Checkpoint Blockade: Not Yet NINJA Status in Ovarian Cancer

Epithelial ovarian cancer (EOC) is the most lethal gynecologic malignancy, and although surgery and platinum-based chemotherapy effectively induce initial remission,¹ most women will ultimately succumb to recurrent and therapy-resistant disease. High-grade serous histology is the most common EOC pathology and is molecularly characterized by defects in DNA repair, copy number alterations, microsatellite stable status, and low tumor mutational burden.^2,3 Platinum-resistant ovarian cancer (PROC), defined as cancer that has progressed within 6 months of last platinum exposure, has a median overall survival (OS) of < 16 months,⁴ and new treatment strategies are exigently needed.

THE TAKEAWAY

• In the article that accompanies this editorial, Hamanishi et al⁷ report a multicenter, randomized, open-label phase III trial (NINJA) that evaluated the safety and efficacy of nivolumab compared with chemotherapy in platinum-resistant epithelial ovarian cancer. Nivolumab demonstrated no improvement in overall survival and significantly worse progression-free survival compared with nonplatinum chemotherapy, adding to the growing body of evidence that single-agent immune checkpoint blockade should not be used in epithelial ovarian cancer without first defining reliable biomarkers to identify subpopulations likely to benefit from these agents.

Results of multiple studies of immune checkpoint blockade (ICB) therapy in recurrent EOC have been disappointing, and EOC remains a cancer with no ICB-specific approvals. In both the KEYNOTE-100⁵ and JAVELIN⁶ trials of pembrolizumab and avelumab in recurrent EOC, respectively, the objective response rate (ORR) of these agents in recurrent EOC was in single digits. The NINJA trial⁷ was launched in response to promising single-agent nivolumab activity observed in a small phase II study of women with PROC showing an ORR of 15%, including two durable complete responses.⁸ The NINJA study was designed to evaluate the efficacy and safety of nivolumab monotherapy compared with chemotherapy in women with PROC, reported by Hamanishi et al⁷ in the article that accompanies this editorial. Eligible patients had PROC, fallopian tube or peritoneal (collectively referred to as ovarian) cancer, and could have received no more than one prior line of therapy in the platinum-resistant setting and no prior exposure to ICB or gemcitabine (GEM) or pegylated liposomal doxorubicin (PLD). Patients were stratified by histology (clear-cell [CC] v non-CC carcinoma) and by number of prior lines of therapy in the PROC setting (0 or 1) and randomly assigned 1:1 to nivolumab monotherapy versus GEM or PLD chemotherapy. The predominant histology of enrolled patients was serous, either high grade or unknown. The primary outcome of the study was OS, and key secondary outcomes were progression-free survival (PFS), ORR, duration of response (DOR), time to response, and safety.

The NINJA investigators are to be congratulated for conducting this randomized phase III study, which gives oncologists a definitive answer to the question: Should single-agent ICB be used for women with PROC, especially given that the predominant EOC histology is high-grade serous histology with low tumor mutational burden (TMB) and microsatellite stable? The results of the NINJA study support that there is no justification for the use of ICB in recurrent EOC, unless the cancer is found to be TMB high or microsatellite instable thus qualifying for pembrolizumab on the basis of the cancer site agnostic approval. ICB performed inferiorly to chemotherapy, and the NINJA trial demonstrated significantly worse PFS for nivolumab compared with nonplatinum chemotherapy. The median PFS was 2.0 months in the nivolumab cohort versus 3.8 months in the GEM/PLD group (hazard ratio [HR], 1.5; 95% CI, 1.2 to 1.9; P = .002). No significant difference in OS was observed between the nivolumab and GEM/PLD treatment groups with a median OS worse in the nivolumab arm compared with the GEM/PLD group (10.1 v 12.1 months). In addition, ORR was nonsignificantly lower in the nivolumab group (7.6% v 13.2%; P = .191). For patients who did respond to treatment, a longer median DOR of 18.7 months (95% CI, 2.5 to not evaluable) versus 7.4 months (95% CI, 3.0 to 10.3) was seen with nivolumab. Fewer grade 3 of 4 and any grade treatment-related adverse events occurred in the nivolumab arm.

