Ovarian carcinosarcoma (OCS) is an uncommon form of ovarian cancer, accounting for approximately 3% of diagnoses, and is distinguished by the presence of both carcinomatous and sarcomatous malignant cell populations (1–3). This biphasic histology led to the hypothesis that OCS may represent collisions of two separately originating tumors; however, the consensus has shifted over the last decade to recognize OCS as metaplastic carcinomas, with the sarcomatous population formed through complete epithelial-to-mesenchymal transition (1, 4). The unique history of OCS has resulted in its exclusion from many pan-histotype studies of ovarian carcinoma, leading to a paucity of research on OCS when compared to other uncommon histotypes (2).

Several studies have examined retrospective cohorts of OCS cases to identify factors associated with patient outcomes (5–11). These studies report a median survival time of 12-24 months across the broader OCS patient population. Earlier FIGO stage at diagnosis and achievement of complete macroscopic resection are both associated with more favorable prognosis, but recurrence and mortality rates appear high even in patients diagnosed with early-stage disease (11–13).

As most OCS cases have carcinomatous components of high grade serous type, some have conceptualized OCS as a rare variant of high grade serous ovarian carcinoma (HGSOC), the most common ovarian cancer histotype (2). However, a significant proportion have carcinomatous components of endometrioid type (3, 11), and limited comparisons of OCS and HGSOC have suggested significant differences in the behavior of these two histotypes (11, 13). Compared to HGSOC, OCS demonstrates greater levels of intrinsic chemoresistance (objective response rate between 30-60%) and is associated with an overall poorer prognosis (11, 14).

While efforts at characterizing the clinical behavior of OCS have improved our understanding of prognostic factors within OCS patients, and limited comparisons have been made against HGSOC (11, 15), there has been little comparison of OCS versus other ovarian carcinoma histotypes. Here, we compare OCS against all major epithelial ovarian carcinoma histotypes using two independent cohorts to improve our understanding of the clinical behavior of these uncommon tumors.

OCS is now recognized as a histotype of ovarian carcinoma, but has received relatively little research attention to date (2). Limited comparisons have been made against HGSOC (11, 13), the most common histotype, but there is a paucity of data comparing these unusual tumors against the spectrum of major ovarian cancer histotypes. Here, we utilize two independent cohorts of ovarian cancer patients to comprehensively characterize the clinical behavior of OCS.

Our findings highlight several distinct features of OCS compared to other ovarian carcinoma histotypes. Firstly, OCS presents in women at an older age compared to other histotypes: the median age at diagnosis in OCS was 69 years in the Scottish cohort and 67 years in the SEER cohort, and this was statistically significantly older than all other histotypes across both cohorts. Within HGSOC, the other histotype with a median diagnosis age of over 60 years, copy number gain of CCNE1 has been associated with older age at diagnosis (17). OCS have recently been reported to commonly demonstrate CCNE1 gain (4), and their older age at diagnosis may be linked to the frequency of this defect; however, direct comparison of CCNE1 status and age of OCS diagnosis has not been reported.

OCS also appears to have a distinct stage distribution; the majority of OCS present at FIGO stage III-IV – unlike MOC, EnOC and CCOC – but around 25% are FIGO stage I/II at diagnosis, and this is significantly more than in HGSOC. As OCS frequently present with advanced stage disease, many patients undergo neoadjuvant chemotherapy prior to cytoreductive surgery; this approach is widely considered safe and effective for HGSOC (27, 28), but neoadjuvant chemotherapy versus primary debulking surgery has not been specifically compared for OCS. Given reports of higher levels of intrinsic chemoresistance in OCS (objective response rate 25-60%) (11–13), neoadjuvant chemotherapy may feasibly represent a less effective management strategy. Indeed, neoadjuvant chemotherapy is not the preferred approach for other histotypes with high levels of intrinsic chemoresistance (29). However, challenges in identifying OCS on diagnostic biopsies – where the sarcomatous component may not be sampled, leading to a diagnosis of more common carcinoma hisotypes – may interfere with the ability to tailor early first-line management decisions for OCS patients.

Univariable analysis identified OCS as the histotype associated with poorest survival across both cohorts. Within the Scottish cohort, multivariable analysis demonstrated that this was independent of other prognostic factors for comparisons of OCS against all other histotypes, with the exception of CCOC. The poorer outcome of OCS compared to HGSOC, EnOC, LGSOC and MOC was confirmed in the SEER cohort; this cohort also identified OCS as having significantly shorter survival than CCOC. The difference in comparisons with CCOC between cohorts may be underpinned by greater statistical power in the SEER cohort, though less detailed clinical annotation prevented inclusion of residual disease status in the SEER cohort model, likely contributing to this discrepancy. Together, these data suggest that the overall OCS population represents the highest risk histotype across ovarian carcinomas.

In an analysis specifically of earlier stage patients (FIGO I-II) in the Scottish cohort, OCS was associated with markedly shorter survival than all other histotypes, including CCOC. These findings were replicated when investigating SEER cohort patients with localized disease, though the comparison with HGSOC was not statistically significant (mHR for localized HGSOC vs localized OCS 0.80, 95% CI 0.58-1.11). This discrepancy may be due to the difference in staging between cohort; SEER localized stage equates to the very earliest FIGO stages (IA, IB and stage I not otherwise specified). These data have important implications for decisions around omission of chemotherapy for early-stage disease. Many ovarian cancer cases diagnosed at the earliest stages do not require chemotherapy (28); however, the aggressive nature of early-stage OCS suggests that chemotherapy omission may not be advisable for this group. Similarly, fertility-sparing surgery may not be feasible in this context, though most OCS patients present after reproductive age. In late-stage disease (FIGO III-IV), the Scottish cohort demonstrated that OCS was associated with significantly shorter survival compared to HGSOC, LGSOC and EnOC, but was not associated with significantly poorer outcome than CCOC or MOC. These findings were confirmed in the SEER cohort, where distant stage CCOC demonstrated similar survival to distant stage OCS, and late-stage MOC demonstrated the worst survival of all histotypes in this context. Together, these stage-specific analyses highlight OCS as highly aggressive even when diagnosed at early-stage, while in the context of late-stage disease, OCS, CCOC and MOC represent the histotypes with poorest survival. This is consistent with reports highlighting CCOC and MOC as highly chemoresistant malignancies with exceptionally poor prognosis when diagnosed at advanced stage (2). While OCS is most commonly diagnosed at advanced stage, late-stage diagnosis of CCOC and MOC is relatively uncommon, underscoring treatment of late-stage OCS as a major clinical challenge.

