Key Points
Question Is the activity of immune checkpoint inhibitors augmented by combination therapy with bevacizumab and metronomic cyclophosphamide in patients with recurrent ovarian cancer?
Findings In this single-arm, phase 2 nonrandomized clinical trial, 40 patients with platinum-sensitive or platinum-resistant ovarian cancer received pembrolizumab with bevacizumab every 3 weeks and oral cyclophosphamide daily. The objective response rate was 47.5% with a median progression-free survival of 10.0 months, and this regimen was well tolerated with good quality of life during treatment.
Meaning These findings suggest that the combination of pembrolizumab with bevacizumab and metronomic cyclophosphamide has clinical activity with a favorable toxicity profile in treatment of recurrent ovarian cancer.
Abstract
Importance Treatment options for recurrent ovarian cancer are of limited clinical benefit and adversely affect patient quality of life, representing an unmet need for tolerable effective therapies.
Objective To assess the efficacy and safety of a combination of pembrolizumab with bevacizumab and oral metronomic cyclophosphamide in patients with recurrent platinum-sensitive, platinum-resistant, or refractory epithelial ovarian, fallopian tube, or primary peritoneal cancer.
Design, Setting, and Participants This open-label, single-arm phase 2 cohort study enrolled patients from September 6, 2016, to June 27, 2018, at a single institution in the United States. Eligible patients had recurrent ovarian cancer, measurable disease per immune-related Response Evaluation Criteria In Solid Tumors (irRECIST), and Eastern Cooperative Oncology Group performance status of 0 to 1. Data were analyzed from September 6, 2016, to February 20, 2020.
Interventions Patients received intravenous pembrolizumab, 200 mg, and bevacizumab, 15 mg/kg, every 3 weeks and oral cyclophosphamide, 50 mg, once daily during the treatment cycle until disease progression, unacceptable toxic effects, or withdrawal of consent.
Main Outcomes and Measures Primary outcomes were objective response rate (ORR) and progression-free survival (PFS).
Results Of the 40 women enrolled, 30 (75.0%) had platinum-resistant and 10 (25.0%) had platinum-sensitive ovarian cancer with a mean (SD) age of 62.2 (9.4) years. Three women (7.5%) had complete responses, 16 (40.0%) had partial responses, and 19 (47.5%) had stable disease in response to treatment based on irRECIST criteria, with an ORR of 47.5%, clinical benefit in 38 (95.0%), and durable response in 10 (25.0%). Median PFS was 10.0 (90% CI, 6.5-17.4) months. The most common grade 3 to 4 treatment-related adverse events were hypertension (6 [15.0%]) and lymphopenia (3 [7.5%]). The most frequently reported adverse events included fatigue (18 [45.0%]), diarrhea (13 [32.5%]), and hypertension (11 [27.5%]).
Conclusions and Relevance In this phase 2 nonrandomized clinical trial, the combination of pembrolizumab with bevacizumab and oral cyclophosphamide was well tolerated and demonstrated clinical benefit in 95.0% and durable treatment responses (>12 months) in 25.0% of patients with recurrent ovarian cancer. This combination may represent a future treatment strategy for recurrent ovarian cancer.
Trial Registration ClinicalTrials.gov Identifier: NCT02853318
Introduction
Epithelial ovarian, fallopian tube, and primary peritoneal cancer are the most lethal gynecological malignant neoplasms.1 Although the survival benefits of optimal cytoreduction and primary platinum taxane–based chemotherapy have been well established in conjunction with mounting evidence supporting the role of poly-adenosine diphosphate ribose polymerase inhibitors for BRCA-mutated or homologous recombination deficient ovarian cancer, there remains a paucity of effective treatments providing durable response in recurrent disease. Traditional second-line chemotherapies achieve limited clinical benefit, with objective response rates (ORRs) ranging from 10% to 25% and with significant treatment-related toxic effects, highlighting an enormous unmet need for the development of novel therapies.
