Trametinib activity in patients with solid tumors and lymphomas harboring BRAF non-V600 mutations or fusions: results from NCI-MATCH (EAY131)

Abstract

Purpose

Substantial preclinical evidence and case reports suggest that MEK inhibition is an active approach in tumors with BRAF mutations outside the V600 locus, and in BRAF fusions. Thus, Subprotocol R of the NCI-MATCH study tested the MEK inhibitor trametinib in this population.

Methods

The NCI-MATCH study performed genomic profiling on tumor samples from patients with solid tumors and lymphomas progressing on standard therapies or with no standard treatments. Patients with pre-specified fusions and non-V600 mutations in BRAF were assigned to Subprotocol R using the NCI-MATCHBOX algorithm. The primary endpoint was objective response rate (ORR).

Results

Among 50 patients assigned, 32 were eligible and received therapy with trametinib. Of these, 1 had a BRAF fusion and 31 had BRAF mutations (13 and 19 with class 2 and 3 mutations, respectively). There were no complete responses; 1 patient (3%) had a confirmed partial response (patient with breast ductal adenocarcinoma with BRAF G469E mutation) and 10 patients had stable disease as best response (clinical benefit rate 34%). Median progression free survival was 1.8 months and median overall survival was 5.7 months. Exploratory subgroup analyses showed that patients with colorectal adenocarcinoma (n=8) had particularly poor PFS. No new toxicity signals were identified.

Conclusions

Trametinib did not show promising clinical activity in patients with tumors harboring non-V600 BRAF mutations, and the subprotocol did not meet its primary endpoint.

Introduction

Molecularly-guided therapy has made a major impact in certain cancer types in tumors with particular genomic alterations (e.g. EGFR mutations, BRAF V600E mutations, ALK fusions). Most commonly identified mutations in cancer, by contrast, have no validated targeted therapy, despite extensive preclinical data suggesting effective therapeutic strategies in some cases. The National Cancer Institute Molecular Analysis for Therapy Choice (NCI-MATCH) Trial is a platform trial with multiple phase II tumor agnostic arms designed to evaluate genomically-targeted treatment strategies across multiple identified genomic changes in cancer.

BRAF inhibitors with or without a MEK inhibitor have demonstrated substantial clinical efficacy in BRAF V600 mutated melanoma, lung cancer, thyroid cancer, hairy cell leukemia, and other cancers.^1–5 However, non-V600 BRAF mutations are identified in a substantial portion of patients across cancers (up to 3% total in some publicly available databases), with no validated molecularly guided therapy for them.^6,7 These non-V600 BRAF mutations activate mitogen-activated protein kinase (MAPK) pathway signaling similarly (albeit generally slightly less robustly) than the BRAF V600 mutations.^8,9 Similarly, fusions in BRAF, which remove the inhibitory-RAS binding domain and hyperactivate MAPK signaling, are also present across cancers.^6,7,10,11 Several pre-clinical studies, particularly in melanoma models harboring BRAF L597 mutations and BRAF fusions, suggested BRAF inhibitor monotherapy would not be effective. In contrast, MEK inhibitors demonstrated substantial pre-clinical efficacy in these studies.^12,13 In addition, several case reports have demonstrated that MEK inhibitors could produce excellent clinical responses in patients with these molecular variants.^12,14–16

MEK inhibitors have shown variable degrees of activity in several settings, including BRAF V600 mutant melanoma, NRAS mutant melanoma, low-grade serous ovarian cancer, plexiform neurofibromas, thyroid cancer, and low-grade gliomas, with more limited responses in KRAS mutant pancreatic cancer or lung cancer.^17–22 Trametinib, a selective, allosteric inhibitor of MEK1/2, is approved in BRAF V600 mutant melanoma (alone or in combination with the BRAF inhibitor dabrafenib).^17,23 This agent has also been extensively studied in pre-clinical and clinical scenarios, and is the only FDA approved MEK inhibitor monotherapy. Herein, we report the results for NCI-MATCH Subprotocol R: A Phase II study of trametinib in patients with BRAF fusions, or with non-V600 BRAF mutations.

Methods

Subprotocol overview

The NCI-MATCH trial, developed by ECOG-ACRIN Cancer Research Group (ECOG-ACRIN) and the National Cancer Institute (NCI), aimed to find signals of efficacy for treatments targeted to actionable molecular alterations found in any tumor type. The R subprotocol, reported here, was a single arm, phase 2 trial to test the efficacy and safety of trametinib in patients with cancers harboring fusions or non-V600 mutations in BRAF. The study was reviewed by the NCI central IRB and all patients signed written informed consent. The study was conducted according to the Declaration of Helsinki.

