Phase II Trial of Cabozantinib in Recurrent/Metastatic Endometrial Cancer: A Study of the Princess Margaret, Chicago and California Consortia (NCI9322/PHL86)

Abstract

PURPOSE

The relevance of the MET/Hepatocyte Growth Factor (HGF) pathway in endometrial cancer tumor biology supports the clinical evaluation of cabozantinib in this disease.

PATIENTS & METHODS

PHL86/NCI#9322 ( NCT01935934) is a single arm study that evaluated cabozantinib in women with endometrial cancer with progression after chemotherapy. Co-primary endpoints were response rate and 12-week progression-free-survival (PFS). Patients with uncommon histology endometrial cancer (eg carcinosarcoma and clear cell) were enrolled in a parallel exploratory cohort.

RESULTS

A total of 102 patients were accrued. Amongst 36 endometrioid histology patients, response rate was 14%, 12-week PFS rate was 67% and median PFS, 4.8 mths. In serous cohort of 34 patients, response rate was 12%, 12-week PFS was 56% and median PFS 4.0 mths. In a separate cohort of 32 patients with uncommon histology EC (including carcinosarcoma), response rate was 6% and 12-wk PFS was 47%. Six patients were on treatment for > 12months, including 2 for > 30 months. Common cabozantinib-related toxicities (> 30% patients) included hypertension, fatigue, diarrhea, nausea and hand-foot syndrome. Gastrointestinal fistula/perforation occurred in 4 of 70 (6%) patients with serous/endometrioid cancer and 5 of 32 (16%) patients in exploratory cohort. We observed increased frequency of responses with somatic CTNNB1 mutation (4 PRs in 10 pts, median PFS 7.6 mths) and concurrent KRAS and PTEN/PIK3CA mutations (3 PRs in 12 patients, median PFS 5.9 mths).

CONCLUSIONS

Cabozantinib has activity in serous and endometrioid histology EC. These results support further evaluation in genomically characterized patient cohorts.

INTRODUCTION

Endometrial cancer (EC) is the most common gynecologic cancer affecting North American women. Approximately 30% of women with EC present with advanced disease, and those with recurrence or distant metastases have a poor prognosis, with 5-year survival rates of less than 20%. (1) Platinum-based chemotherapy with paclitaxel and/or doxorubicin remains the standard first-line treatment. (2-4) Unfortunately, tumor response rates are below 50% and duration of disease control is under a year. Responses to 2^nd line chemotherapy are transient, with median progression-free-survival (PFS) of 3 to 4 months. Although several novel cytotoxics and targeted agents have been evaluated, none has demonstrated sufficient activity to gain regulatory approval, and treatment options for women with recurrent disease remain limited. (5, 6)

The association of high microvessel density and pro-angiogenic gene expression with aggressive biology and inferior outcome across different histotypes of EC, provides biologic rationale to target angiogenic pathways across the histological spectrum of this disease (7-10). Several approaches have been evaluated with mixed results. Modest efficacy has been observed with both bevacizumab and cediranib, with response rates of approximately 13% and 6-month PFS rates of 40% and 29% respectively (6, 11); while aflibercept, nintedanib and sunitinib all had insufficient activity in early studies for further development. (12-14) A range of mechanisms have been implicated in intrinsic and acquired resistance to vascular endothelial growth factor (VEGF)-targeting, including activity through various growth factor pathways and epithelial/stromal cross-talk through MET/hepatocyte growth factor (HGF), Tie2 and RET. (9, 10, 15-17) The recognized relevance of growth factor pathways in promoting angiogenesis provides the basis for studying combination strategies simultaneously directed against VEGF and other mitogenic pathways. One such approach of bevacizumab with temsirolimus, demonstrated impressive efficacy but significant toxicity and hasn’t moved forward into further evaluation. (18)

Cabozantinib is a multi-targeted tyrosine kinase inhibitor with potent activity against MET (HGF receptor), VEGFR2, RET and AXL. It has already demonstrated anti-tumor activity in medullary thyroid, (19, 20) renal (21-23) and hepatocellular carcinoma (24). NCI9322/PHL86 was developed to explore the efficacy and toxicity of dual targeting of mitogenic and angiogenic signaling using cabozantinib in women who had previously received chemotherapy for advanced EC.