Ovarian cancer is an immunogenic disease whereby increased infiltration of T cells in tumor islets correlates with significantly longer survival.2,3 Approximately half of patients with ovarian cancer demonstrate T-cell infiltration in the tumor microenvironment (TME); however, these cytotoxic T cells are subject to various suppressive mechanisms, including inhibition by FoxP3+ regulatory T cells4 as well as induced expression of programmed cell death 1 (PD-1) and other inhibitory checkpoint receptors that dampen antitumor immune responses.5
Programmed cell death 1 and its ligand (PD-L1) are expressed in 28% to 40% of patients with ovarian cancer.6,7 The blockade of PD-1 or PD-L1 augments T-cell function in the TME8; however, only a minority of patients with ovarian cancer respond to immune checkpoint inhibition (ICI). A recent large clinical trial9 has shown underwhelming response rates for single-agent ICI, demonstrating an ORR of 8% in all patients with ovarian cancer and a median progression-free survival (PFS) of 2.1 months. This minimal efficacy could be partially attributed to lack of tumoral T-cell infiltration in certain patients and to the high local immunosuppression constituted by tumor cells, regulatory T cells, and tolerance-inducing myeloid cells.
Vascular endothelial growth factor (VEGF) is highly expressed in most patients with ovarian cancer, serving as a major driver of tumor neovascularization and local immune suppression.10 High levels of VEGF inhibit T-cell trafficking to TME and increase Fas ligand expression on aberrant tumor endothelial cells, leading to T-cell apoptosis.11 High VEGF also promotes expansion of inhibitory immune cell subsets, including regulatory T cells and myeloid-derived suppressor cells, resulting in an inherently immunosuppressive TME.10 In addition to targeting the VEGF pathway, metronomic chemotherapy using oral cyclophosphamide daily is a nontoxic, inexpensive strategy to further enhance antitumor immune response. Metronomic cyclophosphamide administration reduces tumorigenicity of cancer stem cells and selectively depletes regulatory T cells, thus tipping the balance toward increased cytotoxic activity.12
The use of bevacizumab, a humanized monoclonal antibody targeting VEGF, combined with oral metronomic cyclophosphamide is an effective treatment in the management of recurrent ovarian cancer with modest clinical response (ORR, 24%; PFS, 7.2 months).13 Therefore, to enhance T-cell infiltration and concomitantly decrease local immunosuppression mediated by VEGF and regulatory T cells, we conducted a phase 2 clinical trial to assess the safety and synergistic effect of the combination of pembrolizumab, bevacizumab, and oral cyclophosphamide in patients with recurrent ovarian cancer.
Methods, Study Design, and Patients
This investigator-initiated, single-arm, phase 2 nonrandomized clinical trial of pembrolizumab and bevacizumab with oral cyclophosphamide enrolled patients from September 6, 2016, to June 27, 2018 (Figure 1). Interim analysis was performed March 1, 2020, for activity and safety. The trial protocol is available in Supplement 1. This study was approved by the Roswell Park institutional review board, and all patients provided written informed consent before initiating any study procedures. Participants did not receive financial compensation. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.
Key eligibility criteria included the following: age of 18 years or older; histologically confirmed serous, endometrioid, clear cell, mucinous, or undifferentiated recurrent epithelial ovarian, fallopian tube, or primary peritoneal cancer; measurable disease with immune-related Response Evaluation Criteria In Solid Tumors (irRECIST) or RECIST, version 1.114,15; and an Eastern Cooperative Oncology Group performance status score of 0 or 1. Participants were defined as having platinum-resistant disease if they experienced recurrence less than 6 months from the most recent platinum therapy, or platinum-sensitive disease with recurrence 6 months or longer after exposure. Individuals with platinum-sensitive disease were enrolled if the patient had contraindication, intolerance, or hypersensitivity to platinum agents. Prior antiangiogenic therapy, cyclophosphamide, or anti–PD-1/anti–PD-L1 treatment exposure was permitted. Exclusion criteria included active autoimmune disease, central nervous system metastasis, treatment-refractory hypertension, renal insufficiency, myelosuppression, or cardiac or liver dysfunction. Eligibility criteria are detailed in eMethods in Supplement 2. Trial design included a lead-in phase of 5 patients to assess drug-related toxic effects. No dose-limiting toxic effects were seen in the safety lead-in cohort; thus, an additional 35 patients were accrued.
