Phase 1 study of safety, pharmacokinetics, and pharmacodynamics of tivantinib in combination with bevacizumab in adult patients with advanced solid tumors

Abstract

Purpose

We investigated the combination of tivantinib, a c-MET tyrosine kinase inhibitor (TKI), and bevacizumab, an anti-VEGF-A antibody.

Methods

Patients with advanced solid tumors received bevacizumab (10 mg/kg intravenously every 2 weeks) and escalating doses of tivantinib (120–360 mg orally twice daily). In addition to safety and preliminary efficacy, we evaluated pharmacokinetics of tivantinib and its metabolites, as well as pharmacodynamic biomarkers in peripheral blood and skin.

Results

Eleven patients received the combination treatment, which was generally well tolerated. The main dose-limiting toxicity was grade 3 hypertension, which was observed in four patients. Other toxicities included lymphopenia and electrolyte disturbances. No exposure-toxicity relationship was observed for tivantinib or metabolites. No clinical responses were observed. Mean levels of the serum cytokine bFGF increased (p = 0.008) after the bevacizumab-only lead-in and decreased back to baseline (p = 0.047) after addition of tivantinib. Tivantinib reduced levels of both phospho-MET (7/11 patients) and tubulin (4/11 patients) in skin.

Conclusions

The combination of tivantinib and bevacizumab produced toxicities that were largely consistent with the safety profiles of the individual drugs. The study was terminated prior to establishment of the recommended phase II dose (RP2D) due to concerns regarding the mechanism of tivantinib, as well as lack of clinical efficacy seen in this and other studies. Tivantinib reversed the upregulation of bFGF caused by bevacizumab, which has been considered a potential mechanism of resistance to therapies targeting the VEGF pathway. The findings from this study suggest that the mechanism of action of tivantinib in humans may involve inhibition of both c-MET and tubulin expression.

Trial registration

NCT01749384 (First posted 12/13/2012).

Introduction

The c-MET receptor tyrosine kinase (RTK) pathway has attracted considerable interest as a potential target for cancer therapy given its involvement in numerous hallmark features of cancer development and progression [1]. Various alterations in this signaling pathway have been implicated in tumorigenesis, including overexpression, gene amplification, and mutation [2]. Furthermore, activation of this pathway has been associated with poor prognosis and resistance to targeted therapies in various tumor types [1, 3].

A large number of agents targeting c-MET or its ligand hepatocyte growth factor (HGF) have been studied or are under current evaluation in preclinical and clinical settings [2, 4]. The most successful agents, to date, have been small molecule receptor tyrosine kinase inhibitors (TKIs), and three agents in this class have been approved by the U.S. Food and Drug Administration (FDA). These are the multityrosine kinase inhibitors cabozantinib (targets c-MET, VEGFR2, RET, FLT3, AXL, KIT, and TIE2), crizotinib (targets c-MET, ALK, ROS1, and RON), and the selective c-MET inhibitor capmatinib [4]. Tivantinib (ArQule, Inc; Daiichi Sankyo Co.) was developed as a selective, orally administered inhibitor of c-MET that is distinct from other c-MET inhibitors given its non-ATP-competitive mechanism of action that stabilizes the inactive configuration of c-MET. This type of allosteric mechanism is thought to potentially increase selectivity over the more typical mechanism of kinase inhibition that involves specific targeting of the ATP binding site, since this site is relatively conserved between many kinases [5]. Tivantinib entered clinical trials in several tumor types based on promising preclinical and initial early-phase clinical data [6].

Several lines of evidence support the potential value of targeting both vascular endothelial growth factor (VEGF) and c-MET signaling. Preclinical studies have suggested that the c-MET pathway may serve as an alternative angiogenic pathway that is upregulated in tumors treated with VEGF-directed therapies [7], and preclinical models have shown synergistic effects of targeting both signaling pathways simultaneously [8]. More recently, clinical success that is at least partially attributable to this approach is highlighted by cabozantinib, an agent that has now been approved for advanced renal cell carcinoma, hepatocellular carcinoma, and medullary thyroid cancer [9, 10] with documented activity in several other tumor types [4]. Finally, a phase 1 study of tivantinib plus sorafenib demonstrated clinical efficacy with partial tumor responses and extended stable disease in patients with advanced solid cancers such as renal cell carcinoma, hepatocellular carcinoma, and melanoma. This combination exhibited clinical activity in patients previously treated with VEGF-directed therapies [11].