METHODS

Study Design and Participants

NCI9322/PHL86 ( NCT01935934) was a non-randomized, multi-center, trial of cabozantinib in advanced EC patients, with participation of 12 North American centers of the Princess Margaret, California and Chicago Phase II Consortia. Eligible patients were ≥ 18 years old, had received one prior line of chemotherapy for metastatic disease, or had recurrence within a year of adjuvant, platinum-based chemotherapy. Patients with endometrioid and serous adenocarcinomas were eligible for accrual to the experimental cohort and uncommon histology types (including carcinosarcoma, clear cell etc.) to a parallel exploratory cohort. Prior hormonal therapy was allowed, as were targeted agents not directed against MET or VEGF/VEGFR; prior bevacizumab treatment was excluded. All patients required histological diagnosis with available archival samples, measurable disease by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1., Eastern Cooperative Oncology Group performance status ≤ 2 and an estimated life expectancy of ≥ 3 months. Normal organ and marrow function had to be confirmed within 7 days of 1^st dose of cabozantinib.

Patients on therapeutic anticoagulation with warfarin, heparin, factor Xa inhibitors or antiplatelet agents were excluded; prophylaxis with low-dose aspirin (≤ 81 mg/day), warfarin (≤ 1 mg/day) or heparin was allowed. In May 2016 inclusion criteria were expanded to allow therapeutic anticoagulation with low molecular weight heparin (LMWH).

Other ineligibilities included gastrointestinal (GI) bleeding requiring intervention within prior 6 months, pulmonary hemorrhage within prior 3 months, endotracheal/bronchial tumor, or tumor invading GI tract or invading/in contact with major blood vessels, and active brain metastases or epidural disease. Conditions with increased risk of GI fistula/perforation (e.g. inflammatory bowel disease, mucosal metastases, prior bowel obstruction, fistula/perforation or intra-abdominal abscess) were also excluded. Those with uncontrolled intercurrent/recent illness, pregnancy, uncontrolled hypertension (systolic blood pressure (BP) > 140 mmHg or diastolic BP > 90mmHg with optimal anti-hypertensive treatment) and active hepatitis were not eligible.

Previous adequately treated, basal or squamous cell cancer and superficial bladder cancer were allowed, as were other prior cancers provided patient had received definitive therapy and maintained disease-free status for at least 3 years.

Major or minor surgeries had to be completed within 3 and 1 month of initiation of cabozantinib respectively, and radiation to thorax, abdomen or pelvis, prior to 3 months. A wash-out period of 14 days (or 5 half-lives of active agent) was required for patients previously treated with small molecule kinase inhibitors or hormonal therapies, 3 weeks for cytotoxic therapy including investigational cytotoxics or biologics, and 6 weeks for nitrosoureas or mitomycin C. All other investigational agents needed to be discontinued 4 weeks prior.

Study was conducted in accordance with provisions of the Declaration of Helsinki and Good Clinical Practice guidelines, with protocol being approved by the institutional ethics review board at each study site. All patients provided written informed consent prior to any study-related procedures.

Treatment

Cabozantinib was administered orally, 60 mg daily, on a 28-day cycle. Treatment continued until disease progression, unacceptable toxicity, consent withdrawal or death. Initial study design included dose escalation to 80mg after cycle 1 in patients with stable or progressive disease, and ≤ Grade 1 toxicities. At stage I analysis, study was amended to remove dose escalation option due to a lack of tolerability.

Response assessment was per RECIST (Response Evaluation Criteria in Solid Tumors) v1.1 after cycle1 and 3, subsequently every 8 weeks irrespective of missed or held doses. Patients enrolled after amendment removing dose escalation option, had their first response evaluation after cycle3.

All patients who initiated treatment with cabozantinib were evaluable for safety and toxicity from first treatment dose. Adverse event grading was per the Common Terminology Criteria for Adverse Events (CTCAE) v 4.0.