Procedures and Assessments
All patients received an intravenous infusion of pembrolizumab, 200 mg, and bevacizumab, 15 mg/kg, every 3 weeks and 50 mg of oral cyclophosphamide once daily during a 21-day treatment cycle.16 Baseline PD-L1 status was measured via immunohistochemistry, with positivity defined by the presence of interface pattern staining or a modified proportion score reflecting overall frequency of PD-L1–expressing mononuclear and tumor cells within tumor nests. A modified proportion score threshold of at least 1% using the 22C3 antibody (QualTek Molecular Laboratories) was adopted similarly to previously described protocols.6
Physical examination was performed every 3 weeks to assess toxic effects and adverse events, which were graded according to the National Institutes of Health common toxicity criteria.17 Management of grades 3 to 4 adverse effects attributed to pembrolizumab, bevacizumab, or cyclophosphamide included interruption or discontinuation of the study drug. In the event of grade 2 toxic effects resulting from pembrolizumab, patients were monitored for resolution within 12 weeks of the last infusion. Pembrolizumab therapy was resumed after adverse events had been reduced to grade 0 or 1; otherwise treatment was discontinued. No dose reductions were permitted on this protocol. Dose interruption was permitted in the event of toxic effects attributed to a single trial compound.
Treatment response was assessed after every 3 cycles via computed tomographic scans using irRECIST criteria by independent, blinded radiographic review. Treatment was continued until radiographic or clinical evidence of disease progression, development of unacceptable toxic effects, or withdrawal of consent. Quality-of-life assessment was conducted at baseline and after 3 and 6 months of treatment using European Organisation for Research and Treatment of Cancer QLQ-C30 and QLQ-OV28 questionnaires.18
Outcomes
The primary objective of the study was to investigate the ORR and PFS using irRECIST criteria. Secondary end points included frequency and severity of adverse events, duration of response (DOR), overall survival, quality of life, and the treatment response based on PD-L1 status. The DOR was defined as the time from documented response until disease progression or last disease assessment (censored). The PFS was defined as the time from treatment initiation until disease progression, death, or last disease assessment (censored). Overall survival was defined as the time from treatment until death or last follow-up.
Statistical Analysis
Data were analyzed from September 6, 2016, to February 20, 2020. With an accrual goal of 40 evaluable patients, the study had 86.1% power to detect an improvement in the 7-month PFS rate from 0.3 to 0.5 (at α = .10). The null hypothesis is based on the presumed activity of standard chemotherapy, single-agent bevacizumab, or combination low-dose oral metronomic cyclophosphamide and bevacizumab in recurrent ovarian cancer with a median duration of response of 3.9 to 4.5 months.13,19 An improvement to a median PFS of 7.0 months (ie, 7-month PFS rate of 0.5) would be considered significant in this study population.
The analysis includes all patients who received study treatment. The treatment-related adverse events were summarized overall and by grade using frequencies and relative frequencies. The DOR, PFS, and overall survival were summarized using Kaplan-Meier methods, wherein estimates of the median were obtained with 90% CIs. The ORRs were presented with 90% CIs obtained by the Jeffery method.20 Two-sided P values were reported for all analyses unless otherwise specified. No adjustments were made for multiple testing. All analyses were conducted in SAS, version 9.4 (SAS Institute, Inc) at a significance level of 2-sided α = .10.
Results
The mean (SD) age of the 40 participants was 62.2 (9.4) years (median, 62 [interquartile range {IQR}, 56-70] years; range, 45-89 years) with demographic data reflected in eTable 1 in Supplement 2. The most common histological subtype was high-grade serous carcinoma. Thirty participants (75.0%) had platinum-resistant disease. The mean (SD) number of prior lines of chemotherapy was 3.4 (2.4) for all patients and 3.8 (2.6) for patients with platinum-resistant disease. Prior exposure to bevacizumab was noted in 14 patients (35.0%), compared with 5 patients (12.5%) with a history of oral cyclophosphamide use. No significant differences were identified between populations with platinum-resistant and platinum-sensitive disease with respect to age, race, performance status, or PD-L1 or BRCA status (eTable 1 in Supplement 2). No participants in this study had prior exposure to anti–PD-1 or another ICI.
