Translating biological insights into improved management of endometrial cancer

Abstract

Endometrial cancer (EC) is the most common gynaecological cancer among women in high-income countries, with both the incidence and mortality continuing to increase. The complexity of the management of patients with EC has evolved with greater comprehension of the underlying biology and heterogeneity of this disease. With a growing number of novel therapeutic agents available, emerging treatment regimens seem to have the potential to help address the concerning trends in EC-related mortality. In this Review, we describe the epidemiology, histopathology and molecular classification of EC as well as the role of the new (2023) International Federation of Gynecologists and Obstetricians (FIGO) staging model. Furthermore, we provide an overview of disease management in the first-line and recurrent disease settings. With increasing use of molecular profiling and updates in treatment paradigms, we also summarize new developments in this rapidly changing treatment landscape.

Introduction

Globally, endometrial cancer (EC) is the 6^th most common malignancy affecting women with an annual incidence of approximately 417,000 cases, disproportionally affecting women in high-income countries^1,2 (FIG. 1). Current epidemiological trends demonstrate a rapid rise in EC incidence^1,3,4. For example, in the USA, the incidence of EC has risen approximately 1% per year since the mid-2000s among women ≥50 years of age and 1–2% per year since the mid-1990s in women of reproductive age^1,4,5. Unfortunately, survival outcomes of women with EC have not improved over the past 4 decades, and the most recent trends indicate a rise in mortality rates of 1.8% per year since 2010^1,4,6. Despite the growing prevalence and mortality rate of EC, funding for research into this malignancy continues to be disproportionately low compared with other malignancies and might be a contributing factor to the historical lack of meaningful progress in improving oncological outcomes^7,8.

Figure 1. Incidence and mortality of cancers of the corpus uteri.

a. Age-standardized global incidence of cancers of the corpus uteri. b. Global mortality of cancers of the corpus uteri. Adapted with permission from ref 2.

Along with regional variations, disparities in incidence and mortality exist among women of different socioeconomic status and ethnicity. In a systematic review, the authors demonstrated an association between lower socioeconomic status and inferior oncological outcomes, which might be attributable to poor access to quality health care⁹. Observations regarding differences in EC across different ethnicities are most visible in data from epidemiological studies involving populations in the USA. Both the incidence and mortality of EC has risen among all racial and/or ethnic groups, nonetheless, this malignancy has disproportionally affected Black women, with a rise in both overall incidence and that of aggressive histological subtypes as well as higher mortality, even when adjusting for stage and histology^6,10. Delayed access to care might contribute to the inferior oncological outcomes of Black women, but these factors do not fully explain the disparities in prevalence and survival outcomes relative to other racial groups^6,11–17.

In the general population, women 60–80 years of age have the highest risk of EC with a cumulative risk of 1–3% by 75 years of age^1,4. Historically, EC has been classified into two broad histological categories: endometrioid (type I) and non-endometrioid (type II) carcinoma¹⁸. These two groups differ in clinicopathological profiles and risk factors^18,19. Endometrioid carcinomas typically develop on a background of unopposed high levels of oestrogen from endogenous (such as obesity) or exogenous sources (such as tamoxifen) compared with non-endometrioid carcinomas (such as serous carcinomas) that are not typically oestrogen-driven but have more-aggressive features and an poorer prognosis^18,19. Among the established risk factors (Supplementary Table 1), obesity is the most prevalent factor associated with EC development, contributing to up to 57% of all cases in high-incidence nations (such as the USA)^20–22. Interestingly, obesity has been associated with an increased risk of developing both type I and type II EC in several studies, which raises concerns given the ongoing obesity epidemic in most economically developed countries^23,24.