Endpoints

Study had co-primary endpoints of investigator-assessed 12-week progression-free-survival (PFS) and response rate (RR) per RECIST v1.1.

Statistical Design

In initial design, patients were enrolled to an experimental cohort of serous, endometrioid and mixed histology EC, employing a standard Simon two-stage design to evaluate for co-primary endpoints of objective RR and 12-week PFS. Target accrual in stage I was 24 pts, and observation of ≥ 4 PRs or ≥ 8 patients with 12-week PFS, was required to proceed to second stage to accrue a total of 42 pts. The parallel exploratory cohort of uncommon histology cancers would be analyzed independently.

At stage I review, with observed activity across several histologic types and brisk patient accrual, design was modified for stratification by histological subtype (serous vs endometrioid) with mixed histology EC analyzed in exploratory cohort. In amended trial design, 18 patients each of serous and endometrioid subtype were enrolled in stage I, and observation of ≥ 3 responses or ≥ 7 patients with 12-week PFS would be sufficient to proceed to target accrual of a total of 36 patients per cohort. Observation of ≥ 8 PRs or ≥ 16 patients with 12-week PFS in a cohort of 36 patients, would be considered a signal of clinically relevant activity with each cohort being reviewed independently.

Modified trial design discriminated between co-primary endpoints of objective RR of 30% (vs 10%) and 12-week PFS of 55% (vs 30%). The design had 86% power to detect a true objective RR of at least 30% and at least 90% power to detect a true 12-week PFS rate of at least 55% (or a median PFS of 3.4 months).

Patients with uncommon histology cancers (clear cell, carcinosarcoma, adenosquamous etc.) were treated in a separate exploratory cohort with planned accrual of 30 patients. Observation of > 4/10 objective responses in a given subtype was considered activity to warrant further evaluation.

Kaplan-Meier analysis and log rank test were used to compare PFS across subgroups with different mutation status and Fisher’s exact test was used to test association between mutation status and response. Since these analyses were exploratory there was no correction for multiple comparisons.

This trial was registered with ClinicalTrials.gov, NCT01935934

Molecular analyses.

All patients consented to molecular profiling of archival tumor tissue. This included somatic and germline variant profiling on formalin-fixed paraffin-embedded (FFPE) samples using the Next Generation Sequencing (NSG) TruSeq Amplicon Cancer Panel on MiSeq platform (1 patient had Sequenom MassArray profiling). Gene variants included in these panels are included in Supplemental Table 1 and variant classification was completed as per OncoKB. Profiling was conducted in the CLIA-certified research laboratory at the Princess Margaret as described previously (25, 26). Fluorescent in situ hybridization (FISH) for MET amplification was performed on FFPE sections using a dual-colour DNA FISH probe (Vysis MET Spectrum Red FISH Probe Kit and CEP7 AlphaSatellite DNA Spectrum Green; Abbott Molecular). Fields with at least 50 tumor cells were captured to analyze 100 non-overlapping tumor cell nuclei to calculate a ratio of MET to CEP7 signal; MET amplification was defined as MET/CEP7 ratio ≥ 2. Specimens for FISH were processed at the Princess Margaret Drug Development Program Biomarker Laboratory.

Role of Funding Source

The study was funded through the N01 Phase II consortium contract (HHS N261201100032C). Molecular analyses were funded independently through grant funding from Ontario Institute of Cancer Research (OICR) and Princess Margaret Cancer Centre Foundation.

RESULTS

Between May 2013 and November 2016, 103 patients were screened, 1 patient was deemed ineligible and 102 patients were registered. This included 70 patients in the experimental cohort and 32 in the rare histology, exploratory cohort. Analysis presented is as of April 2019 and all patients are now off trial. Trial profile is outlined in Fig 1.

Fig 1. CONSORT diagram.

Outlines patients enrolled in both experimental and exploratory cohorts.

* Identifies patients who discontinued study treatment prior to formal response evaluation (i.e. Tumor Response Assessment or TRA): in exploratory cohort this included 6 pts, 5 of which discontinued treatment due to adverse event (AE) and 1 withdrew consent; in endometrioid cohort, this included 3 patients with AEs.

