Safety and Clinical Activity of Atezolizumab Plus Bevacizumab in Patients with Ovarian Cancer: A Phase Ib Study

Ovarian cancer accounts for 3.4% of new cancer cases, but 4.4% of cancer-related deaths globally (1). Low survival rates are driven by advanced-stage diagnosis in most patients (2).

Platinum-based chemotherapy, usually in combination with taxanes, remains a cornerstone of systemic therapy for advanced and recurrent ovarian cancer (3). However, recurrent disease remains difficult to treat due to chemotherapy resistance (4). Antiangiogenic agents (bevacizumab) have demonstrated clear antitumor activity by delaying recurrence and progression in patients with ovarian cancer, but translating that benefit to prolonged overall survival (OS) remains challenging (4). Maintenance therapy with PARP inhibitors after an initial response to platinum-based chemotherapy is an encouraging option; however, its benefits are greatest in patients with BRCA1/2 and similar mutations that result in deficient homologous recombination (3, 5).

Anticancer therapy with immune checkpoint inhibitors, such as programmed death-ligand 1 (PD-L1)/programmed death 1 (PD-1) antagonists, have considerably improved the standard of care for numerous solid cancers, with some patients achieving substantial tumor regression and/or impressive durable responses (6). Atezolizumab is a humanized immunoglobulin mAb that selectively targets PD-L1 on tumor cells (TCs) and tumor-infiltrating immune cells (ICs) in the tumor microenvironment and reinvigorates suppressed T cells to kill cancer cells (7). Atezolizumab has been approved in the United States for the treatment of certain bladder, lung, and breast cancers. Ovarian cancer is an attractive target for immunotherapy because PD-L1 expression has been detected on the surface of TCs and ICs in ovarian cancer specimens and PD-L1 is a prognostic factor (8, 9). Atezolizumab and other mechanistically similar agents have shown preliminary clinical activity in patients with ovarian cancer (10–12), further supporting the therapeutic potential of targeting PD-L1 and its receptor, PD-1, in patients with this disease.

VEGF interferes with antigen presentation and successful T-cell migration into the tumor microenvironment, ultimately creating an immunosuppressive microenvironment within cancers (13). In addition to its antiangiogenic effects, bevacizumab can reverse these VEGF-mediated immunosuppressive effects on the tumor microenvironment, potentially augmenting immune-mediated antitumor activity. Dual blockade of VEGF and PD-L1 with antiangiogenic agents and PD-1/PD-L1 inhibitors has demonstrated encouraging clinical benefit and tolerable safety profiles in other tumor types (14–16). In patients with recurrent platinum-resistant ovarian cancer, the combination of bevacizumab and the PD-1 inhibitor, nivolumab, led to confirmed objective response rates of 16.7% (17).

Here, we report the safety and efficacy of atezolizumab plus bevacizumab in patients with platinum-resistant ovarian cancer.

Study GP28328 (NCT01633970) is an open-label, multicenter phase Ib basket study of atezolizumab plus targeted agents and/or chemotherapy in patients with solid tumors. The safety and clinical activity of atezolizumab plus bevacizumab in patients from the ovarian cancer cohort are reported here.

Eligible patients were aged ≥18 years; had measurable ovarian cancer per RECIST 1.1 that had progressed <6 months after completion of platinum-based therapy; an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1; adequate hematologic and organ function; and no platinum-refractory disease, history of bowel obstruction, or prior anti-PD-L1/PD-1 or anti-VEGF/VEGFR therapy. No ovarian cancer histology was excluded.

Patients were excluded if they had active or untreated central nervous system metastases, inadequately controlled hypertension, significant vascular disease ≤6 months prior to enrollment, or uncontrolled bleeding varices.

The study was conducted in accordance with the Declaration of Helsinki, the FDA regulations, and International Conference on Harmonisation E6 guidelines, and was approved by the relevant institutional review boards of participating research centers.

All patients provided written informed consent and agreed to provide archival and/or freshly collected tumor tissue.

Patients received intravenous atezolizumab (1,200 mg) and bevacizumab (15 mg/kg) every 3 weeks until disease progression or unacceptable toxicity.

