Clinicogenomic Landscape of Histologic Subtypes in Ovarian Cancer: Real-World Evidence From a Japanese Nationwide Cohort.

In all, 5,395 patients with ovarian cancer were identified in the C-CAT database (Fig 1 ). From this group of 5,395 patients, several study cohorts were defined to address specific study objectives. The pretreatment cohort (n = 3,150) comprised patients with available histologic annotation and CGP data from tumor samples collected before systemic therapy. From the pretreatment cohort, two additional cohorts were established, as illustrated in Figure 1 : a pretreatment mutation profile cohort (n = 2,779) and a survival analysis cohort (n = 1,701). In addition, a post-treatment cohort (n = 189) was established comprising patients with CGP data derived from a tumor sample collected after the first line of systemic therapy. A subset of the post-treatment cohort was used to establish the post-treatment mutation profile cohort (n = 158; Fig 1 ). The demographic and clinical characteristics of the pretreatment cohort are summarized in the Data Supplement (Table S3). Across the pretreatment cohort, 51%, 2.7%, 30%, 8.9%, and 7.3% of patients were classified as HGSOC, LGSOC, CCOV, EOV, and MOV, respectively. Although only approximately 30% of patients had FIGO stage information at the time of diagnosis, there was a significant difference in FIGO stages at diagnosis among the different histologic subtypes. Approximately 95% of patients with HGSOC were diagnosed at advanced stages (FIGO stages III/IV); by contrast, 40%-50% of patients with nonserous histologic subtype (CCOV, EOV, or MOV) were diagnosed at early stages (FIGO I/II). Consistent with stage at the time of diagnosis, more patients with a nonserous histology were diagnosed at age <65 years than patients with HGSOC (79%-83% v 65%, respectively). Notably, in patients <65 years, the incidence of CCOV increased markedly from 2016 to 2024 (Data Supplement, Fig S1). This trend is consistent with findings from another Japanese cohort study based on data from 2002 to 2015. 30 Regarding microsatellite status, 7.1% of EOV tumors were classified as having high microsatellite instability (MSI); however, the prevalence of high MSI was insignificant in the other histologic subtypes. All other reported clinical characteristics were relatively balanced among the five histologic groups. Treatment regimens of the first line of systemic therapy for the entire patient cohort (January 2016–December 2024) are shown in the Data Supplement (Figure S2A). Overall, there were no significant changes in treatment regimens during this period. Following the approval of PARPi in Japan in 2018, 31 their use, presumably in the maintenance setting, started to increase but plateaued after 2020. The treatment regimens in the first four lines of therapy are summarized in the Data Supplement (Figures S2B and S3). As a first line of systemic therapy, 77.1% of patients received platinum-based therapy, with or without anti-vascular endothelial growth factor (VEGF) therapy (Data Supplement, Fig S3). The percentage of patients who received platinum-based therapy in the second, third, and fourth lines of treatment was 49.9%, 43.6%, and 33.4%, respectively, reflecting the increasing number of patients who become platinum resistant over the course of their treatment journey. rwOS for patients who received platinum-based chemotherapy as first-line standard treatment across subtypes is shown in Figure 2 . Compared with the most common subtype, HGSOC, patients with LGSOC had longer rwOS, whereas those with CCOV and MOV had shorter rwOS. EOV showed rwOS comparable with HGSOC. A similar trend was observed among all patients and among those diagnosed at advanced stages (Data Supplement, Figs S4 and S5). These findings are broadly consistent with previous studies. 