“Off treatment” box outlines reasons why patients discontinued study with “PD”: progressive disease; “AE”: adverse event and “other” including withdrawal of consent, physician preference and inter-current illness.

All patients who initiated treatment were considered evaluable for both treatment response and toxicity. This included 4 patients with violations of inclusions/exclusion criteria; > 2 lines chemotherapy (n = 2; both in serous cohort), and uncontrolled hypertension at baseline (n = 2; one each in endometrioid and exploratory (carcinosarcoma) cohorts respectively).

Baseline characteristics of all patients accrued are outlined in Table 1. Median age was 64 years (range: 25 to 84). The majority of women (98%) were good performance status (≤ ECOG 1). Sixty one percent had prior pelvic radiotherapy and all patients had received prior chemotherapy, (80%: 1 line, 20% ≥2 lines).

Table 1.

Baseline Characteristics of all enrolled patients

	TOTAL PATIENTS ACCRUED n = 102
	EXPERIMENTAL n = 70		EXPLORATORY n = 32
	ENDOMETRIOID N = 36	SEROUS N = 34	CARCINOSARCOMA N = 19	OTHER N = 13
Median Age (range)	61 (38 – 84)	68 (39 – 77)	64 (25 – 75)	63 (35 – 74)
ECOG PS
0	22	16	10	6
1	13	17	9	7
2	1	1
Chemotherapy
1 line	33	23	16	10
2 lines	3	9	3	3
> 2 lines	0	2	0	0
Pelvic radiotherapy	23	19	11	9
Yes	13	15	8	4
No
Histological Details
Grade 1	5	0	0	0
Grade 2	14	0	0	0
Grade 3	14	0	0	0
unknown	3	34	19	13

Experimental cohort outcomes

In endometrioid cohort, we observed 5 partial responses (14%) and 24 patients (67%) achieved 12-wk PFS. Median PFS was 4.8 months (95% CI: 4.4 to 6.4 months) with a 6-month PFS rate of 43% (95% CI: 27 to 59%). With the caveat that study did not include a centralized expert pathology review, median PFS in women with Grade 1 tumors was 6.4 months (4.7 months – NR), with Grade 2 was 5.5 (2.5 – 10.8) months, and Grade 3, 4.3 (1.0 – 4.8) months. Amongst the serous histology patients, there were 4 PRs (12%) and 19 patients (56%) met the 12-wk PFS endpoint. Median PFS was 4.0 months (95% CI: 2.8 to 4.7 months) with a 6-month PFS of 30% (16 to 47%). Of all patients meeting the 12-wk PFS, more than half (23 of 43 patients) remained progression free at their subsequent (20-week) scan.

Median PFS of all patients in experimental cohort (serous and endometrioid histology cancers) was 4.6 months (95% CI: 3.7 to 4.9mths) with a 6-month PFS rate of 37% (95% CI: 26 to 48%). These estimates are based on observation of 67 events from 70 patients in experimental cohort.

Exploratory Cohort Outcomes

Thirty-two patients were accrued to the rare histology exploratory cohort. This cohort included patients with carcinosarcoma (n = 19), clear cell (n = 5), mixed (n= 6), mucinous (n = 1) and adenosquamous (n = 1). Across all patients in the exploratory cohort, 2 patients had PR as their best response (1 patient with carcinosarcoma and 1 with a mixed histology endometrial cancer), and 15 patients achieved a 12-wk PFS endpoint, 5 of whom remained progression free at subsequent (week 20) scan. Within the carcinosarcoma cohort, 1 patient achieved PR (5%) and 8 pts met 12-wk PFS (42%). Median PFS for both the whole exploratory cohort and the carcinosarcoma subgroup was 3.0 mths (95% CI: 2.5 to 4.6mths). The patient with carcinosarcoma who had a PR on treatment had a PFS of 6.7 months.