Patient selection

Eligible patients were adults with any solid tumor, lymphoma or myeloma who progressed on standard treatment or for whom no standard treatment was available and whose tumor contained an eligible BRAF variant (either by profiling a fresh biopsy with the NCI-MATCH assay²⁴ or after determination by an assay performed on tumor in a CLIA-approved NCI-MATCH accepted laboratory). Adequate hematopoietic, liver and kidney function, an Eastern Cooperative Oncology Group (ECOG) performance status ≤1, were required. Patients were excluded if they had prior treatment with a MEK inhibitor, prior significant cardiac disease (including arrhythmias, treatment-refractory hypertension, decreased cardiac ejection fraction), or prior interstitial lung disease.

Tumor sequencing and subprotocol assignment

Between August 2015 to May 11, 2017, a central network of laboratory reporting was used to determine eligibility. Biopsy specimens in buffered formalin (29 patients) and/or cytology specimens with smears and in Cytolyt^R for cell block preparation (two patients) or archived formalin-fixed paraffin-embedded tissue blocks (2 patients) were shipped overnight to the CLIA-accredited central processing laboratory for the trial. Tumor profiling in these cytology or biopsy specimens was accomplished as previously described²⁴, using an NGS panel of 143 genes that identified single nucleotide variants (SNV), indels, amplifications and selected fusions. Central immunohistochemistry (IHC) assays for expression of PTEN, MLH1, MSH2 as reported previously²⁵ were done on 29 tumors and in a commercial laboratory for one tumor. Patients were assigned using a validated NCI designed informatics rules algorithm (MATCHBOX) (Manuscript under review). After May 11, 2017, patient’s eligibility was initially determined by a referral from a certified genomic laboratory (https://ecog-acrin.org/nci-match-eay131-designated-labs) and later confirmed by the NCI-MATCH central laboratory network. Two patients were enrolled through this referral method. Patients who had pre-specified fusions or mutations in BRAF (Supplemental Table 1) were assigned to Subprotocol R. Pre-specified mutations were identified based on levels of evidence which include the following: Level 1 – gene variant approved for selection of an approved drug, Level 2 – gene variant an eligibility criteria for ongoing clinical trial or has been identified in N of 1 responses, and/or Level 3 – preclinical inferential data that provide biological evidence sufficient to support the use of the variant in treatment.

Evaluation of Response and Toxicity

Patients were treated with trametinib 2 mg daily until disease progression, unacceptable toxicity, or patient/physician choice to discontinue therapy. Dose reductions were permitted to trametinib 1.5 mg daily, then to 1.0 mg daily for severe or persistent toxicities. Objective response was evaluated every 8 weeks using RECIST 1.1 criteria (for solid tumors) or Lugano criteria (lymphomas).^26,27 Patients continued on trametinib until progressive disease, unacceptable toxicity, or self-discontinuation. Toxicity was evaluated using Common Terminology Criteria for Adverse Events (CTCAE) version 4.03.

Statistical Considerations

The primary objective of NCI-MATCH study was to evaluate the objective response rate (ORR) for each subprotocol, defined as proportion of patients with best overall response of complete response or partial response based on applicable criteria (20,21). The ORR was compared against a null benchmark value of 5%. A response rate of 5/31 patients (16%) or more was predefined as a signal of promising activity. This design had approximately 92% power to conclude an agent’s activity is promising if its true ORR is 25%, with one-sided type I error rate of 1.8%. Allowing for 10% ineligibility rate, the accrual goal was 35 patients for this subprotocol. Secondary objectives included progression-free survival at 6 months (PFS6), PFS, overall survival (OS), toxicity assessment, and evaluation of predictive biomarkers (co-mutations or other factors that potentially predict response). PFS was defined as time from treatment start to disease progression or death from any cause; OS was defined as time from treatment start to death from any cause. Both PFS and OS were estimated using Kaplan-Meier method. In an exploratory, unplanned fashion, we assessed PFS and OS based on prior therapies, location of mutation (exon 11 vs. 15), histology, co-occurring mutations, BRAF allele frequency, and BRAF mutation class (as defined by Yao et al)⁹. Co-occurring mutations were classified as concurrent RAS vs. no RAS mutations (mutations in KRAS, NRAS, HRAS), or PI3K pathway vs. no PI3K pathway (mutations in PI3K, AKT, MTOR, TSC1).