32 , 33 Survival outcomes by histologic subtype in patients receiving platinum-based chemotherapy as first-line systemic treatment. (A) Kaplan-Meier survival curves showing rwOS for each of the five ovarian cancer histologic subtypes. (B) Forest plot for each histologic subtype versus the HGSOC subtype based on rwOS. CCOV, clear cell ovarian cancer; EOV, endometrioid ovarian cancer; HGSOC, high-grade serous ovarian cancer; HR, hazard ratio; LGSOC, low-grade serous ovarian cancer; MOV, mucinous ovarian cancer; rwOS, real-world overall survival. Because samples were analyzed using a tumor-only panel, we evaluated mutation-detection bias by comparing HGSOC mutation frequencies between the TCGA 9 , 34 , 35 and C-CAT cohorts. Unfiltered frequencies were higher in C-CAT. When restricted to OncoKB-annotated oncogenic mutations (oncogenic, likely oncogenic, or resistance), frequencies were comparable with TCGA except for BRCA1 (Data Supplement, Fig S6). This likely reflects pathogenic germline mutations based on germline BRCA testing, variant allele frequency, and diagnosis age distribution (Data Supplement, Fig S7). Therefore, subsequent analyses included only oncogenic mutations. We first analyzed mutational profiles of the five EOC subtypes using CGP data from tumor samples collected before first-line systemic therapy (Fig 3 ; Data Supplement, Fig S8; Table S4). To the best of our knowledge, this is the largest study to investigate the mutational profiles of untreated tumors in subtypes other than HGSOC. As reported previously for Western cohorts, 9 there was a high frequency (94%) of TP53 mutations in the HGSOC group. The HGSOC group also had the highest BRCA mutation frequency (18%) among all the subtypes, with both BRCA1 (15%) and BRCA2 (3%) mutations found (Fig 3 A; Data Supplement, Table S4). The mutational profile of the rare LGSOC subtype was distinct from the HGSOC subtype, with KRAS being the most frequently mutated gene (36%) and TP53 mutations occurring in only 13% of patients in the LGSOC group (Data Supplement, Fig S8A). The CCOV group had several driver mutations that are not commonly observed in HGSOC: ARID1A (66%), PIK3CA (51%), and ERBB2 (24%) were the most frequently mutated genes in the CCOV group (Fig 3 B). Another feature of the CCOV group was the high frequency (22%) of TERT promoter mutations, which are rarely seen in other types of gynecologic cancers. 36 The mutational landscape in the EOV group was similar to that reported for endometrial cancers 37 ; for example, alterations in TP53 , ARID1A , PIK3CA , KRAS , and CTNNB1 were frequent in the EOV group (Data Supplement, Fig S8B). The MOV group had high rates of KRAS and TP53 mutations, with both genes altered in more than 50% of patients (Fig 3 C). Furthermore, a deletion of the 9p21.3 locus containing CDKN2A/B and MTAP was observed in nearly half the patients in the MOV group. OncoPlots showing genomic alterations in pretreatment samples by histologic subtype. (A) High-grade serous ovarian cancer (n = 1,414); (B) clear cell ovarian cancer (n = 835) and (C) MOV (n = 199). MOV, mucinous ovarian cancer; TMB, tumor mutational burden. De novo genetic alterations are a common mechanism underlying acquired resistance to targeted antitumor therapy. 38 We compared mutation frequencies in tumors after the first line of systemic therapy (post-treatment mutation profile cohort; Fig 1 ) with those in the pretreatment mutation profile cohort. Notably, the frequency of KEAP1 and NFE2L2 mutations that drive resistance to ferroptosis was significantly higher in HGSOC and CCOV, respectively, in the post-treatment mutation profile cohort (Fig 4 ). Specifically, compared with the pretreatment mutation profile cohort, the frequency of KEAP1 mutations was approximately 7-fold in the HGSOC post-treatment mutation cohort (0.42% v 2.97%) and the frequency of NFE2L2 alterations was approximately threefold in the CCOV post-treatment mutation profile cohort (3.83% v 11.76%; Data Supplement, Table S4). Changes in mutation frequencies before and after