Figure 2 summarizes anti-tumor activity of cabozantinib in both experimental and exploratory cohorts. Waterfall plots (a, d) illustrate tumor response and swimmers’ plots (b, d) time on treatment, color-coded for different histology patients. Median PFS is presented for experimental (c) and exploratory (f) cohorts in accompanying graphs.

Fig 2. Summary of activity of cabozantinib.

Summary of activity of cabozantinib in both experimental (a, b, c) and exploratory (d, e, f) patient cohorts. (a, d) Objective tumor responses with dotted lines indicating boundaries for stable disease. (b, e) Time on treatment (b, e) with dotted line highlighting 12-week time-point. Patients who also achieved a partial response on study are indicated in blue. (c, f) Progression-free-survival in experimental and exploratory cohorts color-coded for tumor sub-groups.

Safety & Adverse Events

Cabozantinib related toxicity was mostly manageable. Adverse events were primarily Grade 1-2 in severity, and the most common toxicities of fatigue, hypertension, diarrhea and hand-foot syndrome responded well to conservative treatment and dose reduction. After experiencing an adverse event, 53 patients (52%) continued on study at daily dose of 40mg, of these 14 patients required a further dose reduction to 20mg.

Relevant toxicities experienced by all patients (n=102) who received at least one dose of cabozantinib are outlined in Table 2, with a focus on treatment-related toxicities experienced by ≥ 10% patients. Also included are adverse events of special interest based on the anti-angiogenic mechanism of action of cabozantinib including perforation/fistula events, thromboembolism and bleeding.

Table 2. Treatment related adverse effects.

Treatment-related adverse events (AEs) experienced by > 10% all patients treated (n = 102), including serious adverse events of interest.

	No. of patients affected (n = 102)
ADVERSE EVENT	< Grade2	Grade 3	Grade 4	Grade 5	Total
Constitutional
Fatigue	62	3	0	0	65
Anorexia	47	2	0	0	49
Weight Loss	27	4	0	0	31
Vascular
Hypertension	26	25	1	0	52
Gastrointestinal
Diarrhea	52	8	0	0	60
Nausea	44	2	0	0	46
Dysgeusia	42	0	0	0	42
Mucositis	38	0	0	0	38
Vomiting	19	0	0	0	19
Gastroesophageal Reflux	17	0	0	0	17
Dyspepsia	13	0	0	0	13
Dry mouth	15	0	0	0	15
Abdominal Pain	16	3	0	0	19
Dermatologic
*PPED	40	0	0	0	40
Dry skin	11	0	0	0	11
Pruritus	10	0	0	0	10
Alopecia	12	0	0	0	12
Hematologic
Anemia	22	5	0	0	27
Leukopenia	33	0	0	0	33
Lymphopenia	23	7	1	0	31
Neutropenia	22	1	0	0	23
Thrombocytopenia	18	1	1	0	20
Biochemical
AST increased	61	4	1	0	66
ALT increased	55	8	1	0	64
Hypomagnesemia	45	3	0	0	48
Hypophosphatemia	25	5	0	0	30
ALP increased	22	1	1	0	24
Hypoalbuminemia	28	1	0	0	29
GGT	13	4	0	0	17
Hyponatremia	18	1	2	0	21
Hypokalemia	15	5	0	0	20
Hypocalcemia	19	0	0	0	19
Hyperbilirubinemia	11	0	1	0	12
Increased lipase	9	3	0	0	12
Proteinuria	12	0	0	0	12
Other
Hypo/hyperthyroidism	12	0	0	0	12
Hoarseness	15	0	0	0	15
Extremity pain	11	0	0	0	11
Headache	11	0	0	0	11
Special Interest AEs
Bleeding: other	11	1	0	0	12
GI bleed	2	1	0	0	3
Thromboembolism (venous)	5	5	1	0	11
Thromboembolism (arterial)	1	0	0	0	1
GI perforation/fistula	2	3	3	0	8