The primary endpoint was the safety and tolerability of atezolizumab plus bevacizumab in patients with platinum-resistant ovarian cancer. Secondary efficacy endpoints included overall response rate (ORR), duration of response (DOR), progression-free survival (PFS), and OS.

Safety was assessed on the basis of adverse events (AEs) and serious AEs (SAEs) according to NCI Common Terminology Criteria for Adverse Events version 4.0.

Tumor assessment by CT scan was performed at screening and every two treatment cycles. Antitumor activity was assessed according to ORR, which was defined as the proportion of patients with a complete response (CR) or partial response (PR) per the investigator's assessment using RECIST 1.1 criteria, and confirmed by repeat assessments ≥4 weeks after initial documentation. The DOR was defined as the time from first occurrence of a documented objective response until the time of progression or death from any cause. A confirmatory scan was required to document CR and PR, but not progression. The disease control rate (DCR) was defined as the sum of the proportions of patients who had CR, PR, and stable disease (SD) for 12 weeks. PFS was defined as the time from study treatment initiation to the first occurrence of documented disease progression or death from any cause during the study, whichever occurred first. OS was defined as the time from the first dose of study treatment to the time of death from any cause during the study.

Biomarker evaluation was performed in archival formalin-fixed, paraffin-embedded (FFPE) primary or metastatic tumor samples collected before study entry. PD-L1 IHC and tumor-infiltrating lymphocyte (TIL) evaluations were performed centrally at Ventana Medical Systems, Inc.; CD8 IHC was performed at HistoGeneX. PD-L1 expression was centrally evaluated using the VENTANA PD-L1 (SP142) Assay (Ventana Medical Systems, Inc.; ref. 18). PD-L1 positivity was defined as PD-L1 expression on ≥1% of ICs. Tumor-infiltrating ICs (lymphocytes, macrophages, dendritic cells, and granulocytes) were identified by hematoxylin and eosin staining and scored as a percentage of the tumor area (TC and desmoplastic stroma). Intratumoral CD8⁺ T cells assessed by IHC (clone C8/144B, Dako North America, Inc.) were digitally quantified. Analysis of tumor mutational burden (TMB) and BRCA1/2 mutation status was performed using the FoundationOne CDx Next-Generation Sequencing Panel (Foundation Medicine Inc.). Patients with tumors that had known or likely deleterious mutations in the BRCA1 and/or BRCA2 gene were classified as having mutant tumors, whereas those displaying variants of unknown significance were grouped with the wild-type. Ovarian cancer molecular subtypes were derived from targeted RNA-Sequencing (Q2 Solutions) based on the Classification of Ovarian Cancer (CLOVAR) classification (19).

All analyses were based on the safety evaluable population, defined as patients who received any study treatment.

The Kaplan–Meier method was used to estimate medians for PFS, OS, and DOR; 95% confidence intervals (CIs) for the median were computed using the method of Brookmeyer and Crowley. ORR was calculated with 95% CI using the Clopper–Pearson method. Statistical analyses were conducted using SAS.

Qualified researchers may request access to individual patient-level data through the clinical study data request platform (https://vivli.org/). Further details on Roche's criteria for eligible studies are available here: https://vivli.org/members/ourmembers/. For further details on Roche's Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see here: https://www.roche.com/research_and_development/who_we_are_how_we_work/clinical_trials/our_commitment_to_data_sharing.htm.

Twenty eligible patients with platinum-resistant ovarian cancer were enrolled in this cohort of the GP28328 study and were evaluable. At the data cutoff on April 23, 2018, two patients were still receiving treatment.

Patient demographics are summarized in Table 1. The median number of prior treatment regimens was three (range, 1–6). Thirteen patients (65%) had received three or more prior systemic treatment regimens. Seven patients (35%) had PD-L1–positive tumors, and the most common histology type [in 12 patients (60%)] was serous.