Results

Patients

Subprotocol R was activated August 12, 2015. Between August 12, 2015 and August 17, 2017, 50 patients were assigned to Subprotocol R. Thirty-five patients were enrolled to the subprotocol and 15 patients were ineligible to enroll (Figure 1). Of these 35 patients, all received at least one dose of protocol therapy, 3 were found to be ineligible after treatment, and thus 32 patients were evaluable for efficacy endpoints. Table 1 lists patient characteristics; median age was 65.5 years (range 40–83), and 18 (58%) were female. Most patients had ECOG PS of 1 (n=26; 81%) and 69% (n=22) of patients had 3 or more prior therapies. Tumor histopathologic classification is listed in Table 1. GI cancers (n=8, 25%, of which 7 were colorectal adenocarcinoma), lung adenocarcinomas (n=9, 28%), and prostate adenocarcinoma (n=4, 12%, 3 with neuroendocrine differentiation) were the most common subtypes enrolled.

Figure 1:

CONSORT diagram showing numbers of patients assigned, enrolled, and evaluable, as well as reasons for lack of enrollment or evaluability. LVEF: left ventricular ejection fraction; ULN: upper limit of normal.

Table 1:

Patient Characteristics

Characteristics	No. of patients (%)
Total no. of evaluable patients	32
Age (median, range)	65.5(40,83)
Sex
Male	14 (44%)
Female	18 (56%)
Race
White	30 (100%)
Unknown	2
Ethnicity
Hispanic	2 (7%)
Non-Hispanic	28 (93%)
Unknown	2
ECOG PS
0	6 (19%)
1	26 (81%)
No. of prior therapies
1	6 (19%)
2	4 (12%)
3	4 (12%)
>3	18 (56%)
Weight loss in previous 6 months
< 5%	26 (81%)
5 to < 10%	4 (12%)
10 to < 20%	3 (6%)
Tumor Histology
Gastrointestinal	8 (25%)
Adenocarcinoma of colon	6
Adenocarcinoma of rectum	1
Intrahepatic cholangiocarcinoma	1
Gynecologic	4 (12%)
Serous adenocarcinoma of ovary	1
Malignant mixed Mullerian tumor of uterus	1
Endometrioid endometrial adenocarcinoma	1
Melanoma of vulva	1
Breast (ductal carcinoma)	1 (3%)
Lung adenocarcinoma	9 (28%)
Adenocarcinoma	6
Adenosquamous carcinoma	1
Hepatoid adenocarcinoma	1
Sarcomatoid adenocarcinoma	1
Genitourinary	5 (16%)
Prostate adenocarcinoma	1
Prostate adenocarcinoma with neuroendocrine differentiation	3
Osteosarcoma of the renal pelvis	1
Lymphoma	2 (6%)
Cutaneous T-cell anaplastic large cell lymphoma	1
Diffuse large B-cell lymphoma	1
Spindle cell component of parotid epithelial- myoepithelial carcinoma	1
Adenocarcinoma of unknown primary site	2 (6%)

Various BRAF mutations were identified, as well as a single BRAF fusion (n=1) (Supplementary Table 2, Supplementary Figure 1). Exon 11 mutations were identified in 13 patients (42%), including in G464 (n=2), G466 (n=4), and G469 (n=7). Exon 15 mutations were present in 18 patients (58%), including N581 (n=3), D594 (n=11), L597 (n=2), and K601 (n=1). Co-occurring mutations were also diverse, including those in APC (n=10), HRAS (n=1), KRAS (n=2), NRAS (n=2), PIK3CA (n=6), and TP53 (n=16).

Efficacy

Of the 32 patients evaluable for efficacy endpoints, there were no complete responses, 1 patient had a partial response, 10 had stable disease, and 15 had progressive disease. Six patients did not have any imaging assessment prior to death (n=5), or withdrawal (n=1) (Figure 2A). Notably, the patient who withdrew after one cycle remained alive at 20.8 months post-registration. Thus, we observed a response rate of 3% (1/32, 90% CI 0.2%−14%), and a potential clinical benefit rate of 34% (90% CI: 21%−50%). The patient with a partial response had invasive breast cancer with a BRAF G469E mutation; at 4 months on treatment she had a maximal partial response (with 50% tumor shrinkage) but died suddenly at 4.3 months (potentially drug-related, and of unknown cause) without progression. Four additional patients had stable disease with PFS >6 months, including one patient with lung adenocarcinoma with BRAF G469A mutation who remains on therapy for 22 cycles (20.4 months) without progression, and a prostate cancer patient with a BRAF K601E mutation with a near partial response with progression at 9.7 months after starting therapy (Figure 2B, Table 2, Supplementary Figure 2).