^*PPED: palmar-plantar erythrodysesthesia

Amongst 102 patients, 21 patients discontinued treatment due to an adverse event. Upon review, 1 of the 11 was ineligible due to a protocol violation (of inclusion/exclusion criteria) with poorly controlled hypertension at baseline. Another patient was identified as having a clinically occult, but radiographically detected, fistula on baseline imaging. Both were considered to have discontinued treatment due to exacerbation of these baseline conditions in cycle 1. Adverse events experienced by the remaining 19 patients, included colonic fistula/perforation (n = 8), thromboembolism (n = 3), pain (n = 2), hypertension (n = 1); sepsis (n = 1), malnutrition (n = 1), dermatologic toxicity (n = 1), weight loss (n = 1) and hemothorax (n = 1); these adverse events were in the context of disease progression in 4 patients. One patient died of complications of a colonic fistula. A further 6 patients died on study but with evidence of disease progression; these deaths were deemed to be disease related.

Somatic Genomic Profiling

Archival tumor was available in 91 patients. Genomic analysis failed in 7. Of the remaining 84, 83 tumors were profiled on the NGS TruSeq amplicon panel and 1 on Sequenom (Fig 3 (a)). Amongst endometrioid cancers (n = 31), most commonly observed mutations were PTEN (n=13; 42%), PIK3CA (n=10; 32%), CTNNB1 (n=9; 29%), KRAS (n=8; 29%) and TP53 (n=6; 19%). In serous cohort (n = 30), most commonly observed mutations were TP53 (n=19; 58%), PIK3CA (n=10; 38%) and KRAS (n=3; 12%). Carcinosarcoma tumors (n = 15) demonstrated high rates of somatic TP53 (n = 7; 47%), PIK3CA (n = 6; 40%) and KRAS (n = 3; 20%) mutation. (Fig 3(b)). No MET mutations were documented across all tumors profiled.

Fig 3. Somatic Profiling Summary.

Summary of somatic profiling of endometrioid, serous and carcinosarcoma histology endometrial cancer patient treated with cabozantinib. (a) Progression-free-survival (in months) is aligned against molecular variants classified by OncoKB; oncogenic variants are highlighted in green, variants that are unclassified or of uncertain significance are marked in orange and germline variants in yellow. Samples not analyzed are indicated in grey. (b) Heat map summarizing frequencies of common mutations by histologic cohort.

There was an interesting trend of high frequency of responses in two patient cohorts with unique molecular contexts. The first group was patients with somatic CTNNB1 mutation. In this group of 10 patients we observed a 40% response rate (n = 4) and 70% 12-week PFS rate (n = 7). Two patients were on treatment for over 12 months, and median PFS (of the cohort of 10 patients) was 7.6 months. In the second group of 12 patients with concurrent somatic KRAS and PTEN or PIK3CA mutations, we observed a 25% response rate (n = 3) and 83% 12-week PFS rate (n = 10), median PFS of this cohort of 12 patients was 5.9 months. In spite of these intriguing trends, these higher response rates were not statistically significant in comparison with responses observed in the full patient cohort.

In addition to profiling genomic variants, the first 29 patient tumors (which included 12 endometrioid, 12 serous, 4 carcinosarcoma and 1 mixed histology tumours) were evaluated for MET amplification by FISH. In this subset, there was no evidence of MET amplification; although 2 patient tumors did have small clones of amplification, with cMET/CEP7 ratio of 1.55, this was not significant.

DISCUSSION

In this large, single-arm study, we observed activity of cabozantinib in both serous and endometrioid endometrial cancer patients who have experienced disease progression after platinum-based chemotherapy. In addition to the co-primary endpoints of response rate and 12-week PFS, we also evaluated the 6-month PFS rates, as per recent consensus guidelines. (27) In our cohort of serous/endometrioid EC, disease control reflected by 12-week PFS rates of 59 to 66%, and 6-month PFS rates of 29 to 40%, is encouraging compared against a historical 6-month PFS of less than 20% and aligns favorably against that of agents evaluated previously. (28)(29) Further, treatment-related toxicities were primarily ≤ Grade 2 with few ≥ Grade 3 events, most of which were manageable with conservative measures and dose reductions. The percentage of patients who required dose reduction aligns with reports of similar strategies in this disease. (30) The majority of patients discontinued therapy for disease progression. Of 102 patients treated, 22 discontinued treatment for adverse events, 5 of whom also had disease progression. There were therefore 17 patients (17%) who discontinued for treatment-related toxicity. That 6 patients continued on cabozantinib for longer than 12 months also speaks further to the tolerability of this regimen.

