Influence of Genomic Landscape on Cancer Immunotherapy for Newly Diagnosed Ovarian Cancer: Biomarker Analyses from the IMagyn050 Randomized Clinical Trial

In recent years, incorporation of immune checkpoint blockade into clinical practice has changed the treatment landscape for many cancers. However, results have been less spectacular in ovarian cancer. Two randomized phase III trials failed to show benefit from avelumab either alone or combined with chemotherapy (1, 2), and more recently, results from the IMagyn050 randomized phase III trial showed no significant progression-free survival (PFS) benefit from the addition of the anti-programmed death-ligand 1 (PD-L1) immune checkpoint inhibitor (ICI) atezolizumab to standard bevacizumab and chemotherapy for newly diagnosed stage III/IV ovarian cancer (3).

Responses and an extended ‘tail of the curve’ in some trials suggest that a small proportion of patients with ovarian cancer may derive long-term benefit from ICIs (4, 5) but to date, efforts to identify these patients prospectively have had relatively little success. Tumor mutation burden (TMB) has shown predictive potential for single-agent ICI in melanoma and lung cancer (6, 7), which tend to have higher TMB (8), and in gastric cancer (9). However, its relevance and applicability across other solid tumors is less clear (10, 11).

In ovarian cancer, data from non-randomized studies have suggested associations between BRCA1/2 alterations, increased mutations, and increased PD-L1 expression, raising the possibility of enhanced sensitivity to cancer immunotherapy (12, 13). To the best of our knowledge, the potential prognostic and predictive role of TMB, BRCA1/2 mutation, and homologous recombination deficiency (HRD) has not been assessed in randomized clinical trials of ICIs for patients with newly diagnosed ovarian cancer. Therefore, in these prespecified exploratory analyses, we evaluated TMB, BRCA1/2 mutation status, and HRD in samples from women treated in the IMagyn050 randomized phase III trial (3) and explored associations with clinical outcome.

The design of the parent study—the multicenter, double-blind, placebo-controlled, randomized, phase III IMagyn050 trial—has been described in detail previously (3). Briefly, patients with previously untreated epithelial ovarian, peritoneal, or fallopian tube cancer (collectively referred to as ovarian cancer), either postoperative stage III with macroscopic residual disease or stage IV, or a candidate for neoadjuvant therapy with planned interval surgery, were randomized in a 1:1 ratio to receive either atezolizumab 1,200 mg or placebo every 3 weeks for 22 cycles, both in combination with carboplatin plus paclitaxel chemotherapy during cycles 1 to 6 and bevacizumab 15 mg/kg every 3 weeks for 22 cycles. The co-primary endpoints were PFS (per Response Evaluation Criteria in Solid Tumors version 1.1) and overall survival (OS) tested in both the PD-L1–positive and the intent-to-treat (ITT) populations. Stratification factors were International Federation of Gynecology and Obstetrics (FIGO) stage (III vs. IV), Eastern Cooperative Oncology Group performance status (ECOG PS; 0 vs. 1/2), treatment approach (adjuvant vs. neoadjuvant), and PD-L1 status [PD-L1 expression in <1% vs. ≥1% of immune cells (ICs) as a percentage of tumor area, as assessed by the VENTANA SP142 PD-L1 assay (VENTANA, Tucson, Arizona)].

The study was conducted in full conformance with the International Council for Harmonisation (ICH) E6 guideline for Good Clinical Practice and the principles of the Declaration of Helsinki, or the laws and regulations of the country in which the research was conducted, whichever afforded the greater protection to the individual. The study complied with the requirements of the ICH E2A guideline on Clinical Safety Data Management: Definitions and Standards for Expedited Reporting, U.S. FDA regulations and applicable local, state, and federal laws, and the EU Clinical Trial Directive (2001/20/EC). The protocol was approved by institutional review boards or ethics committees at each site. All patients provided written informed consent before any trial-specific procedures or treatment.

Patients were enrolled between March 8, 2017, and March 26, 2019. The data cutoff for the primary analysis, used for the post hoc analyses reported here, was March 30, 2020.