treatment. (A) SNV/Indel; (B) SNV/Indel + CNA. Each point represents a gene. The x-axis shows odds ratios, and the y-axis shows nominal P values calculated using Firth logistic regression. Point size reflects mutation frequency in the post-treatment mutation profile cohort. Owing to the small sample size, LGSOC, EOV, and MOV subtypes were not included in the analyses. CCOV, clear cell ovarian cancer; EOV, endometrioid ovarian cancer; HGSOC, high-grade serous ovarian cancer; LGSOC, low-grade serous ovarian cancer; MOV, mucinous ovarian cancer; SNV, single nucleotide variant. To identify potential biomarkers for clinical outcomes, we used multivariable survival analyses to assess correlations between gene alterations and rwOS. Histologic subtype-specific associations were observed (Fig 5 A; Data Supplement, Table S2). In the HGSOC group, PIK3CA and FBXW7 mutations were strongly associated with poor prognosis (Figs 5 A and 5 C; Data Supplement, Fig S9). Although CCNE1 amplification has been reported to confer chemotherapy resistance, 11 we did not confirm this association. CCNE1 amplifications varied by ECOG performance status (Data Supplement, Fig S10), which were included in the multivariable analysis and may account for the lack of prognostic impact. CDKN2A/B deletion was consistently associated with worse prognosis across CCOV, MOV, and EOV (Fig 5 C), suggesting the role of cell cycle pathways in chemoresistance. ERBB2 mutations were associated with shorter survival in CCOV patients receiving a combination of anti-VEGF therapy, chemotherapy, and platinum treatment (Fig 5 D). This trend was not observed in those who received platinum-based chemotherapy (Data Supplement, Fig S11). Associations between BRCA1/2 alterations and longer OS were observed (Figs 5 A and 5 C), particularly in patients who received platinum-based therapy (Figs 5 B and 5 D). Comprehensive prognostic analysis of mutations associated with rwOS in ovarian cancer. Volcano plots show comprehensive multivariable analysis of the association between individual gene mutations (single nucleotide variants and copy number alterations) and rwOS, adjusted for age and ECOG performance status. (A, C) Subtype-specific analysis of the effect of mutations on rwOS: (A) SNV/Indel; (C) SNV/Indel + CNA. (B, D) Subtype- and treatment-specific analysis of the effect of mutations on rwOS: (B) SNV/Indel; (D) SNV/Indel + CNA. Horizontal dashed lines indicate nominal P value thresholds of 0.05 (red) and 0.1 (black), whereas vertical dashed lines represent HR thresholds of 0.8 and 1.2, respectively. The size of the symbols reflects the number of patients harboring each mutation: <10, <50, and ≥50. Owing to the small sample size, LGSOC subtype was not included in the treatment-specific analyses. CCOV, clear cell ovarian cancer; CNA, copy number alternation; ECOG, Eastern Cooperative Oncology Group; EOV, endometrioid ovarian cancer; HGSOC, high-grade serous ovarian cancer; HR, hazard ratio; LGSOC, low-grade serous ovarian cancer; MOV, mucinous ovarian cancer; PARPi, poly(ADP-ribose) polymerase inhibitor; Pt, platinum; rwOS, real-world overall survival; SNV, single nucleotide variant; VEGF, vascular endothelial growth factor. Although there were no significant associations between TP53 mutations and rwOS in any ovarian cancer subtype (Fig 5 ), the complexity of TP53 mutations and their therapeutic implications 26 , 27 prompted us to further assess correlations between TP53 mutation subtypes and clinical outcomes. TP53 mutations were categorized into four groups according to their impact on p53 protein structure and function 27 : DNA contact mutations, structural mutations, truncation mutations, and other missense mutations. A statistically significant association between structural mutations and shorter survival was seen in EOV ( P = .037), and a trend for such an association ( P = .094) was present in HGSOC (Fig 5 A; Data Supplement, Fig S12).