These negative results from the NINJA trial add to the growing list of trials demonstrating lack of activity of ICB in EOC; compared with other ICB monotherapy trials in EOC, the ORR of nivolumab is comparable with that of pembrolizumab (8.0%)⁵ and avelumab (9.6%).⁶ The significance of the worse PFS and nonsignificantly lowered ORR compared with chemotherapy in this population of women with PROC raises the question of whether anti–programmed cell death-1 (PD-1) therapy may be even less effective in the platinum-resistant setting, but KEYNOTE-100⁵ showed that degree of platinum sensitivity was not predictive of response to ICB. Other clinical biomarkers such as number of prior lines of treatment have also not predicted response to ICB,⁵ and this is clear in the NINJA study in which dismal results of ICB are observed in patients with minimally treated PROC. CC histology has also been identified as a potential biomarker of response,^5,8 but in the NINJA study, CC histology was not predictive of response to nivolumab.⁷ Although the data were negative for the overall population, the median DOR with nivolumab doubled indicated that the few patients who did respond to nivolumab had a durable response. Conversely, the response to chemotherapy in PROC tends to be short-lived, underscoring the importance of identifying the small subset of patients who have a higher chance of benefiting from immunotherapy. Overall, this study further supports what is becoming increasingly clear—that ICB has the potential to induce durable responses in only a very limited subset of patients in PROC, reliable biomarkers to select responders have not been identified, and ICB should not be used as single agents in EOC without further defining appropriate subpopulations, clinical settings, and/or combinations that allow for improved activity.

Given the modest activity of single-agent ICBs in ovarian cancer, combinations of anti-PD-1/programmed cell death ligand 1 (PD-L1) therapy with chemotherapy or targeted therapy have been reported while some trials are currently ongoing. The editorial by Konstantinopoulos and Cannistra⁹ highlighting the results of the IMagyn050 trial¹⁰ also reviewed trials several testing ICB-containing chemotherapy combinations in the newly diagnosed^10,11 and recurrent¹² EOC, all of which reported discouraging results; thus, triplet combinations including ICB, poly(ADP-ribose) polymerase inhibitors, antiangiogenics, and chemotherapy are now being tested. The ongoing MEDIOLA trial reported promising initial results with the triplet combination of olaparib, durvalumab, and bevacizumab in recurrent platinum-sensitive ovarian cancer generating an ORR of 77%.¹³ The combination of pembrolizumab, bevacizumab, and oral cyclophosphamide in recurrent EOC showed an ORR of 47.5% and a durable response in 25%.¹⁴ Taken together, the results of the trials to date suggest distinct biological differences between ovarian and other solid tumors that must be identified and addressed for successful immunotherapy implementation in this lethal disease.

An evolving understanding of the mechanisms driving low response rates to ICB in EOC, including tumor cell intrinsic features such as low TMB and lack of inflamed gene expression profile,¹⁵ and suppressed major histocompatibility complex protein expression,¹⁶ as well as a generalized immunosuppressive tumor immune microenvironment (TIME),^17-25 has catalyzed the development of predictive biomarkers for response to ICB. PD-L1 expression in tumor cells and the TIME remains the most well-studied potential biomarker in EOC, given its correlation with therapeutic response across diverse cancer types.^26,27 This correlation has been observed in EOC, although not uniformly across all studies. KEYNOTE-100 reported an ORR to pembrolizumab of 5% and 17.1% with combined positive score (CPS) ≥ 1 and ≥ 10, respectively,⁵ with all seven complete responses in the CPS ≥ 10 subgroup; however, the incidence of cancers demonstrating CPS ≥ 10 was very low. Stratification by PD-L1 staining in immune cells of ≥ 1% in IMagyn050 did not correlate with response to atezolizumab¹⁰; however, a potential benefit from the addition of atezolizumab was seen in prespecified exploratory analysis of two populations, one exhibiting immune cell PD-L1 expression of ≥ 5% (HR, 0.66; 95% CI, 0.44 to 0.98) and another with tumor cell PD-L1 expression of ≥ 1% (HR, 0.45; 95% CI, 0.19 to 1.02). No differences were seen in OS by PD-L1 expression in the NINJA study, although stratification was by tumor cell (tumor-positive score) PD-L1 ≥ 1%, potentially limiting the sensitivity.⁷ Overall, several challenges limit the utility of PD-L1 as a biomarker,^28,29 and optimization of PD-L1 testing and/or combining it with other methods of assessing the immune-inflamed phenotype of a tumor³⁰ will likely be necessary for its clinical use in EOC.