Next-generation sequencing [NGS; FoundationOne CDx assay (Foundation Medicine, Cambridge, Massachusetts)] was performed in samples with evaluable tumor according to local regulations to assess detection of substitutions, insertion and deletion alterations, and copy-number alterations in 324 genes and select gene rearrangements, mutation status in BRCA1 and BRCA2 genes, genomic loss of heterozygosity (gLOH), TMB, and microsatellite instability (MSI) status. Samples with known or likely deleterious tumor germline/somatic BRCA1/2 mutations (excluding variants of unknown significance) were classified as BRCA1/2 mutated. HRD was defined as gLOH ≥ 16%, the cutoff used in the ARIEL3 randomized phase III trial (14). Homologous recombination proficient (HRP) tumors were defined as gLOH < 16%, regardless of BRCA1/2 mutation status. TMB was assessed according to previously described methods (15), with ≥ 10 mutations/megabase (mut/Mb) classified as TMB-high.

All analyses were exploratory and all P values are descriptive. Prevalences of TMB, BRCA1/2 mutation status, and homologous recombination status were compared using Mann–Whitney tests.

This trial is registered with ClinicalTrials.gov, NCT03038100.

NGS data are deposited in the European Genome-phenome Archive at the European Bioinformatics Institute (https://ega-archive.org/) under study accession number EGAS00001006838.

For up-to-date details on Roche's Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see: https://go.roche.com/data_sharing. Anonymized records for individual patients across more than one data source external to Roche cannot, and should not, be linked due to a potential increase in risk of patient re-identification.

Authors from F. Hoffmann-La Roche/Genentech were involved in data analysis and interpretation.

Among the 1,301 patients enrolled in the IMagyn050 trial, samples from 1,050 patients were assessable by NGS. Asian patients were underrepresented in the biomarker-evaluable population compared with the ITT population (15% vs. 23%, respectively), as samples from China were not evaluated, in accordance with local regulations. Gene mutation status was available from all samples, HRD/HRP status from 980, TMB status from 1,024, and MSI status from 1,022.

The genomic landscape of the biomarker-evaluable population is shown in Fig. 1A. gLOH was higher in patients with high-grade serous ovarian cancers (HGSOC) than with other histotypes (median gLOH: 15.8% vs. 7.8%, respectively; P < 0.0001). Deleterious TP53 mutations were associated with both HGSOC and elevated gLOH (P < 0.0001), whereas CCNE1 amplifications found in HGSOC tumors were associated with lower gLOH (P < 0.0001), and were mutually exclusive with BRCA1 and BRCA2 mutations (Supplementary Fig. S1). Patients with BRCA1/2-mutated or HRD tumors tended to be younger than those with BRCA1/2–non-mutated or HRP tumors, respectively, and were more likely to have PD-L1–positive tumors (Table 1). Compared with BRCA1/2–wild-type tumors, BRCA-mutated tumors were associated with: a numerically higher proportion of patients with HRD (76% vs. 33% in the BRCA wild-type subgroup; P < 0.0001), no gross residual disease after surgery (23% vs. 16%, respectively), and baseline ECOG PS of 0 (64% vs. 58%, respectively); and a numerically lower proportion of patients with clear-cell histology (<1% vs. 5%, respectively). Compared with the HRP population, the subgroup with HRD tumors included: a numerically higher proportion of patients reporting as Asian (20% vs. 13% in the HRP subgroup; P = 0.0025), with serous cell histology (90% vs. 82%, respectively), with BRCA1/2-mutated tumors (40% vs. 9%, respectively), and with no gross residual disease after surgery (20% vs. 15%, respectively); and a numerically lower proportion of White patients (74% vs. 82%, respectively) and patients with clear-cell histology (1% vs. 7%, respectively).

The vast majority of patients had low TMB: only 29 (3%) of the 1,024 evaluable samples had TMB ≥ 10 mut/Mb. Only 3 (0.3%) of 1,022 samples were classified as MSI-high (histologies: one mixed, one undifferentiated, one other), all of them with PD-L1 IC expression ≥1%, PD-L1 tumor cell expression <1%, BRCA1/2 wild type, and either HRP or unknown homologous recombination status. All 3 patients with MSI-high tumors were randomized to the control arm. All high-grade serous cases were microsatellite stable. The overall prevalence of BRCA1/2 mutations was 22% [234/1,050; 120/537 (22%) in the atezolizumab-containing arm vs. 114/513 (22%) in the control arm]. The prevalence of HRD was 46% [446/980 overall; 225/502 (45%) in the atezolizumab-containing arm vs. 221/478 (46%) in the placebo arm].