In this study, we analyzed Japanese patients with ovarian cancer using the C-CAT RWD database, focusing on clinical characteristics, treatment outcomes, and tumor mutational profiles across histologic subtypes. Analysis of the C-CAT ovarian cancer cohort revealed that CCOV accounted for 30% of EOCs, consistent with its higher prevalence in Asian than Western populations. 39 , 40 Notably, the incidence of CCOV increased among patients age <65 years between 2016 and 2024, suggesting a rising trend (Data Supplement, Fig S1). A previous large-scale epidemiologic study has suggested a potential association between declining total fertility rates and the rising incidence of ovarian cancer, especially in Japan and South Korea. 39 This relationship may be mediated by the increasing prevalence of endometriosis, which is recognized as a risk factor for CCOV. 41 Further epidemiologic studies are needed to clarify the reasons for the increased incidence of CCOV. This study provides the first comparison of mutational profiles in Asian ovarian cancer patients before and after systemic therapy, with a focus on histologic subtypes. Through this approach, mutations whose frequencies were altered by treatment were identified. Notably, frequencies of KEAP1 mutations in HGSOC and NFE2L2 mutations in CCOV were elevated after treatment (Fig 4 ). Ferroptosis has been implicated in tumor progression and resistance to platinum-based chemotherapy in ovarian cancer. 42 , 43 The KEAP1-NFE2L2 pathway plays a central role in this process. Our findings provide the clinical evidence supporting ferroptosis as a potential therapeutic target in a large-scale ovarian cancer cohort. These results also underscore the importance of mutational profiling that incorporates both histologic subtype and treatment stage. Large-scale, subtype-specific analysis highlighted genetic mutations associated with clinical outcomes. These observations may lead to the development of novel therapeutic approaches. In the case of HGSOC, survival was significantly shorter in patients harboring PIK3CA mutations (Fig 5 ; Data Supplement, Fig S9). Alpelisib, an oral small-molecule inhibitor of PI3Ks, has been evaluated in platinum-resistant HGSOC in a phase III study. 44 In that study, the experimental arm did not exhibit improved progression-free survival (PFS) and OS. Importantly, patients enrolled in the trial were not selected for PIK3CA mutations. 44 In addition, FBXW7 mutations were also associated with poor survival in our cohort. FBXW7 encodes an E3 ubiquitin ligase that regulates the degradation of several oncogenic proteins. 45 Consistent with this role, FBXW7 mutations or reduced expression has been reported to activate several oncogenic pathways, including mTOR signaling. 46 - 48 Together, these findings support the notion that activation of the PI3K-mTOR pathway may contribute to treatment resistance in HGSOC. Therapeutic target genes in other cancer types were frequently mutated in CCOV, including PIK3CA , KRAS , BRCA1/2 , and ERBB2 (HER2). Of these, ERBB2 was the most frequently mutated in CCOV (27%) compared with other subtypes (Fig 3 B). A recent case report showed that trastuzumab deruxtecan (T-DXd), an antibody-drug conjugate targeting HER2, achieved notable therapeutic efficacy in a patient with HER2-positive CCOV. 49 However, the ongoing phase 3 study of T-DXd currently focuses on HGSOC. 50 Considering the high frequency of HER2 amplification and the promising response reported, T-DXd may represent a potential therapeutic option for CCOV. We also observed that CDKN2A deletions were present in approximately half of MOV cases and were associated with worse outcomes (Figs 3 C and 5 ). In a phase 2 study of patients with ovarian cancer, palbociclib demonstrated only modest clinical activity, but exploratory biomarker analyses revealed that CDKN2A deletion is a potential predictor of sensitivity to palbociclib. 51 Given the negative impact of CDKN2A / B alterations on survival seen in our analyses, further evaluation of cyclin-dependent kinase four and 6 (CDK4/6) inhibitors including palbociclib is warranted. In addition, frequent codeletion of CDKN2A / B and MTAP was observed in this context (Fig 3 C). This finding suggests that protein arginine methyltransferase 5 (PRMT5) inhibitors, which induce synthetic lethality in MTAP-deficient tumors, could offer a promising therapeutic strategy. 52 , 53 TP53 was mutated in more than 90% of HGSOC tumors. However, there is a significant complexity of TP53 mutations, with different types of mutations having distinct effects on p53 protein structure and function. 