To further explore the hypothesis that combined inhibition of c-MET and VEGF pathways may represent a potential therapeutic strategy, we conducted a phase 1 trial of tivantinib in combination with anti-VEGF-A antibody bevacizumab in patients with advanced solid malignancies. During the conduct of this trial, studies were published that questioned the mechanism of action of tivantinib and suggested that its preclinical activity may result from effects on tubulin causing microtubule destabilization as opposed to inhibition of c-MET signaling [12, 13]. To further address this issue, we performed pharmacodynamic studies to identify potential biomarkers of tivantinib effects in patients.

Materials and methods

Patient selection

Eligible patients were ≥ 18 years old with histologically confirmed solid tumor malignancy (excluding squamous cell carcinoma of the lung) that was metastatic or unresectable and for which standard curative or palliative measures did not exist or where no longer effective. Other key eligibility criteria included: measurable or evaluable disease by the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1 [14]; ECOG performance status 0 to 2; life expectancy greater than 3 months; ability to swallow ills; and no significant impairment in gastrointestinal absorption. Patients were required to have adequate hematologic (hemoglobin ≥ 9.0 g/dL, leukocytes ≥ 3000/μL and absolute neutrophil count ≥ 1500/μL, platelets ≥ 100,000/μL), hepatic (total bilirubin ≤ 5 × institutional Upper Limit of Normal (ULN), AST (SGOT) ≤ 2.5 × institutional ULN, ALT (SGPT) ≤ 2.5 × institutional ULN), and renal (serum or plasma creatinine ≤ 1.5 × institutional ULN or creatinine clearance ≥ 60 mL/min) function. Patients were assessed for proteinuria at baseline and during treatment; urine protein was required to be ≤ + 1 on spot urinalysis/urine dipstick, and if the value was > + 1 then a 24-h urine for protein was required to be ≤ 1 g/24 h. There were no restrictions regarding prior lines of therapy, and prior treatment with bevacizumab and inhibitors of MET or HGF was specifically allowed.

Key exclusion criteria included: treatment with chemotherapy, monoclonal antibody therapy, or radiotherapy within 4 weeks (6 weeks for nitrosoureas or mitomycin C) prior to the start of study treatment; major hemorrhagic or thrombotic events within 3 months of the start of protocol therapy, or major surgery within 6 weeks or non-healing wounds; untreated brain metastases; uncontrolled intercurrent illnesses, including inadequately controlled hypertension, defined as systolic blood pressure > 160 mmHg and/or diastolic blood pressure > 90 mmHg despite antihypertensive medication; and pregnant or nursing mothers. Investigators were cautioned to consider potential interactions with inhibitors and/or inducers of CYP2C19 and CYP3A4, although these medications were not strictly prohibited.

This study was approved by the Institutional Review Board at the University of Pittsburgh and was conducted in accordance with the Declaration of Helsinki for human subject protection. All patients provided written informed consent prior to enrollment. Both men and women of all races and ethnic groups were eligible for this trial.

Study design and treatment plan

This was a single institution, open-label, phase 1 study, registered under ClinicalTrials.gov Identifier: NCT01749384. The primary objective of the study was to determine the recommended phase 2 dose (RP2D) of the combination of bevacizumab and tivantinib in patients with advanced solid tumors. Secondary objectives were to determine the dose-limiting toxicity (DLT) and other toxicities associated with the combination as assessed by CTCAE v4.0; to document anti-tumor activity; to determine the pharmacokinetics of tivantinib when given in combination with bevacizumab; to assess the effect of the combination on plasma components of the HGF-MET signaling pathway and VEGF signaling pathway; and to assess tissue (skin) protein biomarkers before and after study treatment including MET and phospho-MET-Tyr1234.

Treatment was administered on an outpatient basis and consisted of cycles of 28 days. A lead-in dose of bevacizumab (commercially obtained) was administered on day − 15 of cycle 1 to allow assessment of plasma biomarkers on cycle 1 day 1 prior to administration of tivantinib. Bevacizumab was then continued on days 1 and 15 of each cycle starting with cycle 1, and tivantinib (crystalline B form supplied through CTEP) was administered by mouth twice a day on days 1–28 of each cycle. Tivantinib was administered orally with meals, and missed or vomited doses were not replaced. Bevacizumab was administered at a fixed dose of 10 mg/kg IV. The dose levels of tivantinib are shown in Table 1. Dose escalation followed a modified Storer’s two-stage “up and down” schema using fixed dose levels and a fixed sample size [15]. During the first stage of the study, a single patient was enrolled on each dose level starting at dose level 1. If no drug-related toxicity grade ≥ 2 non-hematologic or grade ≥ 3 hematologic was observed during cycle 1, the next subject was then enrolled in the next higher dose level. Grade 2 adverse events of electrolyte abnormalities, nausea/vomiting, or diarrhea without optimal management or hypertension would not trigger a conversion to the second stage. Of note, the toxicities that would trigger the second stage of the study were defined differently from the study DLT’s, which are described in the following section. Upon observation of a grade 2 non-hematologic or grade 3 hematologic toxicity during cycle 1, stage 2 would start at the next lower dose level. In stage 2, up to six patients were enrolled per dose level in cohorts of up to three subjects enrolled at a time. If ≥ 2 DLTs, as defined below, were observed in a cohort after enrollment of fewer than six patients, then enrollment immediately proceeded to the next lower dose level. Iterative enrollment of up to 6 patients per cohort was to be continued until a total of 40 evaluable patients had been enrolled or the rules for dose escalation indicated that enrollment was completed. No more than 12 patients were to be enrolled to a given dose level over the course of the study. Patients with DLT of hypertension were permitted to stay on study after the blood pressure was adequately controlled with medication and the dose level of tivantinib was reduced.