HRD was associated with BRCA1/2 mutation status (median gLOH of 22% in BRCA1/2-mutated vs. 12% in non-mutated tumors; Fig. 1B). However, TMB was low regardless of BRCA1/2 mutation or HRD (Mann–Whitney P < 0.0001 for comparisons by both BRCA1/2 and HRD; Fig. 1B). High TMB (≥ 10 mut/Mb) was observed in 11 (5%) of 232 BRCA1/2 mutated samples versus 18 (2%) of 792 BRCA nonmutant samples (Fisher exact test P = 0.068), and in 15 (3%) of 444 HRD samples versus 12 (2%) of 529 HRP samples (Fisher exact test P = 0.33). There was no correlation between TMB and PD-L1 status (data not shown). While BRCA1/2 mutations were not significantly associated with PD-L1 status (19% prevalence of BRCA1/2 mutation in PD-L1–negative tumors vs. 24% prevalence in PD-L1–positive tumors; Fisher exact test P = 0.0637; Fig. 1C), deleterious alterations in BRCA1, but not in BRCA2, were moderately associated with PD-L1–positive tumors (Supplementary Fig. S2). HRD was enriched in PD-L1–positive tumors (50% prevalence vs. 37% in PD-L1–negative tumors; Fisher exact test P < 0.0001; Fig. 1D).

In the pooled treatment arms, deleterious mutations in BRCA2, RB1, and NF1 were associated with better PFS, whereas activating alterations and amplifications in KRAS, CCNE1, FGF12, and AKT2 were associated with worse PFS (Fig. 2A).

In the control arm, BRCA1/2 mutations were associated with better PFS [hazard ratio, 0.62; 95% confidence interval (CI), 0.46–0.84; median 21.1 months in BRCA1/2-mutated tumors vs. 16.7 months in BRCA1/2–non-mutated tumors], indicating a prognostic role of BRCA1/2 mutation (Fig. 2B). A similar effect was seen in the atezolizumab combination arm (hazard ratio, 0.67; 95% CI, 0.49–0.91; median 21.9 vs. 18.7 months, respectively).

Likewise, in the control arm, HRD was associated with better PFS (hazard ratio, 0.63; 95% CI, 0.49–0.80; median 20.7 months in the HRD subgroup vs. 15.3 months in the HRP subgroup), indicating a prognostic effect of homologous recombination status (Fig. 2C). A similar effect was seen in the atezolizumab combination arm (hazard ratio, 0.73; 95% CI, 0.57–0.94; median 20.7 vs. 18.0 months, respectively).

None of the individual gene alterations from the NGS panel was associated with enhanced atezolizumab treatment effect on PFS (data not shown). Similarly, there was no clear association between atezolizumab treatment effect and BRCA1/2 mutation status, PD-L1 status, or the combination of both (Fig. 3A). The 95% CI for the PFS hazard ratio overlapped with unity for all of the subgroups except the 509 patients with BRCA nonmutant PD-L1–positive tumors (hazard ratio, 0.75; 95% CI, 0.59–0.96; median PFS 20.7 months with atezolizumab-containing therapy vs. 16.4 months in the control arm). The hazard ratio point estimate for the BRCA-mutant PD-L1–positive subgroup was the same, suggesting that the improved outcome with the addition of atezolizumab to chemotherapy and bevacizumab derived from PD-L1 positivity rather than lack of BRCA1/2 mutation.

In the subgroup of patients with high PD-L1 expression (IC ≥ 5%), there was no difference in clinical outcome according to BRCA1/2 mutation status (Supplementary Fig. S3).

In additional analyses, subgroups were defined according to BRCA1 versus BRCA2 mutations. Tumors harbored BRCA1 mutations in 152 patients (14%; 91 germline, 24 somatic, 37 unknown), BRCA2 mutations in 78 patients (7%; 51 germline, 14 somatic, 13 unknown), and both BRCA1 and BRCA2 alterations in 4 patients (0.4%). In both treatment arms, there was a suggestion that PFS was more favorable in patients with BRCA2-mutated tumors than BRCA1-mutated tumors (Fig. 3B), although this was less pronounced in the atezolizumab-containing arm. However, there was no evidence of a treatment benefit from atezolizumab in patients with BRCA2-mutated tumors, but a suggestion of improved PFS with the addition of atezolizumab to chemotherapy and bevacizumab in patients with BRCA1-mutated tumors, particularly those with PD-L1–positive tumors (median PFS of 25.8 months with atezolizumab-containing therapy vs. 18.4 months in the control arm; Fig. 3C).