26 , 27 In this study, a trend was observed suggesting that structural mutations in TP53 may be linked to worse outcomes (Data Supplement, Fig S12). Rezatapopt (PC14586), a first-in-class, small-molecule reactivator of p53-mutant proteins harboring the structural mutation Y220C, has demonstrated promising antitumor activity in ovarian cancer. In the ongoing PYNNACLE phase II study, rezatapopt achieved an objective response rate of 43% in the ovarian cancer cohort (n = 44), including one complete response and 17 confirmed partial responses. 54 - 56 Our findings provide preliminary support for the hypothesis that normalizing structurally altered p53 proteins could be therapeutically relevant. We recognize that this study has several limitations. First, data completeness is limited; for example, stage information was available for only 30% of patients. Second, clinical end points in RWD are limited. Owing to a lack of disease progression information in the C-CAT database, we could not analyze PFS or accurately assess platinum sensitivity. However, PFS should be interpreted with caution in ovarian cancer. This is because it may not accurately reflect outcomes in the setting of PARP inhibitor maintenance therapy. 57 Third, the sample size for LGSOC was small (n = 76), limiting statistical power. Finally, the findings may not be generalizable to other East Asian populations because of differences in health care systems. Complementary analyses using Korean data sets could provide further insights. 58

Ovarian cancer is the leading cause of death among gynecologic cancers, with an estimated 324,000 newly diagnosed cases and 204,000 deaths annually worldwide. 1 , 2 Epithelial ovarian cancer (EOC) is the major histologic type, accounting for more than 95% of ovarian malignancies. 3 EOCs are classified into five histologic subtypes: high-grade serous (HGSOC), low-grade serous (LGSOC), clear cell (CCOV), endometrioid (EOV), and mucinous (MOV). 4 Each histologic subtype of ovarian cancer is associated with distinct risk factors. 5 These factors differ across racial and ethnic groups, leading to significant regional variation in subtype distribution. 6 , 7 For example, HGSOC is most prevalent in Western countries, while EOV and CCOV are more common in northern Africa and eastern Asia, respectively. 7 Key Objective Can large-scale real-world clinicogenomic data reveal subtype-specific genomic features and clinically relevant biomarkers across epithelial ovarian cancer (EOC) subtypes, particularly those underrepresented in Western cohorts? Knowledge Generated In a nationwide Japanese cohort of 5,395 patients, nonserous EOC—especially clear cell ovarian cancer (CCOV)—was more prevalent than in Western populations. Subsequent analyses revealed distinct subtype-specific clinicogenomic profiles. ARID1A and PIK3CA mutations were frequent in CCOV, while TP53 mutations were nearly universal in HGSOC. Post-treatment samples showed enrichment of KEAP1 and NFE2L2 mutations in HGSOC and CCOV, suggesting ferroptosis-related resistance mechanisms. Biomarker analyses further identified survival associations, including poorer outcomes with PIK3CA mutations in HGSOC and CDKN2A/B deletions across CCOV, endometrioid, and mucinous subtypes, whereas BRCA1/2 alterations in HGSOC were linked to improved survival. Relevance These findings provide real-world evidence to support biomarker-driven strategies, particularly for nonserous EOC subtypes more common in Asian populations. Key Objective Can large-scale real-world clinicogenomic data reveal subtype-specific genomic features and clinically relevant biomarkers across epithelial ovarian cancer (EOC) subtypes, particularly those underrepresented in Western cohorts? Knowledge Generated In a nationwide Japanese cohort of 5,395 patients, nonserous EOC—especially clear cell ovarian cancer (CCOV)—was more prevalent than in Western populations. Subsequent analyses revealed distinct subtype-specific clinicogenomic profiles. ARID1A and PIK3CA mutations were frequent in CCOV, while TP53 mutations were nearly universal in HGSOC. Post-treatment samples showed enrichment of KEAP1 and NFE2L2 mutations in HGSOC and CCOV, suggesting ferroptosis-related resistance mechanisms. Biomarker analyses further identified survival associations, including poorer outcomes with PIK3CA mutations in HGSOC and CDKN2A/B deletions across CCOV, endometrioid, and mucinous subtypes, whereas BRCA1/2 alterations in HGSOC were linked to improved survival. Relevance These findings provide real-world evidence to support biomarker-driven strategies, particularly for nonserous EOC subtypes more common in Asian populations. The mutational landscapes of ovarian cancer have been well characterized. 8 - 10 In addition, because platinum-based chemotherapy has been the standard-of-care primary systemic treatment for ovarian cancers, biomarkers associated with platinum sensitivity or resistance have been identified, including CCNE1 amplification. 