Major strengths of this study include the detailed clinical annotation available for the Scottish cohort, extensive follow-up time and the utilization of multivariable analysis to assess associations of histotype with outcome independent of other prognostic factors. The use of two independent cohorts from distinct geographical locations is also a notable strength; SEER is a pan-cancer database curated across a large number of centres in the US, while the Edinburgh Ovarian Cancer Database is a disease-specific resource curated centrally at a single site. A limitation of the present study is that all cases did not undergo centralized pathology review, though over 75% of the Scottish cohort has either undergone pathology review as part of recent molecular profiling studies or represented contemporary diagnoses (2010 onwards), limiting the potential for histotype misclassification. As it was not possible to perform pathology review of any cases in the SEER cohort, we utilized only recent diagnoses from the SEER database (2010-2019) to minimize potential histotype misclassification. Though we were able to include a relatively large number of cases with uncommon histotypes - and the number of these cases exceeded that in many reported cohorts of these less common diagnoses - power was still limited for some analyses. In particular, the number of advanced stage MOC and early-stage OCS or LGSOC cases was modest, though the large effect sizes detected between analyses of these groups bolstered power.

Our findings highlight the urgent need for additional treatment options for OCS patients. Molecular profiling studies have the potential to identify targeted approaches that may improve OCS patient survival; however, relatively few OCS samples have undergone genomic, transcriptomic or other molecular characterization to date. Limited available data suggest a paucity of targetable oncogenic driver mutations from the genomes of OCS tumors (4, 30), with TP53 mutation representing one of the few recurrent molecular events. A proportion of OCS demonstrate genomic evidence of homologous recombination repair deficiency (31), and these cases may be expected to benefit from poly(ADP-ribose) polymerase (PARP) inhibitors. Case reports of OCS patients deriving clinical benefit from PARP inhibition are available in the literature, but this evidence base is extremely limited (32, 33). The frequency of germline or somatic BRCA1/2 mutation is poorly characterized in OCS; case reports of BRCA1/2-mutant OCS are available, but current data from OCS cohorts suggest the frequency is low (0/12 in (4) and 0/13 in (34)). There is also a lack of data quantifying the extent of homologous recombination deficiency with robustly established techniques due to a lack of whole genome sequencing (35).

Recent data suggest that the sarcomatous compartment of OCS is less well engaged by the host anti-tumour immune response compared to the carcinomatous component (4); immunotherapeutic drugs may therefore represent agents worthy of investigation in the hope of reinvigorating the anti-tumour immune response. In particular, immune checkpoint inhibitors targeting PD1, PDL1 and CTLA4 are of interest. Case reports of responses to such inhibitors in OCS patients provide anecdotal evidence of their potential utility in the wider population (36, 37). However, as with other candidate targeted approaches, there is a marked absence of trial data at any phase. Overexpression of HER2 and VEGF in some OCS has suggested trastuzumab and anti-angiogenics as further potential treatment strategies for investigation, alongside inhibitors of mTOR (38). Recently established preclinical models of OCS have identified eribulin as a candidate therapeutic strategy targeting epithelial-to-mesenchymal transition in OCS (39), and we eagerly await the results from initial clinical evaluations of this strategy (NCT05619913). The relative rarity of OCS is likely to hinder progress of histotype-specific trials for this tumour type; international collaborative efforts have led to successful disease-specific trials in other uncommon ovarian cancer types (40), and it is likely that similar international collaboration will be required to drive advances in the standard of care for OCS patients.

As with other uncommon ovarian cancer histotypes, a multidisciplinary approach is key for determining optimal management for individuals with OCS.

Together, our findings identify OCS as an exceptionally aggressive histotype of ovarian carcinoma. OCS patients represent an older patient group that are frequently diagnosed at advanced stage. Despite its aggressive behavior, OCS is a relatively under-researched tumour type, hindering progress toward new treatment options which are urgently required to improve outcomes.

The Scottish dataset presented in this article is not readily available because sharing of patient data is only possible within the constraints of our local ethics framework, which means line-by-line data cannot be shared without seeking an associated ethical approval. Such approval can be sought via contact with the corresponding author (robb.hollis@ed.ac.uk). Data from the SEER database can be accessed through the SEER program website.

Institutional review board approval for the Scottish cohort was received from the South East Scotland Cancer Information Research Governance Committee (Caldicott guardian reference CG/DF/E164, study reference CIR21087). The studies were conducted in accordance with the local legislation and institutional requirements. The need for informed consent for every participant was waived due to the retrospective nature of the study.

IM: Data curation, Formal analysis, Visualization, Methodology, Writing – original draft. JMP: Data curation, Writing – review and editing. EB: Data curation, Writing – review and editing. NG: Resources, Writing – review and editing. KCC: Resources, Writing – review and editing. CSH: Resources, Writing – review and editing. RLH: Conceptualization, Data curation, Formal analysis, Methodology, Visualization, Writing – original draft.