Table 1.

Dose escalation scheme, study enrollment, and dose-limiting toxicities

Dose level	Tivantinib (mg PO AM/PM)	Bevacizumab (mg/kg IV every 2 weeks)	Enrolled	Evaluable	# DLT (cycle 1)	DLT description (cycle 1)
− 1a	120/0	10	0	n/a	n/a	n/a
1b	120/120	10	1	1	0	n/a
2	240/120	10	1	1	0	n/a
3	240/240	10	1	1	0	n/a
4	360/240	10	1	1	1	G3 hypertension
3d	240/240	10	7	6c	2	G3 hypertension in both
2d	240/120	10	1	1	0	n/a
5a	360/360	10	0	n/a	n/a	n/a

^aNo patients enrolled at these dose levels

^bStarting dose

^cOne patient went off study prior to initiation of tivantinib

^dThese rows represent the second stage of the study

Safety and efficacy evaluation

All patients were considered evaluable for toxicity from the time of their first treatment with tivantinib and bevacizumab. NIH CTCAE v4.0 was used to grade toxicities. DLT was defined as any of the following toxicities observed during cycle 1: grade ≥ 3 non-hematologic toxicity with additional specified criteria of nausea/vomiting or diarrhea that is persistently grade 3–4 despite 48 h of optimal management; grade 3 hypomagnesemia/hypokalemia/hyperkalemia not corrected to grade 1 within 48 h; and hematologic toxicities comprising grade 4 neutropenia; febrile neutropenia; and grade 4 thrombocytopenia. The dose level at which the DLT rate was the closest to 1/6 was to be considered as the RP2D.

Patients were evaluated for response every 8 weeks via radiographic imaging and tumor measurements. In addition to a baseline scan, confirmatory scans were planned 8 (not less than 4) weeks following initial documentation of objective response. Response and progression were evaluated using RECIST version 1.1 [14]. Patients without progression on the first re-staging scan at 8 weeks were considered to have a best response of stable disease.

Pharmacokinetics

Blood was collected for tivantinib pharmacokinetic (PK) analysis during the first cycle on day 1 before and at 0.5, 1, 1.5, 2.5, 3, 4, 6, and 8 h after tivantinib administration. Trough samples were obtained on days 2 and 15. Vacutainers were pre-chilled and wrapped in aluminum foil to protect tivantinib, which is light-sensitive. Plasma was obtained by centrifuging for 10 min at 1500×g at 4 °C and stored at − 70 °C or colder until analysis. Tivantinib plasma concentrations were analyzed by validated LC–MS/MS analyses as previously described [16], and tivantinib metabolites M4845, M4846, M5991, and M6385 were also analyzed (Supplemental Fig. 1; [17]). Assay range, accuracy, and precision per analyte was as follows: tivantinib (20–10,000 ng/mL; 96.5–104.5%; < 4.5%), M4845 (5–1000 ng/mL; 97.5–104.0%; < 11.3%), M4846 (5–3000 ng/mL; 92.4–104.0%; < 6.4%), M5991 (5–3000 ng/mL; 96.0–99.3%; < 8.0%), and M6385 (5–1000 ng/mL; 95.9–98.0%; < 7.1%). Plasma pharmacokinetic parameters were extracted from the data by non-compartmental methods with PK Solutions 2.0™ (Summit Research Services, Montrose, CO, USA). Exposure response analysis was performed by exploring the difference in C_max and AUC values of tivantinib and metabolites between patients with grade 2–4 toxicities vs those without toxicity.