Figure 2:

(A) Waterfall plot showing depth of response at best change from baseline in target lesion size; best confirmed response noted as PD (progressive disease), SD (stable disease), or PR (partial response) in 24 patients whose best change in target lesion size was evaluable (6 patients without any follow-up imaging assessment and 3 patients without confirmatory follow-up assessment of their target lesions were not included), triangle denotes new lesion as cause of PD; (B) “Swimmer’s” plot showing treatment duration for all 32 evaluable patients and their occurrence of response (*), disease progression (+) and death (#); (C) Progression-free survival (1 patient was censored at registration and did not contribute to the analysis); (D) Overall Survival

Table 2:

Patients with response or prolonged stable disease

Histology	Mutation type	Concurrent mutated genes	Best conf. response	# Cycles	PD	Dead	PFS time (months)	OS time (months)
Ductal carcinoma of breast	Gly469Glu	NF1, PIK3CA, RHOA, TP53	PR	5	No	Yes	4.1	4.3
Spindle cell neoplasm	Gly464Glu	CBL, MAP2K1, TP53	SD	2	Yes	Yes	6.9	15.5
Endometrioid endometrial adenocarcinoma	Leu597Val	KRAS, TP53	SD	6	Yes	Yes	7.9	13.1
Adenocarcinom of prostate with neuroendocrine differentiation	Lys601Glu	No	SD	8	Yes	Yes	9.7	18.5
Adenocarcinoma of lung	Gly469Ala	No	SD	22	No	No	20.0	20.4

The median PFS was 1.8 months (90% CI: 1.7, 3.4), with an estimated 6-month PFS rate of 17% (90% CI 8%, 30%) (Figure 2C). The estimated 6-month OS rate was 46% (90% CI: 30%, 59%), and the median OS was 5.7 (90% CI: 4.1, 8.1) months (Figure 2D). At last follow up, 29 (of 32) patients had died (3 patients were alive at 2.2, 20.4, and 20.8 months).

In exploratory analyses, we assessed whether histology, co-occurring mutations, BRAF allele frequency, and type of BRAF mutation affected benefit from trametinib (Figure 3;). Given the small sample size and post-hoc analyses, we did not formally statistically compare subgroups, rather, we provided the hazard ratios (HR) and associated 95% confidence intervals (CI) from univariate Cox proportional hazard models. We did not observe obvious differences in clinical outcomes in patients based on prior therapies (Supplementary Figure 3). A trend toward improved OS was observed in patients with exon 11 BRAF mutations (HR=0.45, 95% CI: 0.20, 1.00) -(Supplementary Figure 4). Patients with colorectal adenocarcinomas had particularly poor PFS (HR=3.22, 95% CI: 1.29, 8.02) (Supplementary Figure 5). Some trends toward improved PFS (HR=0.36, 95% CI: 0.15, 0.90) and OS (HR=0.50, 95% CI: 0.21, 1.20) were observed in patients lacking concurrent PI3K pathway gene mutations, albeit with small numbers, with no differences observed in patients with or without concurrent RAS mutations (Supplementary Figures 6–7). Interestingly, lower than median BRAF allele frequency also seemed associated with slightly better PFS (HR=0.76, 95% CI: 0.36, 1.61) and OS (HR=0.67, 95% CI: 0.32, 1.43) (Supplementary Figure 8). Finally, we assessed mutation class (class 2 vs. class 3; see discussion); class 2 mutations appeared associated with improved PFS (HR=0.50, 95% CI: 0.22, 1.14) and OS (HR=0.62, 95% CI: 0.29, 1.31) (Supplementary Figure 9).

Figure 3:

Oncoprint showing spectrum and allele frequency of identified BRAF mutations, co-occurring mutations, histology, and best overall response).

Safety

Adverse events at least possibly related to treatment are listed in Supplementary Table 3. All 35 patients enrolled started protocol therapy, but one patient declined all intervention and symptom assessment shortly after starting treatment, and adverse event data was not assessed, so the analysis population for toxicity was the 34 patients who were treated and reported adverse event data. Nine deaths on study were noted; two were considered possibly related to treatment (one patient with sudden death several days after the development of extreme fatigue and was found to have decreased cardiac ejection fraction and one patient with a thromboembolic event approximately one month following an ankle fracture, both events judged possibly due to drug or disease). Worst grade toxicity otherwise was grade 1–2 (n=16, 47%) or grade 3 (n=12, 35%). Toxicities were consistent with other MEK inhibitor studies overall, and included anemia (n=13, 38%), nausea (n=12, 35%), peripheral edema (n=11, 32%), and acneiform rash (n=11, 32%). Of the 35 treated patients, the median number of cycles was 2 (range 1–22). Among 31 eligible patients who had discontinued therapy, six (19%) discontinued due to toxicity, fifteen patients (48%) discontinued treatment due to disease progression, 4 due to death on study (2 due to disease, 2 to possible drug-related toxicities), 3 due to other complicating disease, 1 due to other reason, and 2 patients withdrew (Supplemental Table 4).