Women with Lynch syndrome, which accounts for 3% of all ECs and 9% of those diagnosed in women <50 years of age, have the highest risk of developing EC^25,26. Lynch syndrome is an autosomal dominant disorder characterized by germline mutations in DNA mismatch repair genes (MLH1, MSH2, MSH6, PMS2) or EPCAM. This hereditary tumour syndrome substantially increases the lifetime risk of multiple cancers, most notably EC (~50%) and colorectal cancers (~60%)²⁷. Given this increased lifetime risk of multiple malignancies, patients with Lynch syndrome should undergo regular screening as well as use prevention strategies according to the most up-to-date clinical practice guidelines, with potentially affected family members offered cascade genetic testing²⁸.

In this Review, we describe the clinicopathologic features, molecular classification, and new staging model of EC. Additionally, an overview of EC management in the first-line and recurrent disease setting will be presented. Lastly, updates on the integration of molecular profiling in treatment paradigms and promising new therapeutic agents in the rapidly changing EC treatment landscape will be discussed.

Of note, the definition of ‘woman/women’ in this review will represent individuals who were born with a uterus and the authors acknowledge that not all members of this group may self-identify as a woman.

Disease biology and risk stratification

Traditional classification and histopathological features

The framework for conceptualization of EC into two broad categories of endometrioid (type I) and non-endometrioid (type II) tumours was based on differing clinicopathological features proposed by Jan V. Bokhman in 1983¹⁸. Encompassing the majority of ECs, type I EC consists of tumours of a less-aggressive (grade 1–2) endometrioid histology with a more favourable prognosis, typically occurring in states of relative oestrogen excess to progesterone (such as obesity)¹⁸. Atypical endometrial hyperplasia, a precursor lesion of endometrioid carcinoma, develops in an environment of unopposed oestrogenic signalling through a complex process of subclonal evolution featuring an increased ratio of endometrial glandular tissue to stroma^29,30. Endometrioid carcinomas are graded on a scale of 1–3 based on proportions of glandular and solid-tumour components on histological examination according to the International Federation of Gynecology and Obstetrics (FIGO) system³¹.

In contrast to type I tumours, type II EC develops via hormone-independent processes, has less well-defined precursor lesions, and is associated with a poorer prognosis¹⁸. Non-endometrioid tumours encompass several histological subtypes such as serous and undifferentiated carcinoma, carcinosarcoma, undifferentiated carcinoma, and other high-grade carcinomas^18,30,32(Supplementary Table 2). Serous EC is the most common non-endometrioid histological subtype (accounting for ~10% of all ECs) and commonly arises on the surface of an endometrial polyp in the background of an atrophic endometrium³³. For these cancers, TP53 mutations are postulated to be the carcinogenesis-initiating event^34,35. Despite apparent uterine-confined disease, serous ECs develop aggressively with a propensity for frequent occult metastases and account for 40% of EC-related deaths^33,36,37. Clear-cell EC is rare (<5% of all ECs) and thought to be chemotherapy-resistant relative to other ECs, and is associated with inferior oncological outcomes^38–40. Carcinosarcoma, previously known as mixed mullerian tumour, is a rare biphasic carcinoma (<5% of all ECs) with both epithelial and sarcomatous elements^30,41. Carcinosarcoma is further characterized based on whether the sarcomatous element is native to the uterine tissue (homologous) or not (heterologous). Although previously categorized as a sarcoma, carcinosarcomas are thought to originate from a single precursor lesion with the sarcomatous element undergoing metaplastic differentiation with patterns of recurrence and metastases that mirror those of endometrial carcinomas^42–44. Carcinosarcomas are typically associated with a worse prognosis than endometrioid, serous or clear-cell ECs^41,45. Undifferentiated/dedifferentiated carcinomas are aggressive and frequently underrecognized high-grade carcinomas that are poorly understood relative to other ECs^46–48. Undifferentiated carcinoma is characterized as monotonous cells arranged in sheets with frequent loss of PAX8 and oestrogen and/or progesterone receptor expression^46,47. With co-existing areas of grade 1–2 endometrioid carcinoma, undifferentiated carcinomas are termed dedifferentiated carcinoma^46,47. Rarer histological subtypes of EC include mesonephric-like adenocarcinoma and squamous cell carcinoma. Thus, despite providing some insight into the natural history and pathogenesis of EC, the type I/II classification system fails to provide a nuanced view of the heterogeneity in tumour biology and clinicopathological outcomes of patients with EC. As a result, the classification of EC has evolved over time to include comprehensive profiling of the molecular features of each histological subtype^49,50 (Table 1, Supplementary Figure 1).