Pharmacodynamics

Preclinical immunoblot experiments of α- and β-tubulin expression in human RKO colon cancer cells and the tubulin polymerization assay were performed as previously described [12, 18]. With respect to the trial-related assessments, tissue samples (plasma, PBMCs, and skin biopsies) were obtained from patients prior to starting treatment (cycle 1, day − 15), 15 days post-bevacizumab administration (pre-dose on cycle 1, day 1), and 15 days post-bevacizumab/daily tivantinib administration (cycle 1, day 15). HGF, HGFA, and VEGF family members A, B, C, D, and PlGF were quantified in plasma using Meso Scale Discovery (Cat#K15190D), Luminex (HGF; Cat#LHC6003M), and ELISA (VEGF-B; Cat#SEA144Hu; Cloud-Clone Corp.) platforms. Median basal levels of these growth factors were obtained from reference data. For VEGF-A, VEGF-C, VEGF-D, PIGF, bFGF, VEGFR1, the reference values from EDTA plasma were provided in the product insert of the corresponding assay kit from Meso Scale Discovery [19]. The reference values for VEGF-B and HGF, respectively, have been previously described [20, 21]. Wilcoxon signed rank tests were used to compare the marker levels between two different timepoints, with p < 0.05 considered significant. For analysis of skin biopsy tissue biomarkers, biopsies of skin, typically from locations of the back, were performed in the office and tissues were placed immediately in formalin for fixation overnight. On the following day, tissues were transferred to 70% ethanol and subsequently embedded in paraffin for sectioning. The slides are deparaffinized, rehydrated, and antigen retrieval was performed using a citrate buffer (Dako, Carpinteria, CA, USA) at 123 °C. The slides were stained using an Autostainer Plus (Dako) platform with TBST rinse buffer. The beta-tubulin antibody (rabbit monoclonal—9F3, Cell Signaling, Danvers, MA) was applied using a 1:200 dilution. The Met antibody (rabbit monoclonal—D1C2, Cell Signaling) was applied using a 1:300 dilution. The p-Tyr1234/5-Met antibody (rabbit monoclonal—D26, Cell Signaling) was applied using a 1:200 dilution. The goat-anti rabbit secondary antibody consisted of Envision Dual Link + (Dako) HRP polymer. The substrate used was 3,3,diaminobenzidine + (Dako). Lastly, the slides were counterstained with hematoxylin.

Results

Patient characteristics

Twelve patients were enrolled between February 2013 and December 2014. The baseline characteristics of these patients are presented in Table 2. Of note, although eligibility criteria technically included patients with ECOG performance status 0–2, only patients with performance status 0–1 were actually enrolled on the trial. Additional patient information including diagnosis for each patient and prior treatments is shown in Supplemental Table 1.

Table 2.

Summary of patient characteristics

	Patients (n)
Patients enrolled	12
Patients treated with both tivantinib and bevacizumab	11
Age at study entry (years)
Median	61.5
Range	29–82
Gender
Female	2
Male	10
Race
Caucasian	12
ECOG performance status
0	7
1	5
Cancer
Colorectal adenocarcinoma	4
Mesothelioma	2
Non-small cell lung	1
Renal cell carcinoma, papillary	1
Alveolar soft part sarcoma	1
Myxofibrosarcoma	1
Leiomyosarcoma	1
Paraganglioma	1

Dose escalation and toxicity

A summary of the dose escalation is provided in Table 1. During stage 1 (fast dose escalation), one patient each was treated in dose levels 1, 2, and 3 before the first dose-limiting toxicity (DLT, grade 3 hypertension) was recorded in the patient on dose level 4. The study then proceeded to stage 2 and enrolled seven additional patients on dose level 3, of which one patient was taken off study before receiving tivantinib due to adverse events that were possibly related to his dose of bevacizumab on day − 15. Of the evaluable patients in dose level 3, 2/6 had DLT’s in the first cycle (grade 3 hypertension in both cases). The next patient was then enrolled on dose level 2. No patients were subsequently enrolled, as the study was prematurely terminated given concerns about lack of clinical efficacy in this and other clinical studies of tivantinib, as well as uncertainty regarding the mechanism of action of tivantinib.

Adverse events observed in all administered cycles are summarized in Table 3. The most common serious (grade 3 or higher) non-hematologic toxicity was hypertension, which was seen at any grade in 6 of 11 evaluable patients. One patient experienced a grade 3 thrombosis (inferior vena cava filter-related deep venous thrombosis). In terms of hematologic toxicities, lymphopenia was seen in 6 of 11 patients overall, with 2 patients experiencing grade 3 toxicity. Other hematologic toxicities were mild and included anemia (grade 1 in six patients, grade 2 in one patient), and leukopenia (grade 1 in one patient, and grade 2 in one patient). Hypophosphatemia was seen in 5 of 11 patients, with 4 of 11 patients experiencing grade 1 toxicity. Finally, one patient experienced grade 4 cardiotoxicity with complete heart block and hypotension one week after being taken off study due to disease progression. Although the patient had pre-existing cardiac issues and the toxicities were recorded as unlikely to be related to either drug, the event was reported as a serious adverse event and is, therefore, included herein.