Germline or somatic BRCA variants were identified in 14 patients (35.0%), 23 patients (57.5%) had BRCA-negative status (germline and somatic), and 3 patients (7.5%) had an unknown BRCA status. Evaluable baseline tumor specimens were obtained in 36 participants, and PD-L1 status was positive in 19 patients (47.5%) via immunohistochemical analysis based on a modified proportion score threshold of 1% or presence of interface PD-L1 staining. Using a PD-L1 expression threshold of 1%, those with PD-L1–positive tumors had an ORR of 52.6% (10 of 19 patients) compared with 35.3% for PD-L1–negative disease (6 of 17 patients; P = .34), with no difference in PFS between the 2 groups (P = .13).
Three patients (7.5%) achieved a complete response and 16 (40.0%) achieved a partial response based on irRECIST criteria, demonstrating an ORR of 47.5% (90% CI, 34.9%-60.3%) (Table). Median time to best response from the start of treatment was 5.8 months (IQR, 3.1-10.7 months [range, 0-26.4 months]). All patients with a complete response to therapy were noted to have platinum-resistant disease, and 1 of these 3 patients at present has no evidence of disease on surveillance (Figure 2). There were 6 partial responses (6 confirmed) in patients with platinum-sensitive disease (ORR, 60.0%; 90% CI, 26.2%-73.8%) compared with 13 partial and complete responses (9 confirmed) in patients with platinum-resistant disease (ORR, 43.3%; 90% CI, 29.6%-58.2%) (eFigure 1 in Supplement 2). In addition, stable disease was noted in 11 patients (27.5%) at 24 weeks (Table). Overall, 38 of 40 patients (95.0%) had a clinical benefit in the clinical trial (complete plus partial responses plus stable disease), and 10 of 40 (25.0%) had a durable response.
Median PFS in the study population was 10.0 (90% CI, 6.5-17.4) months (eFigure 2 in Supplement 2), and the 7-month PFS rate was 0.67 (90% CI, 0.53-0.78), exceeding the predefined threshold of 0.3 for a heavily pretreated ovarian cancer cohort. Patients were monitored with a median follow-up of 25.5 (90% CI, 2.5-34.0) months. When stratified by best response using irRECIST criteria, patients with a complete response had not reached median PFS at time of analysis (90% CI, 12.0 months to not recorded), whereas those with a partial response had a median PFS of 10.6 (90% CI, 6.5-20.2) months vs patients with stable disease with a median PFS of 7.1 (90% CI, 4.7-10.8) months. Median DOR was 5.9 (IQR, 3.6-11.3) months for all patients who achieved clinical benefit (n = 38), and 8.3 (IQR, 4.2-15.6) months for patients who showed partial or complete response (n = 19) to treatment. There was no statistically significant difference in PFS in patients with stable disease with decrease in tumor burden from the baseline (0%-30% decrease) compared with patients with partial or complete response (median PFS, 10.3 [90% CI, 4.3-27.1] vs 17.4 [90% CI, 7.8-24.4] months, respectively; P = .47) (Figure 3A). Patients with platinum-sensitive disease had a median PFS of 20.2 (90% CI, 6.0 to not recorded) months compared with 7.6 (90% CI, 5.7-10.3) months in patients with platinum-resistant disease (P = .16) (eFigure 3 in Supplement 2). Patients who received 3 or fewer lines of prior chemotherapy treatment before trial start had significantly better PFS and overall survival compared with patients with more than 3 prior lines of treatment: PFS: 10.8 (90% CI, 7.6-24.4) vs 6.5 (90% CI, 4.3-10.2) months (P = .03) (Figure 3B); overall survival, 26.1 (90% CI, 12.7 to not reported) vs 12.5 (90% CI, 8.3-17.1) months (P = .03) (eFigure 4 in Supplement 2).