Table 1 |.

EC histological subtypes and their morphological and molecular characteristics^49,50

Histological subtype	Morphological features	Molecular features	Potentially significant molecular traits
Endometrioid adenocarcinoma	Glandular or villoglandular architecture, while solid growth within the tumour is the basis for FIGO grading (grade 1: ≤5%; grade 2: 6–50%; and grade 3: >50%). These cells have a columnar appearance with pseudostratified nuclei. Nuclear atypia is mild to moderate. Exhibit positivity for ER and PR that is stronger and diffuse in low-grade tumours (FIGO 1–2) than high-grade tumours. Rare low-grade tumours have been reported to exhibit aberrant p53 staining, whereas a larger proportion of high-grade tumours are likely to have aberrant p53.	Low-grade POLE mut: 6.2% MMRd: 24.7% p53 abn: 4.7% NSMP: 63.5%	Mutations in PTEN, PI3KCA, ARID1A, PI3KR1, KRAS, TP53, CTNNB1, and POLE; loss of MMR proteins; L1CAM overexpression, ER expression, PR expression
		High-grade POLE mut: 12.1% MMRd: 39.7% p53 abn: 21.3% NSMP: 28%
Serous carcinoma	Papillary or glandular architecture in most cases, although a solid growth pattern can also be seen. Cytological features are typically high-grade and easily identified, such as severe nuclear pleomorphisms with easily identifiable mitotic activity. Aberrant p53 staining (such as the diffuse-positive or completely negative ‘null’ phenotype) is seen in serous carcinomas, along with diffuse positivity for p16. ER and PR expression is variable.	POLE mut: 0% MMRd: 0% p53 abn: 100%* NSMP: 0%	Mutations in TP53, PIK3CA, PPP2R1A, FBXW7, ARID1A and ERBB2; ERBB2 amplifications and/or HER2 overexpression
Clear-cell carcinoma	Several major architectural patterns (alone or admixed), including papillary, tubulocystic and solid. Cancer cells typically have a cuboidal morphology (and also can be polygonal or flat) and can be found with ‘hobnail’ extension into the lumen. The cytoplasm might be clear; however, eosinophilic cytoplasm can also be seen. Extent of nuclear atypia can be variable, albeit focal atypia is common. Level of mitotic activity can also be variable. ER and PR expression are typically lacking, whereas a significant portion of samples will have aberrant p53 staining. Specific markers include HNF1β, napsin A and racemase (AMACR/p504s).	POLE mut: 3.8% MMRd: 9.8% p53 abn: 42.5% NSMP: 40.9%	Mutations in TP53, PIK3CA, PIK3R1, PPP2R1A and ARID1A
Carcinosarcoma	A mixture of malignant epithelial and mesenchymal elements. The carcinoma component usually is comprised of serous or endometrioid cells, although other elements can be seen and are of their typical histology. The malignant mesenchyme usually is similar to that of nonspecific high-grade sarcoma, though heterologous elements can be seen, including (but not limited to) rhabdomyosarcoma or chondrosarcoma. Immunohistochemical findings are dependent on the sub-elements found within the tumour. The majority of tumours have underlying TP53 mutations, and thus aberrant p53 immunohistochemistry is typically seen.	POLE mut: 5.3% MMRd: 7.3% p53 abn: 73.9% NSMP: 13.5%	Mutations in TP53, FBXW7, PIK3CA, PPP2R1A and PTEN
Mesonephric-like carcinoma	Appears as small, simple glands with eosinophilic secretions within the lumen. Architectural patterns can also include papillary, retiform, solid or spindled forms. The nuclei typically have moderate levels of atypia, and vesicular chromatin reminiscent of papillary thyroid carcinoma. ER and PR expression is typically negative (a key feature from morphologic mimics of low-grade endometrioid cells, which are typically diffusely positive). GATA3 and TTF1 are usually present and can have an inverse staining pattern (where one marker is positive the other negative, and vice versa). Secondary markers include calretinin and luminal CD10.	POLE mut: 0% MMRd: 0% p53 abn: 0% NSMP: 100%	Mutations in KRAS/NRAS, ARID1A, PIK3CA and CTNNB1
Undifferentiated and dedifferentiated carcinoma	Morphologically similar, except that dedifferentiated carcinomas are diagnosed in the setting of adjacent differentiated carcinoma components (typically a low-grade endometrioid carcinoma). Cancer cells are small-to-intermediate in size and uniform, arranged in sheets of discohesive cells. Markers of epithelial differentiation are positive but usually in scattered cells, with CK8/18 and EMA more likely than other epithelial markers to be positive. ER, PR and PAX8 are usually negative. Neuroendocrine markers chromogranin and synaptophysin can be present and should not be mistaken for a neuroendocrine carcinoma, with a cutoff <10% used for undifferentiated/dedifferentiated carcinoma. A third of cases might have SMARCA4 loss	POLE mut: 12.4% MMRd: 44% p53 abn: 61.1% NSMP: 0%	Mutations in ARID1A and PTEN, loss of MMR proteins