TMB is another leading biomarker candidate based on the assumption that a high mutational load fosters enhanced immunogenicity.^31,32 Pembrolizumab is now approved for use in all unresectable or metastatic tumors with high TMB as determined using the FoundationOne CDx assay; however, these are rare in EOC.^33,34 No association between TMB and response to pembrolizumab was seen in a prespecified exploratory analysis in KEYNOTE-100, although two tumors with high TMB did exhibit an ORR and PFS double that of low TMB tumors.³⁴ In IMagyn050, most tumors had a low TMB regardless of BRCA1/2 mutation or homologous recombination deficiency status, and neither was associated with response to atezolizumab,³³ underscoring the likely low utility of TMB assessment in EOC. BRCA mutations have also not been shown to be predictive of cancer responsiveness to ICB.^5,6,11 The quest for more sensitive and precise biomarkers to predict ICB response in EOC is robust and includes transcriptomic analyses to identify tumors with an inflamed phenotype or other signatures related to response,^2,34,35 single-cell spatially resolved analyses of the TIME,³⁶ liquid biopsy−based approaches,^37-41 and assessment of the host microbiome^42,43 (Fig 1).

FIG 1.

Summary of potential biomarkers under investigation for predicting response to therapy with immune checkpoint inhibitors. Representative candidates for biomarkers in tumor cells, the tumor microenvironment, and the circulation are depicted. In addition, clinical and host factors under investigation are included. B2M, beta-2 microglobulin; CTC, circulating tumor cell; ctDNA, circulating tumor DNA; GEP, gene expression profile; IFN, interferon; JAK1/2, Janus kinase 1/2; MMR, mismatch repair; ncRNA, noncoding RNA; NK, natural killer; PD-L1, programmed cell death ligand; TAM, tumor-associated macrophage; TCR, T-cell receptor; TMB, tumor mutational burden; TIL, tumor-infiltrating lymphocyte; Treg, regulatory T cell. Created with BioRender.⁴⁴

Despite the low response rates thus far to ICBs in EOC, studies such as the NINJA trial highlight their capability to produce durable responses in a very small subset of patients, challenging the medical and scientific community to uncover the specific barriers limiting the activity of ICB and apply novel technologies to develop predictive strategies but also to develop immunotherapies that extend beyond ICB or help these agents work better. The identification of transformative biomarkers for predicting response to immunotherapy—which ultimately requires a method of testing if a patient has a tumor-reactive T-cell repertoire with the ability to both access the TME and also effectively eliminate cancer cells once there—remains challenging and will likely require incorporation of multiple parameters. Combining patient data from across ICB trials with low numbers of responders may prove useful for discovery of new candidates, which could then be validated in prospective trials. Building on what has already been learned and applying emerging technologies that can characterize tumors and their microenvironment at the single-cell level will undoubtedly lead to improved biomarkers that will not only allow for optimal patient selection for these therapies but will also improve our ability to test novel ICB combinations and sequences in the most appropriate populations.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Collection and assembly of data: Ursula A. Matulonis

Data analysis and interpretation: Ursula A. Matulonis

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO’s conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).