At the data cutoff, the median duration of safety follow-up was 3.8 months (range, 0–20.8). The median duration of treatment exposure was 3.8 months for atezolizumab and 3.5 months for bevacizumab (range for both, 0–21 months). Five patients (25%) received atezolizumab plus bevacizumab for >12 months.

Treatment-related AEs (TRAE) of any grade occurred in 19 (95%) patients; grade 3/4 AEs were observed in seven (35%; Table 2). No grade 5 events were reported. The most common grade 3/4 treatment-emergent AEs were bowel obstruction, nausea, hypertension, and hyponatremia.

TRAEs of special interest with atezolizumab were immune-related rash, hepatitis, and pneumonitis (Table 2). Most occurred at grade 1/2 severity. Grade 3/4 immune-related rash occurred in three patients (15%) and grade 4 immune-related pneumonitis occurred in one (5%). Two patients (10%) experienced AEs (pneumonitis and small bowel obstruction) that led to atezolizumab and bevacizumab discontinuation.

The median duration of follow-up was 20.7 months [range, 1.2 (censored)–26.6 months]. Three patients had a PR; hence, the ORR was 15% (95% CI, 3.2–37.9; Table 3; Fig. 1A). The median DOR was not reached (95% CI, 11.3 months–not reached) by RECIST 1.1. Eight additional patients (40%) had SD maintained for ≥12 weeks, resulting in a DCR of 55% (95% CI, 31.5–76.9) with this treatment combination.

All three PRs were durable (11.3–18.9 months), with the longest response still ongoing at the clinical data cutoff (Fig. 1B). Four of eight patients had SD for >8 months, two of whom experienced prolonged disease control for 19.6 and 19.9 months. These two patients remained on study uneventfully until May and August 2019, respectively, when they transitioned to a separate treatment maintenance study (NCT03768063) due to this study's termination.

The median PFS was 4.9 months [range, 1.2–20.2 months (upper end of range censored)]. The median OS was 10.2 months [range, 1.2–26.6 months (upper end of range censored)]. The OS rate at 12 months was 48.75% (95% CI, 26.40–71.10).

Eight patients had FFPE tumor tissue available for FoundationOne genomic profiling, but only six samples were of sufficient quality to evaluate TMB. The median time between sample collection and study treatment administration was 356 days (interquartile range, 267–749 days; range, 8–1,418 days). While we did not formally test the association between biomarkers and outcomes (e.g., with alpha spend), PD-L1 expression on ICs was mildly associated with CD8⁺ T cells and TIL infiltration (Fig. 1A), but none of these biomarkers was clearly associated with response (Fig. 1A). No specimens had PD-L1 expressed on TCs (data not shown). There was no evidence of association between molecular subtypes (immunoreactive, proliferative, mesenchymal, or differentiated) and clinical response using CLOVAR. Of the six patients evaluable for TMB, none had TMB-high tumors (≥16 mutations/Mbp) and there was no apparent association between TMB-intermediate/low status and response. Two patients with BRCA1 deleterious mutations achieved SD as their best overall response.

Analysis of response to treatment by tumor histology showed no clear association (Fig. 1C). Of the three patients who had a PR, two had serous and one had endometrioid ovarian cancer. Of the five patients with disease progression, three had serous and two had endometrioid histology. Of the eight patients with SD, five (including the two patients who had >19-month disease control) had serous histology, one had clear cell carcinoma, and two had adenocarcinoma (not further specified).

There was also no association between treatment response and the number of prior treatment regimens (Fig. 1D). Two of the three patients who had a PR had received ≥3 prior regimens, and one had received ≤2 prior lines of treatment. Of the four patients with SD who maintained disease control for >8 months, three had received ≥3 prior regimens, and one had received ≤2 prior lines of treatment. Of the five patients with disease progression, three had received ≥3 prior regimens, and two had received ≤2 regimens.

The prognosis of patients with platinum-resistant ovarian cancer is poor, with a median PFS of less than 7 months (4). This is because the disease is usually diagnosed at advanced stages and has a high recurrence risk, with most recurrences demonstrating some degree of chemotherapy resistance. Consistent with this, the median PFS in this study was 4.9 months, but treatment with atezolizumab plus bevacizumab led to durable (i.e., >8 months) disease control, with PR and/or SD in seven patients (35%), and a median OS of 10.2 months in this platinum-resistant study population.