There was no association between atezolizumab treatment effect and homologous recombination status or PD-L1 status (Fig. 4). When combining these two factors, the predictive effect of PD-L1 status seemed more pronounced in patients with HRD tumors. However, in the subgroup of patients with PD-L1 IC ≥ 5% there was no difference in PFS hazard ratio between subgroups with HRD versus HRP tumors (Supplementary Fig. S3).

Subgroup analyses of PFS according to TMB showed a numerically improved effect of atezolizumab in patients with TMB ≥ 10 mut/Mb, but this was a very small subgroup and 95% CIs were wide (Supplementary Fig. S4).

IMagyn050 is the first randomized double-blind trial in ovarian cancer to demonstrate that neither deleterious BRCA1 or BRCA2 mutations nor HRD improves sensitivity to therapeutic PD-L1 blockade. Similarly, TMB is generally not increased and plays no clear predictive role in ovarian cancer. None of these biomarkers can be recommended for use as a selection criterion for PD-L1–targeting immunotherapy in newly diagnosed ovarian cancer.

In tumor types with higher TMB, such as melanoma and lung cancer, BRCA1/2 alterations are associated with increased neoantigen load and greater sensitivity to ICIs. In a retrospective study of more than 37,000 samples across multiple indications, BRCA1/2-altered tumors had higher median TMB than BRCA1/2 wild-type tumors (16). However, ovarian tumors represented only 2% of samples and of those, only 41 (5%) were BRCA1/2 mutated. Survival analysis in a subset of these patients treated with ICIs showed that those with BRCA2 alteration and high TMB appeared to have the best OS outcome, but outcomes specifically in the ovarian cancer subgroup were not described (16). Furthermore, as all patients received an ICI, potential differences may simply reflect the prognostic effect of BRCA2 alterations.

In the IMagyn050 trial in ovarian cancer, we observed low TMB (<10 mut/Mb) in almost all tumors (97%), irrespective of BRCA1/2 or homologous recombination status. We also found that genomic instability due to BRCA1/2 mutations or HRD was associated with statistically significant but not biologically meaningful increases in TMB. These biological findings are consistent with previous reports of a higher neoantigen load in HRD versus HRP HGSOC (12). We show that the minor TMB increase in HRD or BRCA1/2-mutated tumors is not associated with sensitivity to ICIs nor hypermutation, such as described for tumors with high MSI that are deficient in DNA mismatch repair.

The prognostic role of BRCA mutations (particularly in BRCA2) and HRD observed in IMagyn050 is consistent with previous reports (17, 18), highlighting the importance of stratifying according to BRCA1/2 and/or HRD status in future trials in newly diagnosed ovarian cancer. Our findings are also consistent with the lack of predictive value of BRCA1/2 alterations in patients receiving an ICI (atezolizumab) in the randomized IMpassion130 trial of atezolizumab combined with nanoparticle albumin-bound (nab)-paclitaxel in triple-negative breast cancer (TNBC; ref. 19).

More specifically, BRCA2 status was associated with improved prognosis in IMagyn050 but without a predictive role for atezolizumab. Of note, there was a numerical effect favoring atezolizumab-containing therapy among the subgroup of patients with BRCA1-mutated tumors, notwithstanding the caveat of the small sample size. Superficially, this contrasts with findings reported by Samstein and colleagues (20), which suggested that mutations in BRCA2 but not BRCA1 were associated with improved outcomes. However, all patients received ICIs and therefore it is impossible to differentiate between prognostic and predictive effects. Moreover, few patients in Samstein and colleagues’ study had ovarian cancer and neither of the patients with HGSOC and deleterious BRCA2 mutations showed a clinical response to the ICI. It is plausible, therefore, that those who responded to ICIs had a very different tumor microenvironment from the BRCA2-mutated HGSOCs. Interestingly, PARP inhibitors are clinically active in both ovarian cancer and TNBC; therefore, the tumor characteristics where the BRCA2 mutation resides may differentially predispose to ICI sensitivity.

There is no evidence from the present analysis to support use of TMB as a predictive biomarker for immunotherapy in ovarian cancer. Emerging data suggest that weighting all mutations identically when calculating TMB score may miss important information about the type of mutation, with certain mutations generating more immunogenic neoantigens than the more common nonsynonymous single-nucleotide mutations. There may also be differences between inflamed and non-inflamed tumors (10). In an analysis of almost 1,000 patients with ovarian cancer reported by Fan and colleagues (21), higher TMB was associated with higher CD8+ T-cell infiltration but also better PFS and OS, lower clinical stages, and tumor-free status.

