, and being an investigator for Lilly. M. Teufel and B.H. Childs are employees of Bayer HealthCare Pharmaceuticals, Inc. K. Roth and H. Krissel are employees of Bayer AG. H. Krissel also reports patent BHC 133035 EP 01 pending. D. Reis is an employee of and has received personal fees from Bayer S.A. J.M. Llovet reports research/education grants from Bayer, Blueprint Medicines, Boehringer Ingelheim, and Incyte and consulting fees from Eli Lilly, Bayer, BMS, Blueprint Medicines, Eisai, Celsion, and Boehringer Ingelheim. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H.Y. Lim, B.-Y. Ryoo, M. Teufel, D. Reis, B.H. Childs, H. Krissel, J.M. Llovet

Development of methodology: H.Y. Lim, T. Yau, B.-Y. Ryoo, M. Teufel, D. Reis, B.H. Childs, H. Krissel, J.M. Llovet

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.Y. Lim, P. Merle, K.H. Weiss, T. Yau, P. Ross, V. Mazzaferro, J.-F. Blanc, Y.T. Ma, C.J. Yen, J. Kocsis, S.P. Choo, W. Sukeepaisarnjaroen, R. Gérolami, J.-F. Dufour, E.J. Gane, B.-Y. Ryoo, M. Peck-Radosavljevic, T. Dao, W. Yeo, S. Thongsawat, B.H. Childs, H. Krissel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.Y. Lim, P. Merle, T. Yau, S.P. Choo, B.-Y. Ryoo, M. Peck-Radosavljevic, M. Teufel, K. Roth, D. Reis, B.H. Childs, H. Krissel

Writing, review, and/or revision of the manuscript: H.Y. Lim, K.H. Weiss, T. Yau, P. Ross, V. Mazzaferro, J.-F. Blanc, Y.T. Ma, C.J. Yen, S.P. Choo, R. Gérolami, J.-F. Dufour, B.-Y. Ryoo, M. Peck-Radosavljevic, T. Dao, W. Yeo, W. Lamlertthon, M. Teufel, K. Roth, D. Reis, B.H. Childs, H. Krissel, J.M. Llovet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.Y. Lim, T. Yau, J. Kocsis, S.P. Choo, B.-Y. Ryoo, W. Lamlertthon, H. Krissel

Study supervision: H.Y. Lim, T. Yau, Y.T. Ma, B.-Y. Ryoo, W. Lamlertthon, D. Reis, B.H. Childs, H. Krissel

The authors wish to thank the patients and their families. These studies were supported by Bayer AG. Medical writing assistance was provided by Laura Badtke, PhD, at Complete HealthVizion, Chicago, IL, based on detailed discussion and feedback from all authors. Medical writing assistance was funded by Bayer AG. The study sponsor played a role in the design of the study, data interpretation, writing of the manuscript, and the decision to submit for publication. At each study center, the principal investigator was responsible for the study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.