Table 3.

Grade ≥ 3 toxicities at least possibly related to treatment in all cycles (for a specific toxicity, if a grade ≥ 3 toxicity was observed, then Grades 1 and 2 toxicities were also summarized)

Toxicity	Grades 1 and 2	Grade 3	Grade 4
Headache	3 (27%)	1 (9%)
Hypertension	2 (18%)	4 (36%)
Hypomagnesemia	1 (9%)	1 (9%)
Hypophosphatemia	4 (36%)	1 (9%)
Lymphocytopenia	4 (36%)	2 (18%)
Thromboembolic event		1 (9%)
Atrioventricular block completea			1 (9%)
Hypotensiona			1 (9%)
Any toxicity	11 (100%)	7 (64%)	1 (9%)

Number of patients experiencing adverse event (percentage, out of 11 evaluable patients). Each toxicity is considered separately, such that a given patient may appear more than once

^aConsidered unlikely related to treatment but was reported as a serious adverse event

Antitumor activity

No objective responses were observed in this study. Of the 11 patients who received the combination of tivantinib and bevacizumab, 6 patients experienced stable disease as their best response, while 5 patients experienced progressive disease. The median number of cycles of therapy received was 3, with a range of 0 to 20. Four patients received at least 6 months of treatment. Nine patients were taken off of study due to disease progression on radiologic imaging, one patient stopped therapy after a 19.5% increase in the size of lesions on imaging and worsening of disease-related symptoms, and one patient decided to switch to an alternative treatment.

Of note, one patient with papillary renal cell carcinoma who had progressed on prior VEGF pathway-directed therapies including sunitinib, pazopanib, and axitinib was found to have a reduction in his target lesions as determined by imaging that did not meet the criteria of objective response. However, this patient had an extended duration of stable disease during which he received 20 cycles of therapy. One patient with metastatic colorectal cancer who had previously progressed on a regimen containing bevacizumab received eight cycles of treatment with tivantinib and bevacizumab before experiencing progressive disease. Three other patients who had previously progressed on treatments targeting the VEGF pathway (bevacizumab or pazopanib) had a best response of progressive disease during the current study.

Pharmacokinetics

PK data were available for all 11 patients who received both tivantinib and bevacizumab (see Table 4 and Supplemental Tables 2–6). A typical patient concentration vs time profile is presented in Supplemental Fig. 2. Tivantinib concentrations were higher than metabolite concentrations, but metabolite levels appeared to persist while tivantinib was being eliminated from plasma. Tivantinib concentrations peaked at about 3 h while metabolites peaked between 3 and 7 h. No apparent relationship between C_max or AUC values and toxicity was observed (data not shown), although given the small sample size a relationship cannot be excluded.

Table 4.

Tivantinib mean (SD) plasma pharmacokinetic parameters

Tivantinib (mg AM/PM)	Cmax (μg/mL)	Tmax (h)	t½ (h)	AUC0–8 (μg/mL h)	Cl/Fa (L/h)	Cmin D15/D1	AUCratio M4845	AUCratio M4846	AUCratio M5991	AUCratio M6385
120/120 (N = 1)	0.93	2.5	8.2	5.57	9.43	2.85	0.069	0.16	0.025	0.032
240/120 (N = 2)	1.53 (0.05)	2.0 (0)	3.3 (0.6)	7.46 (0.60)	25.1 (1.0)	NA	0.23	0.35	0.11	0.087
240/240 (N = 7)	1.37 (0.46)	3.1 (0.9)	4.6 (1.9)	6.88 (2.37)	27.0 (17.7)	2.64 (1.78)	0.14	0.32	0.12	0.098
360/240 (N = 1)	1.27	3.0	9.2	5.02	25.8	NA	0.091	0.26	0.12	0.069
Total (N = 11)	–	2.9 (0.8)	5.1 (2.4)	–	24.9 (14.6)	2.67 (1.69)b	0.14 (1.7)	0.30 (1.4)	0.098 (1.7)	0.084 (1.78)