Bevacizumab and atezolizumab monotherapy have each been associated with durable responses in patients with recurrent platinum-refractory ovarian cancer (10, 20). In a phase II study in patients with platinum-sensitive and -resistant recurrent ovarian cancer, bevacizumab monotherapy was associated with a 21% response rate including two CRs, a median PFS of 4.7 months, and a median OS of 16.9 months (20). In a phase I study of atezolizumab, two of nine patients with ovarian cancer experienced durable responses (8.1 and >30.6 months), including one CR, with a median PFS of 2.9 months and a median OS of 11.3 months (10).

Despite a modest ORR of 15%, the 55% DCR and prolonged DORs observed in this study are encouraging and consistent with the durable clinical responses seen with other anti-VEGF/anti-PD-L1/PD-1 combinations. The ORR in this study was similar to that of 16.7% observed with nivolumab plus bevacizumab in patients with platinum-resistant ovarian cancer (17), despite the higher number of median prior therapies (three vs. two in the nivolumab study). As in our study, a small proportion of platinum-resistant patients (2/18) had durable responses of approximately 16–18 months to nivolumab plus bevacizumab (17). This finding is also consistent with observations of prolonged disease control in other studies involving immune checkpoint inhibitors (6), where the disease control was potentially improved by the immunomodulatory effects conferred by bevacizumab (13).

Atezolizumab plus bevacizumab continues to be investigated in several cancers. In the randomized arm of a phase I study in patients with unresectable hepatocellular carcinoma (HCC), a significant improvement in the median PFS (the primary endpoint) was observed with atezolizumab plus bevacizumab versus atezolizumab alone (21). Atezolizumab plus bevacizumab was also associated with significant and clinically meaningful improvements in OS (HR, 0.58; P = 0.0006) and PFS (HR, 0.59; P < 0.0001) compared with sorafenib in a phase III study in patients with unresectable HCC, and an ORR of 27% (22). Efforts to elucidate the complementary mechanisms of blocking both angiogenesis and the PD-L1 immune checkpoint are ongoing.

The frequency of AEs observed with this combination was consistent with the well-established side-effect profile of bevacizumab (4) and the known AEs associated with atezolizumab (10, 18). Two patients discontinued treatment because of treatment-related AEs, and no additional or unique safety signals were observed. Trials are being conducted to evaluate whether bevacizumab or tyrosine kinase inhibitors (TKI) are the best anti-VEGF partners for immunotherapy, and TKIs have been associated with a greater frequency of AEs than bevacizumab, likely due to off-target effects. In patients with recurrent endometrial cancer, lenvatinib was associated with treatment-related AEs and grade ≥3 treatment-related AEs in 87% and 59% of patients, respectively; 18% of patients discontinued treatment because of AEs (23). In a phase II study of cediranib for recurrent ovarian cancer, grade 3–4 dose-limiting toxicity was observed at daily doses of 45 mg, requiring a protocol amendment to reduce the dose to 30 mg, and 37% of patients discontinued this dose due to AEs (24).

Expression of PD-L1 on ICs, along with the presence of TILs and CD8 T cells, have been associated with clinical activity of checkpoint inhibitors, alone and in combination with VEGF inhibition (9, 12, 13). None of these biomarkers, however, were found to be linked to response or long-term outcomes in this small study. TMB is another biomarker linked to the activity of immune checkpoint inhibitors, hypothetically as an indicator of tumor neoantigens (10), but no clear association was seen between TMB and response or progression. This is perhaps not surprising, given the small sample size, which makes it difficult to draw definitive conclusions from the biomarker results. In addition, the samples were archival, which may preclude comment about the predictive nature of these findings following multiple prior treatments (25).