Women with non-serous/non-endometrioid EC are often excluded from clinical trials. Through our exploratory cohort, we were able to characterize the molecular features of these rare histology ECs. We observed similar somatic genomic aberrations in uterine carcinosarcomas as in serous and endometrioid EC, aligning with recent reports. (31) These observations provide further justification to evaluate targeted agents like cabozantinib across the histologic and molecular spectrum of EC. Not surprisingly, carcinosarcoma patients in our exploratory cohort had an especially poor prognosis. We observed a PR rate of 7%, 12-wk PFS of 42% and 3-month median PFS, somewhat encouraging in comparison with results of pazopanib in carcinosarcoma EC, where no PRs were observed and median PFS was 2 months. (32) We remain cognizant however that these outcomes remain dismal and that these patients experienced a high frequency of GI perforation/fistula events, which may be reflective of disease burden. Still, we feel that there are some preliminary signals of activity that warrant further evaluation, potentially in earlier stages of disease for this patient cohort with aggressive tumor behavior and few available options.

One serious adverse event of significance is that of GI perforation/fistula (GIPF) which affected 6% serous/endometrioid patients and 16% patients in exploratory cohort. These hazards are reminiscent of the early evaluation of bevacizumab in ovarian cancer (18) and more recent studies in endometrial cancer.(33) Although smaller trials of other anti-angiogenics have noted lower frequency of these events, the lower number of patients treated must be considered in these observed risks. (12, 30) The 4-patient group in experimental cohort who developed GIPF, includes 1 patient with unrecognized fistula on baseline imaging, highlighting the importance of careful radiologic review of patients with bowel involvement from recurrent endometrial cancer. Two had received prior (external beam pelvic) radiotherapy and there was no observed correlation between prior radiation and fistula development. None of the 5 patients in the exploratory cohort who developed GIPF had received prior radiotherapy. Half of GIPF events were in the context of confirmed or suspected progression. A randomized evaluation in a larger cohort would better delineate cabozantinib-related toxicity from disease-related events and we feel is justified based on the preliminary signals of activity observed here. Further, incorporating formal patient reported outcome measures in a randomized evaluation, would provide important insights into the potential impact of this treatment on patient quality of life.

Although cabozantinib has demonstrated efficacy across different tumor types, molecular markers predictive of response have been elusive; several studies have been unable to establish a robust relationship between MET expression status and tumor response. (34, 35) We initially explored a potential relationship between MET amplification and response. In a preliminary assessment of a subset of 29 tumors across different histologies, we identified no tumors with MET amplification, aligning with data in publicly available databases where frequency of MET amplification was < 1%. (36, 37) Recognizing that MET was likely a poor candidate biomarker of response in this setting we discontinued this line of investigation to prioritize scarce tumor tissue for other correlative molecular studies.

We completed profiling of archival tumor samples using a 48-gene panel encompassing genes relevant to EC pathogenesis and MET signaling (see Suppl Table 1 for details). No MET mutations were documented, aligning with previous reports. (38) Together with the lack of MET amplification observed, it would seem that that aberrant MET function is not a major determinant in predicting cabozantinib sensitivity in endometrial cancer, unless driven through other pathways.

We did observe a trend of increased clinical benefit in two discrete molecular cohorts of patients. The first group included patients with somatic CTNNB1 mutation. CTNNB1 mutations have long been implicated as an early event in endometrial carcinogenesis, appearing limited to endometrioid histology cancers. (39) An interesting analysis integrating genomic and proteomic profiling identified CTNNB1 mutations in young, obese women who, in spite of an initial presentation with early stage disease, had inferior overall survival in comparison with women diagnosed with similar stage/grade of disease but lacking CTNNB1 mutation. (40) Further, in vitro and in silico analyses have linked CTNNB1 mutation with increased angiogenesis (41), providing mechanistic rationale for observations from GOG-86P, where CTNNB1 mutation identified a subgroup of patients who benefited from the addition of the anti-angiogenic bevacizumab to carboplatin/paclitaxel. (42) Given the recognized poor prognosis of patients harboring somatic CTNNB1 mutation, our exploratory observations of a partial response rate of 40% and median PFS of 7.6 months in this small cohort of 10 patients is encouraging and warrants further evaluation, ideally in a randomized trial setting.