Accumulation ratio expressed as C_min D15/D1 is reported as geometric mean and standard deviation of the ratio. C_min D1, C12h after first dose was calculated based on C_8h and half-life. Expected accumulation ratio based on dose interval τ = 12 and individual half-lives was 1.27-fold (1.17) (geometric mean and SD). Observed accumulation ratios were 2.10-fold (1.83) more than predicted based on day 1 PK (geometric mean and SD). AUC ratios were calculated as geometric means of metabolite/tivantinib AUC_0–8 with associated geometric standard deviations (used multiplicative)

SD standard deviation

^aBased on AUC_0–inf, which had an average 38% extrapolated beyond AUC_0–8

^bBased on N = 7

Pharmacodynamics

Given data suggesting that tivantinib may exert its biological activity via microtubule destabilization and not through inhibition of c-MET signaling [12, 13], we hypothesized that tivantinib treatment may result in loss of tubulin expression and that analysis of tubulin protein levels may be a biomarker of tivantinib activity. In preliminary in vitro experiments, we demonstrated that treatment of human colon cancer RKO cells with tivantinib resulted in loss of both α- and β-tubulin protein expression, similar to vincristine, a known microtubule destabilizer (supplemental Fig. 3A). We also demonstrated that tivantinib inhibited tubulin polymerization, but not to the same extent as vincristine (supplemental Fig. 3B). This observation has been previously reported in other human cancer cell lines [12]. These findings suggest that loss of tubulin protein expression may represent a potential biomarker of tivantinib’s mechanism of action. Total β-tubulin expression was evaluated in patient PBMCs by immunoblot analysis. No significant differences were observed in tubulin expression following tivantinib treatment (data not shown). However, immunohistochemical (IHC) analysis of skin biopsies from patients receiving oral tivantinib for two weeks demonstrated substantial reductions in tubulin expression in 4 of 11 patients (Fig. 1). In six of the remaining patients, there was no apparent change in tubulin expression, and in one patient tubulin expression appeared to increase slightly (see Supplemental Fig. 4 for representative images for every patient). The changes in tubulin expression were not related to dose level in these patients, although sample size was too small to evaluate this fully. Suppression of activated p-MET expression in the nucleus was evident after tivantinib treatment (Fig. 1), while total MET expression in the cytoplasm remained unchanged. Previous studies have shown that activated (i.e. phosphorylated) MET migrates to the nucleus whereas unactivated MET expression remains plasma membrane- and cytoplasm-associated [22, 23]. Expression of p-MET was reduced in 7 of 11 patients following tivantinib and stable in the rest.

Fig. 1.

Representative effects of combination treatment on protein expression in skin. Biopsies were obtained at C1D-15, C1D1, and C1D15, formalin-fixed, paraffin-embedded and slides were stained with indicated antibodies. Panels A and B represent two different patients. The bar represents 100 µm, and part B has the same magnification as part A

Growth factors and cytokines were measured in patient plasma. bFGF levels increased in 8 of 11 patients following bevacizumab treatment (p = 0.008, Fig. 2A, see Supplemental Table 7 for mean values and p values for all comparisons). On day 15 of the treatment cycle following 14 days of tivantinib treatment, concentrations of bFGF returned to baseline in most patients (p = 0.047). Placental growth factor (PIGF) concentrations increased in 10 of 11 patients by an average of 60% after bevacizumab treatment (p = 0.005, Fig. 2B). Addition of tivantinib did not alter PIGF concentrations. Three patients had HGF concentrations above 300 pg/mL at baseline but no responses were observed (Fig. 2C). After bevacizumab treatment, HGF concentrations increased in six patients. With the addition of tivantinib, HGF concentrations decreased in these same patients. With regard to soluble VEGFR1 (sVEGFR1), five patients had elevated plasma concentrations at baseline compared to normal volunteers (Fig. 2D, references for normal values provided in methods). These concentrations decreased after bevacizumab administration. Following tivantinib administration, sVEGFR1 concentrations in three of these five patients remained low. However, sVEGFR1 concentrations following tivantinib increased in four patients, including those patients with low basal expression levels at baseline. VEGF-A was detected in all patients at concentrations higher than in normal volunteers (Fig. 2E). Bevacizumab treatment resulted in decreased expression of VEGF-A in two patients (#01; #07) and increases in eight patients. The addition of tivantinib to bevacizumab did not alter VEGF-A concentrations. VEGF-B, VEGF-C, and VEGF-D were detected in all patients. VEGF-D plasma concentrations were lower than the reported reference value (Fig. 2H). This may be due to our use of platelet-poor plasma for the analysis, which can lower cytokine concentrations [24]. No effects of treatment on VEGF-B, VEGF-C, and VEGF-D expression were observed (Fig. 2F–H).

Fig. 2.