Classification of patients with ovarian cancer according to their tumor histologies and molecular classifiers is prognostic in untreated patients with ovarian cancer (2). Although all of the most common histologies (serous, endometrioid, and clear cell) and molecular subtypes (mesenchymal, immunoreactive, proliferative, and differentiated) were represented in our study, none were clearly associated with tumor response to atezolizumab plus bevacizumab. We also observed no association between treatment response and the number of pretreatment regimens. Interestingly, the patients who had the longest DOR (11.3 and 18.9 months) and three of four patients who had durable disease control (8.8–19.9 months) had received three or more prior treatment regimens.

These results helped inform several studies of the combination of atezolizumab plus bevacizumab with chemotherapy in patients with early diagnosed stage IV (NCT03038100), platinum-sensitive relapsed stage III (NCT02891824), and early relapsed ovarian cancer (NCT03353831 and NCT02839707). The hypothesis that PD-L1 is the main biomarker associated with clinical activity in ovarian cancer is being explored using PD-L1 IHC in ICs as a coprimary endpoint in IMagyn050 (NCT03038100).

Study limitations include the small cohort size and the lack of a comparator arm, which are typical of early-stage cancer trials conducted in patients with a wide range of solid tumors. Few samples were available for biomarker studies; these were archival, and thus, may not represent the pretreatment status of the tumor immune microenvironment.

Nevertheless, treatment response or disease stabilization for ≥8 months was demonstrated in seven of 20 patients (35%) with platinum-resistant ovarian cancer who typically have a poor prognosis and limited options for effective treatment. The complementary mechanisms of action of atezolizumab and bevacizumab appear to result in augmented antitumor activity and durable disease control without jeopardizing tolerability, making this an attractive treatment combination for further investigation.