Some investigators recommend endometrioid carcinoma with serous morphology with POLE mutations or MMRd be classified as serous-like high-grade endometrioid carcinoma rather than serous carcinoma⁴⁹. abn, abnormal; EC, endometrial cancer; ER, oestrogen receptor; MMRd, DNA mismatch repair-deficient; MMR, mismatch repair; mut, mutation; NSMP, no specific molecular profile; PR, progesterone receptor; p53abn, p53 abnormal.

^aRarely, immunohistochemically normal p53 expression with TP53 mutations and/or amplifications of wildtype TP53 might be detected.

Molecular characterization

The Cancer Genome Atlas (TCGA) was a landmark cancer genomics programme led by the National Cancer Institute and National Human Genome Research Institute in the USA that characterized multiple tumour types at the genomic, epigenomic, transcriptomic and proteomic level. Analyzing 373 endometrioid, serous and mixed histological subtypes, the TCGA categorized EC into four molecular subtypes: POLE ultramutated, microsatellite instability hypermutated, copy-number low, and copy-number high⁵⁰.

The POLE-ultramutated subgroup comprises approximately 8% of all ECs and is defined by the presence of functional pathogenic variants in exons 9–14 of the exonuclease domain of POLE, which encodes the catalytic subunit of DNA polymerase ε (Polε) that is involved in DNA replication and repair^50,51. Although the smallest subgroup of EC, POLE-ultramutated tumours have the highest average tumour mutational burden (TMB), ~100-fold greater compared to the average TMB for all EC subtypes combined⁵⁰. Given the magnitude of the TMB in POLE-ultramutated tumours, ascertaining the contributions of specific somatic mutations to carcinogenesis (passenger versus driver mutations) can be challenging⁵⁰. Interestingly, POLE-ultramutated tumours are strongly associated with a grade 3 endometrioid histological appearance but can, uncommonly, have a non-endometrioid histology^51,52. Among the four TCGA subtypes, POLE-ultramutated tumours are associated with the most favourable prognosis and longest median survival duration (e.g. 60-month progression-free survival rate of 100% reported in the TCGA); this might be attributed to the strong association with high levels of immune cell infiltration into the tumour as well as the presence of tertiary lymphoid structures^50,53,54.