Effect of tivantinib/bevacizumab combination on levels of growth factors. Plasma was obtained at day − 15, day 1, and day 15 from each patient. Various growth factors were measured by Meso Scale Discovery, Luminex, or standard ELISA platforms. The dotted line on the y-axis indicates median basal levels of each growth factor, with references provided in methods section

Discussion

This phase 1 clinical trial was designed to assess safety, tolerability, and preliminary efficacy of the combination of tivantinib and bevacizumab. During the conduct of this trial, two major lines of evidence were published that suggested that this combination might not be viable. First, as noted above, preclinical studies suggested that tivantinib may actually exert its effects through microtubule destabilization as opposed to inhibition of c-Met signaling. Second, phase 3 studies were conducted in patients with hepatocellular carcinoma [25, 26] and non-small cell lung cancer [27, 28] that showed disappointing efficacy results. Although the phase 1 study reported herein was not powered to assess efficacy, the lack of clinical responses was consistent with the results of these other studies of tivantinib. For these reasons, the decision was made to terminate this phase 1 study before the RP2D could be established.

In terms of the primary clinical endpoint of safety, the observed adverse events were generally consistent with those of the individual drugs. The three patients with DLT’s in cycle one all had grade 3 hypertension, which in all cases resolved with temporary holding of the study medications and initiating (in two patients with no pre-existing hypertension) or increasing (in the third patient) antihypertensive medications. Another patient had an inferior vena cava filter-associated venous thromboembolism. Both of these are well-established side effects associated with bevacizumab, and it is difficult to quantify any contribution of tivantinib. Mild cytopenias and electrolyte abnormalities, including hypophosphatemia, have been observed in previous studies of tivantinib [11, 16].

The pharmacokinetics of tivantinib in combination with bevacizumab were similar to what has been previously reported for tivantinib [29]. The pharmacokinetics of tivantinib appear to be non-linear when evaluating the relationship between dose and exposure parameters such as AUC_0–8/D, and C_max/D, which had been previously alluded to without explicit data in an earlier report [30]. However, based on our data, clearance (based on AUC_0–inf) does not appear to be dose-dependent, and possibly with higher doses, the C_max occurs later and is lower than expected, resulting in more of the AUC_0–inf shifting to times beyond that captured in AUC_0–8. Based on the ratio of tivantinib day 15 C_min and day 1 12 h, tivantinib accumulation ratio was approximately 2.67-fold. This level of accumulation is higher than what was presented in a previous report where C_max/D and AUC_0–12/D appeared to accumulate 1.25-fold, mostly due to a higher reported exposure on day 1 of 9.27 ng/mL/mg relative to our current 5.82 ng/mL/mg [29]. In our study, the accumulation ratio between the day 2 and day 15 C_min of 1.06 combined with the 2.67-fold accumulation between day 1 and day 15 suggests that any accumulation has reached steady-state within 24 h. This finding is as expected given that this time frame represents approximately 4–5 times the half-life of 5 h. However, based on a dose interval of 12 h and this half-life, the expected accumulation ratio is only 1.27-fold. This finding suggests that clearance decreases approximately twofold soon after the first dose. Such a decrease in clearance might explain why our 1st dose apparent clearance of 24.9 L/h is twofold higher than the day 5 steady-state clearance of 11.8 L/h previously reported [31]. We suspect tivantinib may inhibit its own metabolism, and while it was reported not to be a strong direct CYP450 inhibitor with IC₅₀ values 10 µM or greater (~ 3700 ng/mL), tivantinib exhibits time-dependent inhibition of CYP3A4 at 50 µM (~ 18 µg/mL) ([32], ArQule data on file) concentrations that could easily be reached locally in the gut and liver during the first-pass. The inability to detect an impact of temsirolimus (a weak CYP3A4 inhibitor) on tivantinib PK in an earlier report was presumably because temsirolimus was added and the interaction was assessed after a week of initial tivantinib dosing, when CYP3A4 had already been inhibited by tivantinib [29].

To our knowledge, this study is the most extensive analysis conducted, to date, of the pharmacokinetics of the tivantinib metabolites M4845, M4846, M5991, and M6385. The pharmacological activity in vitro of these four metabolites is 3.3- to 8.4-fold less potent than the parent compound tivantinib and because, when taken together, they represent an additional 50% relative to tivantinib AUC on day 1, their pharmacological contribution relative to tivantinib is expected to be small [33]. Apart from M4845, the T _max values of the metabolites occur later than that of tivantinib, in line with their apparent longer half-lives. Individual metabolite AUC_0–8 values were 10–30% of parent tivantinib AUC. However, as can be seen from the representative PK profiles, while tivantinib elimination phase was mostly captured in the sample interval of 0–8 h after dosing, that of the metabolites was not, suggesting metabolites might accumulate more than tivantinib with chronic dosing. The D15/D2 plasma concentration ratio was 1.06 for tivantinib and 1.18, 1.21, 1.61, and 1.21 for metabolites M4845, M4846, M5991, and M6385, respectively. These ratios are likely less than expected based on their day 1 profile, which may be explained by the decrease in tivantinib clearance as detailed above. Of note, M5991 accumulates most after day 2, perhaps because the formation, but not the clearance, of this metabolite is mediated by CYP3A4/5 [17].