Our analyses showing no correlation between PD-L1 status and BRCA1/2 mutation in ovarian cancer contradict early reports that BRCA1/2-mutated HGSOC was associated with increased PD-L1 expression in tumor-infiltrating ICs (but not tumor cells) compared with HRP tumors (12) but are consistent with recently published analyses from the randomized IMpassion130 trial in metastatic TNBC (19).

This analysis of a double-blind, randomized, placebo-controlled trial offers an important strength compared with most previous reports in the literature. While Liu and colleagues (22) found no association between clinical benefit from immunotherapy and BRCA1/2 mutation, HRD, or TMB in recurrent ovarian cancer, it is not possible to differentiate between prognostic and predictive effects in a single-arm study. In contrast, the efficacy of immunotherapy versus placebo was assessed in our analyses, thus enabling separation of disease-related versus treatment-related effects.

HRD is a frequent feature of HGSOC, as observed in this analysis, thus a potential weakness of the present trial is the grouping of all histologies for analyses according to homologous recombination status. Non-HGSOC tumors are usually HRP and BRCA1/2 wild type; therefore, segmenting histologic subgroups within post hoc biomarker-identified subgroups would result in sample sizes too small for meaningful interpretation. Another potential criticism is the lack of information on tumor-infiltrating lymphocytes (TIL), which have also shown prognostic value in ovarian cancer independent of HRD (23). Analyses of TILs and other tumor immune biomarkers, such as T-cells (cytotoxic and regulatory), myeloid populations, and other immune-based gene expression signatures are ongoing in the IMagyn050 trial.

Notwithstanding these limitations, the analyses reported here provide important new information from a randomized phase III trial challenging the hypothesis that BRCA2 mutation status, HRD, and/or high TMB predict clinical benefit from immune checkpoint blockade in ovarian cancer. On the other hand, we observed an intriguing hint that BRCA1 mutation may predict for enhanced effect of atezolizumab-containing therapy on PFS. There was a hint that the prognosis for patients with BRCA1-mutated tumors, which was less favorable than for those with BRCA2-mutated tumors, can perhaps be improved with the addition of atezolizumab to chemotherapy and bevacizumab. Sample sizes are small, but this finding merits exploration in other datasets to try to establish robust markers potentially enabling identification of those patients with newly diagnosed ovarian cancer who may benefit from immunotherapy. This may also have implications for ongoing trials of immunotherapy in combination with PARP inhibitors, which may show higher benefit in patients with BRCA1-mutated disease.