J. Powderly reports funding support from Genentech (clinical trial funding, speakers bureau) during the conduct of the study; funding from AstraZeneca (clinical trial funding, consultancy, advisory), BMS (clinical trial funding, speakers bureau, consultancy), Merck (speakers bureau, advisory role), EMD Serono (clinical trial funding), MacroGenics (clinical trial funding), Incyte (clinical trial funding), RAPT Therapeutics (clinical trial funding), Alkermes (clinical trial funding), Tempest (clinical trial funding), Curis (clinical trial funding, consultancy, advisory role), Corvus (clinical trial funding), AbbVie (clinical trial funding), Top Alliance Biosciences (clinical trial funding), Precision for Medicine (clinical trial funding), MT Group (clinical trial funding), Stemcell Technology (clinical trial funding), and BioCytics (founder equity) outside the submitted work; and discloses that Carolina BioOncology Institute PLLC and BioCytics Inc. are both developing intellectual property for personalized autologous cell therapies. J. Bendell reports grants and other from Arrys Therapeutics (payment to institution for conduct of clinical trial and consulting services), BeiGene (payment to institution for conduct of clinical trial and consulting services), Boehringer Ingelheim (payment to institution for conduct of clinical trial and consulting services), Bristol Myers Squibb (payment to institution for conduct of clinical trial and consulting services), Celgene (payment to institution for conduct of clinical trial and consulting services), Evelo Biosciences (payment to institution for conduct of clinical trial and consulting services performed), Five Prime Therapeutics (payment to institution for conduct of clinical trial and consulting services), Genentech (payment to institution for conduct of clinical trial and consulting services), Innate Pharma (payment to institution for conduct of clinical trial and consulting services), Merck (payment to institution for conduct of clinical trial and consulting services), Merrimack Pharmaceuticals (payment to institution for conduct of clinical trial and consulting services), Novartis (payment to institution for conduct of clinical trial and consulting services), Roche (payment to institution for conduct of clinical trial and consulting services), Seattle Genetics (payment to institution for conduct of clinical trial and consulting services), Taiho Oncology (payment to institution for conduct of clinical trial and consulting services), and Rgenix (payment to institution for conduct of clinical trial and consulting services), other from Cerulean (payment to institution for consulting services), Daiichi Sankyo (payment to institution for consulting services performed), FORMA Therapeutics (payment to institution for consulting services), Janssen (payment to institution for consulting services), Moderna Therapeutics (payment to institution for consulting services), Tolero Pharmaceuticals (payment to institution for consulting services), and Translational Drug Development (payment to institution for consulting services), grants from AbbVie (payment to institution for conduct of clinical trial), Stemcentrx (payment to institution for conduct of clinical trial), ADC Therapeutics (payment to institution for conduct of clinical trial), Agios (payment to institution for conduct of clinical trial), Arcus Biosciences (payment to institution for conduct of clinical trial), Array BioPharma (payment to institution for conduct of clinical trial), AstraZeneca (payment to institution for conduct of clinical trial), Bayer (payment to institution for conduct of clinical trial), Bellicum Pharmaceuticals (payment to institution for conduct of clinical trial), Bicycle Therapeutics (payment to institution for conduct of clinical trial), Blueprint Medicine (payment to institution for conduct of clinical trial), Boston Biomedical (payment to institution for conduct of clinical trial), Calithera Biosciences (payment to institution for conduct of clinical trial), Cyteir Therapeutics (payment to institution for conduct of clinical trial), CytomX (payment to institution for conduct of clinical trial), eFFECTOR Therapeutics (payment to institution for conduct of clinical trial), Eisai (payment to institution for conduct of clinical trial), Lilly (payment to institution for conduct of clinical trial), EMD Serono (payment to institution for conduct of clinical trial), Forty Seven (payment to institution for conduct of clinical trial), Gilead Sciences (payment to institution for conduct of clinical trial), Gossamer Bio (payment to institution for conduct of clinical trial), Incyte (payment to institution for conduct of clinical trial), Leap Therapeutics (payment to institution for conduct of clinical trial), MacroGenics (payment to institution for conduct of clinical trial), MedImmune (payment to institution for conduct of clinical trial), Merus (payment to institution for conduct of clinical trial), Millennium Pharmaceuticals (payment to institution for conduct of clinical trial), NGM Biopharmaceuticals (payment to institution for conduct of clinical trial), Novocure (payment to institution for conduct of clinical trial), Oncologie (payment to institution for conduct of clinical trial), Pieris (payment to institution for conduct of clinical trial), Revolution Medicines (payment to institution for conduct of clinical trial), SynDevRx (payment to institution for conduct of clinical trial), Synthorx (payment to institution for conduct of clinical trial), Takeda Pharmaceuticals (payment to institution for conduct of clinical trial), Tarveda (payment to institution for conduct of clinical trial), Tempest Therapeutics (payment to institution for conduct of clinical trial), TRACON Pharmaceuticals (payment to institution for conduct of clinical trial), and Unum Therapeutics (payment to institution for conduct of clinical trial) outside the submitted work. S.G. Eckhardt reports grants from Genentech during the conduct of the study and grants from Genentech outside the submitted work. L. Molinero reports employment with Genentech (full time employee), and Roche (stockholder). J. Spahn reports other from Genentech (employment, receive salary and stocks) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

J.W. Moroney: Writing-review and editing, material (patient) support. J. Powderly: Writing-review and editing, material (patient) support. C.H. Lieu: Data curation, writing-review and editing, material (patient) support. J.C. Bendell: Data curation, writing-review and editing, material (patient) support. S.G. Eckhardt: Conceptualization, data curation, formal analysis, supervision, writing-review and editing. C.-W. Chang: Formal analysis, writing-review and editing. L. Molinero: Conceptualization, data curation, formal analysis, supervision, methodology, writing-review and editing. J. Spahn: Conceptualization, data curation, formal analysis, supervision, methodology, writing-review and editing. P. Williams: Conceptualization, data curation, formal analysis, supervision, methodology, writing-review and editing. Y.G. Lin: Conceptualization, data curation, formal analysis, supervision, methodology, writing-review and editing. F.S. Hodi: Writing-review and editing, material (patient) support.

This study was supported by F. Hoffmann-La Roche, Ltd. We thank the patients, their families, and the clinical study site investigators and staff for their contributions to the study. Statistical analyses were conducted by Jatinder Singh, MSc. Medical writing assistance for this article was provided by Samantha Santangelo, PhD, of Health Interactions, Inc., and funded by F. Hoffmann-La Roche, Ltd.

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