One of the preclinical studies suggesting that tivantinib was actually a cytotoxic agent commented that the initial preclinical effects of tivantinib on c-MET were present under very specific conditions (only if the drug is preincubated with dephosphorylated MET in the presence of ATP). The authors of this study questioned whether tivantinib would actually be able to inhibit c-MET kinase activity in human tumors given that it could not induce autophosphorylation of c-MET in any of the cell lines tested [13]. Our pharmacodynamic studies suggest that, at least in human skin, tivantinib appears to act on both tubulin and c-MET. It should be noted that while tubulin levels were clearly decreased in four patients, this effect was not seen in the majority of treated patients, which may limit the generalizability of this finding. Unfortunately, we were unable to perform serial analysis of tubulin or phospho-MET levels in tumor samples, so it is unclear whether the biomarker findings in skin are truly reflective of changes in tumor tissue. Additionally, the immunohistochemical analysis done in our study was qualitative, and it is unclear whether MET was inhibited to a sufficient degree to produce clinically meaningful effects.

Analysis of serum biomarker levels before and during tivantinib treatment revealed several notable findings. First, several investigators suggested that bFGF elevation could represent a resistance mechanism to anti-angiogenic therapy [34, 35]. In our study, bFGF levels were elevated in 8 of 11 patients following the lead-in treatment with bevacizumab and normalized after tivantinib was added. In an exploratory analysis, both changes were statistically significant. These findings support the concept that tivantinib may prevent resistance to VEGF-targeted therapies through this mechanism. Second, the increase in PlGF observed after anti-angiogenic treatment is also consistent with what has been previously observed with other anti-VEGF therapies [36]. Third, a recent phase 3 study showed that high plasma levels of HGF may predict response to tivantinib-based therapies [27]. In contrast, other studies were unable to identify such a correlation [37]. HGF has been suggested as another biomarker that becomes elevated prior to progression on bevacizumab and may contribute to resistance [35]. In our study, HGF levels were elevated in six patients following bevacizumab treatment and subsequently normalized after administration of tivantinib. However, given the lack of clinical responses, it is difficult to determine the significance of this finding. Fourth, the increase in sVEGFR1 levels in several patients after treatment with tivantinib may be related to tivantinib’s anti-microtubule effect as Meissner et al. have shown that microtubule-targeting agents inhibit VEGFR2 expression [38]. Finally, although the mean values were essentially stable, serum levels of VEGF-A were found to increase in several patients after treatment with bevacizumab, which is consistent with previous reports. This increase is thought to predominantly represent VEGF complexed with antibody [39].

There may be several potential reasons for the lack of clinical efficacy of tivantinib. Clearly, if the agent does not primarily work through c-MET, its use alone or in rationally designed combinations targeting c-MET-dependent tumors would not be expected to be effective. Second, resistance to MET-directed therapy can emerge through various mechanisms, and alternative or combination approaches may be needed to overcome the development of cellular drug resistance [40]. Third, given the fact that we did see evidence of c-MET inhibition in human skin, it is conceivable that the degree of c-MET inhibition at the dose levels of tivantinib used in our study was simply not sufficient to produce a clinically relevant effect. Finally, it is not yet clear that targeting c-MET can be clinically effective except in carefully selected patient populations [3]. Therefore, it will be important to further refine the use of molecular biomarkers that will predict response to c-MET inhibition. Improved patient selection has shown success in non-small cell lung cancer, where patients with exon 14 skipping mutations or patients who have MET upregulation after other targeted therapies are thought to be most likely to benefit [2]. Of note, in our phase 1 study, patients were not selected based on any specific molecular features, and this lack of patient selection may have further reduced our ability to observe any clinical responses.

In conclusion, the findings of this study do not support the further clinical development of the combination of tivantinib and bevacizumab in patients with advanced cancer. The idea of combining anti-VEGF treatment with c-MET inhibition may still be worth pursuing albeit with a more effective c-MET inhibitor and with careful selection of patients most likely to benefit.