Discussion

In this study of trametinib in patients with BRAF non-V600 mutations, we found that trametinib had relatively low activity and the primary endpoint was not met. Too few patients with BRAF fusions (n=1) were included to characterize the activity of trametinib in this population. A few patients did experience clinical benefit with responses (3%) or prolonged stable disease. The toxicities observed were consistent with other studies of trametinib, without obvious new safety signals.

The explanation for this lack of benefit is not entirely evident. Patients were heavily pre-treated and multiple histologies were enrolled. Exploratory subgroup analyses were assessed to potentially identify signals of benefit or particularly poorly performing populations, although it should be noted that these were post-hoc and underpowered for definitive conclusions. Patients with concurrent PI3K pathway mutations seemed to experience worse PFS and OS, possibly indicating that parallel signaling networks may have driven resistance in many patients. BRAF and/or MEK inhibition has had little success in many tumor types, for example in colorectal cancer, which comprised 20% of patients in this study (and had particularly poor outcomes).^28,29 By contrast, only one patient (with melanoma of the vulva; who failed to respond) had a tumor type historically more sensitive to MEK inhibitors based on previously available data (e.g. melanoma, thyroid cancer). Thus, the available evidence suggests that histology (or molecular features that accompany histology) continues to play a role and provide context for mutations common to distinct cancer types. Further, 5 patients had concurrent RAS mutations, which typically do not respond to MEK inhibition; this may have contributed to the poor responses.

Another potential explanation lies in the types of BRAF mutations identified. One potentially useful framework and nomenclature is the Class 1–3 mutations recently described.^8,9 Class 1 mutations (limited to BRAF V600 mutations) signal as constitutively active monomers, whereas class 2 (G469, L597, K601) signal as constitutively active dimers. By contrast, class 3 mutations (D594, G466, A581) have impaired kinase activity (or are kinase dead), bind more tightly to wild type RAF, and often exhibit RAS activation triggered by other mechanisms (e.g. RAS mutations, NF1 deletions, growth factor signaling). Thus, class 3 BRAF mutated tumors may signal through multiple pathways, similar to RAS mutated tumors. Historically, most responses to MEK inhibitors^16,30 and newer agents (ERK inhibitors)³¹ have come in tumors with class 2 rather than class 3 mutations. Most patients in this series had class 3 mutations (n=19), potentially explaining the relative lack of activity; most patients that benefited in our study had, by contrast, class 2 mutations. Of note, the patient with the transient partial response had a class 3 BRAF mutation (G469E) as well as an NF1 inactivating mutation, but the patient with prolonged stable disease had a class 3 mutation (D594G) without a RAS or NF1 co-mutation (Supplemental Table 3), thus suggesting this is just one piece of the puzzle. One could suggest that assessing the degree of pERK expression might be a potential readout to determine MAPK signaling dependency and MEK inhibitor sensitivity (and lack of pERK as a possible marker of resistance), although this was not feasible for this study.

While trametinib did not show substantial activity in this population, newer agents to target MAPK signaling are in development. These include inhibitors of ERK, the final canonical member of the MAPK cascade, which has shown some modest clinical activity in early studies.³¹ In addition, next generation BRAF inhibitors (so-called paradox-breaker, or dimer-disrupting BRAF inhibitors) may also hold promise.³² These agents have shown promise in BRAF V600 mutant melanoma resistant to BRAF/MEK inhibitors, as well as those with non-V600 mutations and fusions in BRAF. However, no large scale studies in this population have been performed.

In conclusion, single agent trametinib had low rates of clinical activity in patients with heavily pre-treated, metastatic cancers harboring non-V600 mutations in BRAF. This contrasts with a number of case reports, largely in melanoma, showing responses in patients with these mutations. Further study might help distinguish subpopulations that benefit from trametinib. In the interim however, trametinib cannot be recommended as a single agent in patients harboring these mutations.

Statement of translational relevance.

Mutations in BRAF outside of the 600^th codon (BRAF non-V600) or BRAF fusions activate MAPK pathway signaling, and may be targetable by MEK inhibitors. We conducted a phase II study of trametinib in patients with solid tumors and lymphomas harboring BRAF non-V600 mutations or fusions to characterize the activity of this population. Overall, trametinib had low activity (3% response rate) and is thus not a recommended treatment option in this population. Additional treatment options are needed for patients harboring these genomic alterations.