The microsatellite instability-hypermutated subgroup accounts for approximately 30% of all ECs and is defined by high levels of microsatellite instability (MSI-H), which can be observed as shifting lengths of repeated DNA sequences called microsatellites⁵⁰. The presence of MSI-H is generally a consequence of mismatch repair deficiency (MMRd)^50,55,56. The majority of MMRd/MSI-H ECs are attributable to MLH1 promoter methylation (~70%) with the remaining cases arising from germline predisposition (Lynch syndrome ~10%) or somatic mutations (~20%) in MLH1, MSH2, MSH6, PMS2, or EPCAM⁵⁵. MSI-H tumours are characterized by a high TMB (~10-fold greater than the average EC TMB) and high levels of immune infiltration⁵⁰; therefore, similar to POLE-ultramutated tumours, distinguishing between passenger and driver mutations can be challenging. These tumours are associated with an intermediate prognosis relative to the other subtypes⁵⁰. In comparison with Lynch syndrome-associated MSI-H ECs, those attributed to MLH1 promoter methylation typically have a less-favourable prognosis and inferior responses to immune-checkpoint inhibitors (ICIs)^55,57,58.

The remaining two molecular subtypes are based on the extent of somatic copy-number alterations⁵⁰. Copy-number high EC primarily comprises serous carcinomas including approximately 25% of high-grade endometrioid tumours with TP53 mutations as the defining feature, frequent FBXW7 and PPP2R1A mutations, and increased levels of cell-cycle dysregulation (owing to alterations in CCNE1, PIK3CA, MYC and/or CDKN2A)⁵⁰. This molecular subtype portends a poor prognosis⁵⁰. The majority of ECs (38%) profiled in TCGA are of the copy-number low subtype and are mostly low-grade endometrioid tumours associated with a good prognosis⁵⁰. However, the presence of chromosome 1q amplifications, CTNNB1 mutations (~50% tumours), or oestrogen receptor (ER) loss are all associated with worse than expected clinical outcomes despite otherwise low-risk features^50,59,60.

Integration of molecular profiling into clinical practice

Despite the clear differences in prognosis and clinical characteristics of the four molecular subgroups of EC, integration of the complex genomic analyses performed in the TCGA into everyday clinical practice is currently prohibited by costs^61,62. However, two groups of investigators have developed and demonstrated the performance of testing algorithms capable of analogous (but not identical) molecular subtyping corresponding to the four TCGA subgroups^61,62. In contrast to attempting to replicate select TCGA analyses using fresh-frozen tissue, these investigators used formalin-fixed, paraffin-embedded (FFPE)-based methods that resulted in logistically practical molecular classification with the potential for integration into clinical practice^61–64. One group, the TransPORTEC investigators, used p53 immunohistochemistry (IHC) combined with an RNA sequencing panel comprising MSI-related genes and selected somatic hotspot mutations and with POLE proofreading mutation testing using Sanger sequencing to analyse 116 FFPE samples from women with high-risk EC⁶¹. These tumours were categorized into four subgroups with distinct prognostic significance: POLE mutant, MSI-H, no specific molecular profile (NSMP), and p53 mutant⁶¹. Similarly, the ProMisE investigators were able to molecularly subcategorize EC based on MMR protein IHC, p53 IHC and analysis of POLE mutations in 152 FFPE samples from women with EC⁶². The ProMisE investigators used a stepwise algorithm to classify EC into four molecular subgroups: POLE, MMRd, p53 abnormal, and p53 wildtype, which are analogous to the TCGA subtypes⁶². Like the TransPORTEC investigators, the ProMisE investigators demonstrated that these molecular subtypes are associated with differing disease-free and overall survival (OS) outcomes^61,62. Both the TransPORTEC and ProMisE algorithms have been validated as accurate EC classification tools using large retrospective datasets^63,64 (FIG. 2).

Figure 2. Suggested molecular profiling workflow in patients with endometrial cancer.