C.N. Landen reports personal fees from Roche during the conduct of the study as well as personal fees from Mercy Bio outside the submitted work. L. Molinero is an employee of Genentech and holds stock in Roche. H. Hamidi is an employee of Roche and holds stock in Genentech. J. Sehouli reports grants from Roche during the conduct of the study as well as grants and personal fees from GSK, AstraZeneca, Clovis, MSD, and Pfizer outside the submitted work. K.N. Moore reports personal fees from AstraZeneca, Aravive, Addi, Alkemeres, Blueprint Medicines, Clovis, Eisai, EMD Serono, GSK/Tesaro, Genentech/Roche, Hengrui, Immunogen, Inxmed, Imab, Lilly, Merck, Mersana, Mereo, Novartis, OncXerna, Onconova, Tarveda, VBL Therapeutics, and Verastem during the conduct of the study as well as personal fees from Genentech/Roche outside the submitted work; in addition K.N. Moore is associate director for GOG Partners and is on the GOG Foundation board of directors (uncompensated). M. Bookman reports advisory board participation with Genentech/Roche and Merck Sharp & Dohme and is chair of an independent data monitoring committee at Immunogen. K. Lindemann reports other support from Roche during the conduct of the study as well as personal fees from MSD, Eisai, AstraZeneca, and Nycode and grants from GSK outside the submitted work. R. Berger reports other support from Roche during the conduct of the study as well as other support from Merck, PharmaMar, AstraZeneca, Novartis, Roche, and BIOCAD outside the submitted work. M. Beiner reports other support from Meir Medical Center during the conduct of the study. A. Okamoto reports grants from Taiho Pharmaceutical Co. Ltd., Fuji Pharma Co. Ltd., Kissei Pharmaceutical Co. Ltd., ASKA Pharmaceutical Co. Ltd., Kaken Pharmaceutical Co. Ltd., Meiji Holdings Co. Ltd., Nippon Shinyaku Co. Ltd., Tsumura & Co., Mochida Pharmaceutical Co. Ltd., Linical Co. Ltd., Daiichi Sankyo Co. Ltd., Pfizer Japan Inc., Terumo Corporation, Eisai Co. Ltd., and Gyne Mom Co. Ltd.; grants and personal fees from MSD K.K., Chugai Pharmaceutical Co. Ltd., Takeda Pharmaceutical Company Ltd., and AstraZeneca K.K.; and personal fees from Zeria Pharmaceutical Co. Ltd. and Johnson & Johnson K.K. outside the submitted work. C. Aghajanian reports personal fees from Eisai/Merck, Roche/Genentech, AbbVie, AstraZeneca/Merck, and Repare and grants from AbbVie, Roche/Genentech, and AstraZeneca outside the submitted work; in addition, C. Aghajanian serves on the board of directors of GOG Foundation (travel cost reimbursement for attending meetings) and NRG Oncology (unpaid). P.H. Thaker reports grants and personal fees from Merck, and GlaxoSmithKline and personal fees from Clovis Oncology, AstraZeneca, Iovance, Aadi Biosciences, Novocure, Celsion/Immunon, Immunogen, Seagen, Agenus, Mersana, Eisai, and Incyte outside the submitted work. S.V. Blank reports grants and other support from Roche during the conduct of the study. V.K. Khor is an employee of Genentech and holds stock in Roche. C.-W. Chang is an employee of Genentech and holds stock in Roche. Y.G. Lin is an employee of Genentech and holds stock in Roche. S. Pignata reports grants and personal fees from Roche during the conduct of the study. No disclosures were reported by the other authors.

C.N. Landen: Resources, investigation, writing–original draft, writing–review and editing. L. Molinero: Conceptualization, data curation, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. H. Hamidi: Data curation, formal analysis, validation, visualization, writing–review and editing. J. Sehouli: Resources, investigation, writing–review and editing. A. Miller: Writing–review and editing. K.N. Moore: Conceptualization, resources, investigation, methodology, writing–review and editing. C. Taskiran: Resources, investigation, writing–review and editing. M. Bookman: Conceptualization, resources, investigation, writing–review and editing. K. Lindemann: Resources, investigation, writing–review and editing. C. Anderson: Resources, investigation, writing–review and editing. R. Berger: Writing–review and editing. T. Myers: Resources, investigation, writing–review and editing. M. Beiner: Resources, investigation, writing–review and editing. T. Reid: Resources, investigation, writing–review and editing. E. Van Nieuwenhuysen: Resources, investigation, writing–review and editing. A. Green: Resources, investigation, writing–review and editing. A. Okamoto: Resources, investigation, writing–review and editing. C. Aghajanian: Resources, investigation, writing–review and editing. P.H. Thaker: Resources, investigation, writing–review and editing. S.V. Blank: Resources, investigation, writing–review and editing. V.K. Khor: Conceptualization, writing–review and editing. C.-W. Chang: Data curation, formal analysis, validation, visualization, writing–review and editing. Y.G. Lin: Conceptualization, methodology, writing–review and editing. S. Pignata: Conceptualization, resources, investigation, methodology, writing–review and editing.

This work was supported by F. Hoffmann-La Roche Ltd. The trial was sponsored, designed, and conducted by F. Hoffmann-La Roche/Genentech in collaboration with the Gynecologic Oncology Group Foundation (GOG-F) and the European Network for Gynaecological Oncological Trial Groups (ENGOT) according to what was subsequently described in 2019 as ENGOT model C. The sponsor (F. Hoffmann-La Roche/Genentech) was involved in data collection, analysis, and interpretation. Medical writing assistance was provided by Jennifer Kelly, MA (Medi-Kelsey Ltd, Ashbourne, UK), funded by F. Hoffmann-La Roche Ltd. No grant number is applicable.

We thank the patients and their families and the investigators, study groups, and clinical sites for their contributions to the IMagyn050 trial and this analysis. IMagyn050/GOG 3015/ENGOT-OV39 is sponsored by F. Hoffmann-La Roche Ltd.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.