11 However, these studies focused almost exclusively on HGSOC because of its high prevalence in Western countries. Consequently, less information is available for the less common histologic subtypes. 12 Real-world data (RWD) offer an opportunity to address these gaps. The Center for Cancer Genomics and Advanced Therapeutics (C-CAT) database, a centralized RWD repository in Japan, provides a unique resource. The C-CAT database was established after the start of the nationwide cancer genomic medicine program to optimize systemic therapy based on comprehensive genomic profiling (CGP). 13 - 15 The database includes comprehensive clinical and genomic data, such as cancer type, treatment details, and adverse events. 16 Importantly, C-CAT data are enriched with cancers and subtypes more prevalent in East Asia, offering a unique advantage over frequently used RWD sources in the literature. Although the primary goal of C-CAT is to support personalized cancer therapy, the integration of clinical and CGP data also enables identification of drug resistance biomarkers and novel therapeutic targets. For example, a pan-cancer analysis using C-CAT data revealed associations between genetic alterations and certain chemotherapy regimens. 17 Moreover, the success rate of clinical trials has improved when the association between the disease and the target molecule is supported by genetic evidence. 18 - 21 This underscores the importance of integrating genomic and clinical data to accelerate novel drug development and precision oncology. The aim of this study was to use C-CAT data to investigate several important scientific and clinical aspects of ovarian cancers that currently remain largely unaddressed. A previous study analyzing data for 1,606 patients with ovarian cancer from the C-CAT database (data cutoff May 2022) suggested distinct mutational profiles across histologic subtypes. 22 However, that study did not analyze associations between gene mutations and clinical outcomes. Moreover, CGP data derived from both before and after the first line of systemic therapy were included without considering the potential effect of treatment on gene alterations. Thus, the overall objectives of the present study were to (1) summarize the clinical characteristics and treatment patterns of Japanese patients with EOC; (2) characterize and compare the mutational landscapes across different histologic ovarian cancer subtypes, stratified by tumor sample collection time (before or after the first line of therapy); and (3) discover potential molecular biomarkers associated with clinical outcomes.

This retrospective observational study used data from the C-CAT database, which includes deidentified clinicogenomic data collected nationwide in Japan under the national health insurance program. As of June 17, 2025, 5,395 patients with ovarian cancer were registered, and two cohorts were created: pretreatment (n = 3,150; samples collected before first-line systemic therapy) and post-treatment (n = 189; samples collected after first but before second line; Fig 1 ). Patient selection flowchart. STROBE diagram summarizing the selection process for patients with epithelial ovarian cancer in this study. C-CAT, Center for Cancer Genomics and Advanced Therapeutics; CGP, comprehensive genomic profiling; OV, ovarian cancer. Genomic data were obtained from tissue-based CGP tests approved in Japan (FoundationOne CDx, NCC Oncopanel, GenMineTOP). 23 Variants included SNVs, InDels, CNAs, and noncoding alterations; annotations used OncoKB v6.1 (GRCh38). 24 , 25 TP53 mutations were subclassified as DNA-contact, structural, truncating, or other. 26 - 28 Clinical features included age, Eastern Cooperative Oncology Group (ECOG) performance status, smoking history, treatment lines, and regimens. Regimens were defined by drugs initiated on the same date and grouped by mechanism of action (Data Supplement, Table S1). Real-world overall survival (rwOS) was calculated from the start of first systemic therapy to death from primary disease. Survival analyses were limited to patients diagnosed before December 31, 2024, to ensure a minimum follow-up of 6 months, using the survival analysis cohort (Fig 1 : n = 1,701). Patients who underwent CGP testing before initiation of first-line systemic therapy were excluded to avoid potential bias, as CGP testing in Japan typically follows standard treatment. 13 - 15 , 29 Survival analyses evaluated associations between genomic alterations and rwOS across histologic subtypes and treatment groups (Data Supplement, Table S2). Kaplan-Meier curves and Cox proportional hazards models adjusted for age and ECOG performance status were used. For full details, see the Data Supplement, Methods. This study was approved by the Ethics Committee of Eisai Co, Ltd (REP-2024-0970-005-E) and the C-CAT Review Board (CDU2022-020E04). All participants provided written consent for their data to be used for research purposes.