Any grade of treatment-related adverse events occurred in 33 patients, with grade 3 or greater adverse events in 13 (32.5%). Most frequently reported treatment-related adverse events (any grade) included fatigue (18 [45.0%]), diarrhea (13 [32.5%]), and hypertension (11 [27.5%]) (eTable 2 in Supplement 2). The most common grade 3 or greater adverse reactions were lymphopenia attributed to oral cyclophosphamide (3 [7.5%]) or bevacizumab-induced hypertension (6 [15.0%]). One grade 4 reaction was noted due to low white blood cell and platelet counts that resolved without clinical intervention. No deaths attributed to the investigational protocol occurred during this study. Treatment-related adverse events led to dose interruptions in 17 patients (42.5%), most frequently due to hypertension attributed to bevacizumab at time of assessment. Four patients (10%) discontinued therapy due to treatment-related adverse events, and the most common cause of treatment discontinuation was grade 2 skin toxic effects attributed to pembrolizumab.
Quality-of-life questionnaires were completed at baseline and after 3 and 6 months of treatment using the QLQ-C30 and QLQ-OV28. No association between treatment response and quality-of-life domains, including global health status, was identified. Global and physical functioning scores remained high and stable throughout the clinical trial with improved mean (SD) scores in body image (baseline, 64.3 [25.0]; 6 months, 72.0 [27.9]), emotional functioning (baseline, 70.0 [22.8]; 6 months, 77.2 [17.7]), and social functioning (baseline, 68.0 [30.4]; 6 months, 75.6 [27.6]) measured during the first 6 months of the treatment (eFigure 5 in Supplement 2).
Discussion
In the absence of curative therapy for treatment-refractory disease, the incorporation of target-specific treatments and immunotherapy with meaningful response in ovarian cancer are crucial for women diagnosed with recurrence. Combining anti–PD-1 and antiangiogenic therapy with metronomic cyclophosphamide represents a novel opportunity to augment cytotoxic lymphocyte function while potentially modulating harmful regulatory T-cell responses and optimizing vascularization within the TME.21,22 This single-arm phase 2 trial demonstrates an encouraging ORR of 47.5%, a median PFS of 10.0 months, and clinical benefit of 95.0% in a heavily pretreated cohort of patients with recurrent ovarian cancer. As with similar trials using combination ICI in recurrent ovarian cancer, this study included individuals with platinum-sensitive and platinum-resistant disease to evaluate treatment responses among a spectrum of women.7
Several studies investigating the use of ICI combined with traditional chemotherapy or targeted treatments are ongoing.7,13,23,24 An investigation of pembrolizumab and pegylated liposomal doxorubicin hydrochloride demonstrated an ORR of 11.5% among 26 patients with platinum-resistant ovarian cancer, with a higher than anticipated rate of pneumonitis.23 The combination of pembrolizumab with cisplatin and gemcitabine followed by maintenance pembrolizumab therapy in platinum-resistant ovarian cancer offers a promising ORR of 57% but a median DOR of 3.5 months.24 A similar phase 2 clinical trial in patients with recurrent ovarian cancer using nivolumab with bevacizumab7 showed an ORR of 28.9% and a median PFS of 8.1 months. Bevacizumab and oral cyclophosphamide alone had an ORR of 24% and a median PFS of 7.2 months in a less heavily pretreated cohort of patients with recurrent ovarian cancer and mostly platinum-sensitive disease.13
Exploration of dual ICI with ipilimumab and nivolumab in recurrent ovarian cancer25 suggests a meaningful response in 51 patients (31.4%) at the cost of grade 3 or greater toxic effects in 67% of participants. Another randomized phase 2 study comparing nivolumab with nivolumab and ipilimumab26 showed increased ORR for dual ICI (12% vs 31%) with slightly increased PFS (2 vs 3.9 months) and higher grade 3 toxic effects noted in the combination arm (33% vs 49%). Poly-adenosine diphosphate ribose polymerase inhibition combined with ICI offers another promising strategy, but preliminary findings27,28 suggest response rates of 18% to 20% mirror conventional therapies for ovarian cancer in the absence of platinum-sensitive, BRCA-mutated disease.