The second cohort of interest included patients with tumors harboring concurrent KRAS and PTEN or PIK3CA mutations. This finding aligns somewhat with one made in medullary thyroid cancer (MTC), where patients with tumors with RET or RAS mutation had a survival benefit with cabozantinib. (43) This is particularly interesting in endometrial cancer given the prominence of RAS/MAPK pathway activation in this disease. Further, although RET mutation is a rare occurrence (44) (we detected RET mutation in 1 patient tumor), other mechanisms of RET pathway dysregulation, occult to traditional panel profiling, have been characterized in endometrial cancer and may be relevant in the context of dependence on pro-angiogenic signaling. (45) Again, we highlight the fact that these analyses were not pre-specified and are best exploratory and hypothesis generating; still, given the biological plausibility and intriguing trends, we believe these potential relationships warrant further evaluation in a larger randomized trial.

There remain a few limitations of this work. A centralized review by an expert gynecologic oncology pathologist was not included and may have identified discrepancies in histotype and grading of some patient tumors. Further, contemporary strategies of somatic profiling based on exome or whole genome sequencing, are more comprehensive than the targeted hot spot panels that were in common use at the time this study was conducted. We recently described discrepancies in identification of TP53 mutation in high grade serous ovarian cancer using targeted sequencing versus standard immunohistochemistry. (46) It is likely that both histological misclassification and the use of less sensitive sequencing approaches both contributed to the lower than anticipated rate of TP53 mutation (63%) in our serous histology cohort. Finally, correlative molecular studies were completed in archival tumor samples which may not accurately reflect the mutational status of tumors at time of recurrence and after treatment with chemotherapy or radiation. (47) We feel that the correlative work completed in this study however, provides some interesting avenues for further consideration, and suggest that future studies in this area employ more in-depth tumor characterization, including status of mismatch repair genes, fusion genes and quality of immune infiltrate, in baseline tumor biopsies for a higher yield with respect to identifying candidate biomarkers.

In conclusion, cabozantinib has single-agent activity with acceptable toxicity across the molecular spectrum of EC. Although a clear predictive biomarker of response remains to be determined, the data suggests that patients whose tumors harbor aberration in either CTNNB1 or both MEK and PIK3CA/AKT pathways might be more sensitive to this treatment approach. Our study adds to the body of literature to support a benefit to targeting angiogenesis in EC. (30)

Fig 4. Spiderplots of tumor responses in molecular subgroups of interest.

Tumour volume changes over time in patients with (a) CTNNB1 variant (green) vs all others (blue) and (b) concurrent KRAS and PTEN/PIK3CA mutations (red) vs all others (blue).

TRANSLATIONAL RELEVANCE.

Although cytotoxic chemotherapy remains the standard treatment for recurrent/progressive endometrial cancer, clinical benefit is short-lived and toxicity significant. Despite the high frequency of genomic aberrations characterizing endometrial cancer, the development of molecularly targeted agents has been challenged by an incomplete understanding of molecular predictors of response. This single arm study demonstrates an encouraging signal of activity for cabozantinib across different histologic and molecular subtypes of endometrial cancer. The observation of increased frequency of responses in patients with tumors harboring CTNNB1 mutation or concurrent KRAS and PTEN/PIK3CA mutations, is hypothesis-generating for future studies. The results reported here not only provide support for the further evaluation of cabozantinib in endometrial cancer, but also justify the critical need for endometrial cancer drug studies to be inclusive, enrolling broad, molecularly-characterized, patient populations to facilitate insights into the heterogeneity of clinical benefit, and factors predictive of resistance and response.