This decision matrix provides a general overview of the molecular profiling workflows required to identify distinct molecular subtypes of endometrial cancer. Specific details will differ based on the clinical situation and resource availability and/or prioritization for each institution. ER, oestrogen receptor;FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; MMR, mismatch repair; MMRd, mismatch repair deficient; MMRp, mismatch repair proficient; MSI, microsatellite instability; MSI-H, microsatellite instability-high; MSS, microsatellite stable; NSMP, no specific molecular profile; POLE, DNA polymerase epsilon; PR, progesterone receptor.

Data provided by the TCGA, TransPORTEC, and ProMisE studies indicate that patients with POLE-mutant tumours have a better prognosis relative to other subtypes, irrespective of other uterine histological features or characteristics^50,63,64. Despite the potential application in triaging patients who are most appropriate for adjuvant therapy, the identification of POLE-mutant tumours in clinical practice continues to pose logistical challenges. Currently, 11 pathogenic missense mutations within exons 9, 11, 13, and 14 are recognized to define a clinically significant subgroup of POLE-mutant ECs that are associated with an excellent clinical outcome⁶⁵. However, in the absence of IHC markers, identification of these POLE-mutant tumours requires DNA sequencing, which can be cost-prohibitive and overly time-consuming for universal use, even in resource-rich settings⁶⁶. To address this diagnostic barrier, a multicentre collaboration developed and evaluated QPOLE, a POLE-hotspot test, designed to utilize three quantitative PCR assays to identify both common and rare pathogenic POLE mutations⁶⁶. In an analysis of 282 FFPE specimens from patients with EC (including 99 with POLE-mutant disease), QPOLE demonstrated an overall accuracy of 98.6%, sensitivity of 95.2% and specificity of 100%⁶⁶. This test has the additional advantage that POLE status can be determined within 4–6 hours of DNA isolation and thus has the promise for implementation in routine clinical practice⁶⁶. The use of artificial intelligence tools to improve the efficiency of tasks such as morphological–molecular correlation and risk-stratification, and thus improve the cost-effectiveness of molecular profiling is an area of considerable research interest that should be examined in future studies⁶⁷.

Integration of histopathological and molecular features into FIGO staging

Historically, FIGO staging of EC has been solely based on the anatomical distribution of disease, although updates to the FIGO staging system in 2023 have incorporated histopathological and molecular features with the goal of more-refined staging^68,69 (FIG. 3, Supplementary Table 3). Most of the major (albeit also the most controversial) changes in the 2023 FIGO staging system pertain to the classification of stage I/II disease. Along with depth of myometrial invasion and cervical stromal invasion, other stage modifiers for early stage disease include histological subtype, grade and extent of lymphovascular space invasion⁶⁹. Additionally, histology is dichotomized into non-aggressive and aggressive subtypes⁶⁹. Although not mandatory for FIGO staging, comprehensive molecular profiling is encouraged for all ECs and is denoted by ‘m’ followed by a subscript of the molecular subtype (for example, POLEmut, MMRd, NSMP, or p53abn) after stage designation (for example stage IB MMRd tumour = stage IBm_MMRd)⁶⁹. Irrespective of other histopathological features or anatomical spread for stage I/II disease, POLE-mutant tumours are downstaged to stage IA (stage IAm_POLEmut) and p53 abnormal tumours are upstaged to stage IIC (stage IICm_p53abn)⁶⁹ based on the prognostic implications of these characteristics.

Figure 3. Endometrial cancer staging criteria.