In the present study, women with platinum-resistant disease had an ORR of 43.3% (90% CI, 29.6%-58.2%) with a median PFS of 7.6 (90% CI, 5.7-10.3) months compared with an ORR of 60.0% (90% CI, 26.2%-73.8%) and median PFS of 20.2 (90% CI, 6.0 to not recorded) months among those with platinum-sensitive disease. Although patients with platinum-sensitive disease had a trend toward greater median PFS than those with platinum-resistant disease, these differences did not reach statistical significance (P = .16) (eFigure 3 in Supplement 2). When patients were analyzed based on the number of prior chemotherapy lines, patients with no more than 3 prior chemotherapy lines had a significantly longer PFS (10.8 [90% CI, 7.6-24.4] vs 6.5 [90% CI, 4.3-10.2] months; P = .03), further suggesting that using immunotherapy earlier in the recurrent setting leads to greater PFS benefit.
Subanalyses of long-term responders (10 [25.0%]) with prolonged benefit (>12 months) on this regimen were also performed. Expression of PD-L1 in baseline tumor biopsy did not correlate with response or PFS. This is similar to findings demonstrated in a single-agent trial of avelumab in ovarian cancer.29 In the present study, there was no statistically significant correlation with BRCA variant status and PFS, although BRCA-positive patients had an ORR of 71.4% compared with 30.4% of BRCA-negative patients (P = .02). Further characterization of tumor mutational burden, neoantigen load, and T-cell phenotypes on translational specimens may help to identify genomic and immunologic correlates in patients associated with durable treatment response.
Interestingly, we saw no difference in PFS in patients with stable disease who had decreased tumor burden from baseline compared with patients with partial or complete response based on irRECIST criteria. Clinical observation supported our findings that many patients with stable disease had durable response and that partial or complete response alone did not always translate to durable clinical benefit. Of note, 2 patients with complete response did not have prolonged survival: one died of an unrelated cause while not receiving treatment, whereas the other demonstrated a new lesion shortly after a complete response. Secondary cytoreduction was performed under physician discretion for 2 women owing to a symptomatic pelvic mass who were censored at the time of first progression for survival analysis. Pathological findings for both patients were significant for liquified tumor with extensive necrosis (>60% of tumor tissue), and thus they were permitted to continue with the trial regimen postoperatively. Once treatment was restarted, their second PFS was much shorter.
Quality-of-life parameters of high global health status and cognitive, emotional, and social functioning remained consistent throughout the study period. These tools have been previously validated in patients with ovarian cancer and serve as a reliable means to assess disease and the effect of treatment and may further act as a surrogate marker for disease survival.30 The treatment of recurrent ovarian cancer is complex, and the goals of treatment are not only to optimize survival but also to palliate symptoms with minimal treatment-related toxic effects.
Limitations
The study has several limitations typical of early-phase trials. This study was a single-center, 1-arm, phase 2 trial lacking a comparison arm to characterize the effect of each trial drug vs a dual regimen or traditional second-line therapy. In addition, although the inclusion of heterogeneous tumor histological findings could be perceived as a potential study limitation, evaluation of these patients may better represent patient diversity managed in typical gynecological oncology practice.
Conclusions
There is an unmet clinical need to identify new strategies to improve ICI efficacy and durability, not only to achieve improved overall survival but also to buy meaningful time with good quality of life for patients with ovarian cancer. This study’s findings suggest that the combination of pembrolizumab with bevacizumab and oral metronomic cyclophosphamide was safe and well tolerated in patients with recurrent ovarian cancer and may present a promising treatment option. These results are currently being followed up with further analysis of biospecimens to understand the synergistic mechanisms among ICI, antiangiogenic therapy, and regulation T-cell depletion and thereby allow many more patients to benefit from immunotherapy treatments in the future.