Endometrial cancer is staged according to the International Federation of Gynecology and Obstetrics (FIGO) criteria (2009), including an update published in 2023, stipulating that molecular testing for relevant pathological characteristics (such as POLE mutations, mismatch repair deficiency (MMRd), and/or p53 status) should be performed when feasible. a, Stage I disease (limited to the corpus uteri). b, Stage II disease (tumours invading the cervical stroma but remaining confined to the uterus). Stage I–II disease with POLE mutations or p53 aberrations should undergo modifications for the final FIGO stage (e.g. stage IA for POLE mutated tumours and stage IIC for p53 aberrant tumours), despite initial stage based on anatomopathologic features. c, Stage III (with local and/or regional tumour spread) and d, Stage IV (with tumour invasion of the bladder and/or bowel mucosa and/or distant metastases). Stage III–IV disease is not modifiable by molecular classification; however, the molecular classification should be recorded if known⁶⁹. Aggressive histological types include high-grade endometrioid endometrial carcinomas (grade 3), serous, clear cell, undifferentiated, mixed, mesonephric-like, gastrointestinal mucinous type carcinomas, and carcinosarcomas. DMI, deep myometrial invasion; LVSI, lymphovascular space invasion. Adapted from ref. 166, Springer Nature Limited¹⁶⁶.

Globally, responses to the FIGO 2023 staging system have thus far been mixed^70–73. More precise prognostication is the greatest advantage of integrating established prognostic pathological and molecular characteristics into the FIGO staging criteria^72,74,75. Multiple studies have retrospectively applied the FIGO 2023 staging to large EC cohorts and demonstrated improved predictive validity for oncological outcomes whe these criteria are used^74,76–78. These retrospective studies demonstrate that, compared to FIGO 2009 staging, the 2023 criteria enable more-precise risk stratification and potentially greater personalization of treatment for patients, especially given that EC is not a homogenous entity.

Incorporation of a larger number of parameters and granularity into FIGO staging might increase prognostic capabilities and help triage care, although full implementation of this system also substantially increases the complexity of staging compared to the 2009 predecessor (increase from 9 to 19 subcategories, excluding molecular subclassification) and might not be as intuitive given the use of non-anatomical variables to determine stage^68–70. Other disadvantages to the FIGO 2023 staging model include the use of highly subjective histopathological variables (such as quantification of lymphovascular space invasion), which will probably lead to substantial levels of interobserver variability (even among experienced pathologists)⁷⁰. A lack of consistent reproducibility might hinder patient care given that such differences in stage can result in very different treatment recommendations. Another layer of complexity includes the inclusion of molecular classifiers, a feat that is challenging to implement in resource-poor settings. Given this controversy, appraisals and discussions regarding implementation of the new FIGO 2023 staging system are ongoing but should strive to maintain the advantages of greater prognostic precision and personalization of treatment while improving practicality for patients, clinicians, and health-care providers^70,71,75. For the remainder of this Review, descriptions of staging will refer to the FIGO 2009 staging system.

Management of patients with primary resectable disease

Surgical management

Along with adequate abdominopelvic examination and the standard preoperative tests, preoperative evaluations should include radiographic imaging to rule out lung metastases. Pelvic ultrasonography might also be ordered to determine uterine size for surgical planning purposes, if this is not discernable on physical examination or if no prior imaging is available. CT or PET–CT imaging is often performed if concerning features associated with metastases and/or high-risk EC are found on examination of preoperative biopsy samples. MRI might be informative for patients with abnormalities on physical examination such as gross cervical, vaginal or parametrial extension. CT and PET–CT are both known to have poor predictive ability for the detection of extrauterine disease and routine imaging rarely alters the approach to surgical management^79,80.

Surgical staging followed by total extra-fascial hysterectomy, bilateral salpingo-oophorectomy and lymph node biopsy sampling is the gold standard surgical approach. For clinically confined EC, minimally invasive surgery (traditional laparoscopy or robotic-assisted laparoscopy) is recommended given the improved perioperative morbidity and quality of life indices and equivalent oncological outcomes when compared to laparotomy^81–85. The optimal extent of lymph node evaluation is controversial, although sentinel lymph node (SLN) mapping with ultrastaging has gained popularity. SLN mapping, which involves sampling fewer, higher-yield lymph nodes to detect metastatic disease (particularly low-volume disease) while minimizing the risks associated with more-extensive lymph node dissection is an attractive alternative to full lymphadenectomy. SLN mapping has demonstrated excellent accuracy for the detection of lymph node metastases both in women with low-risk and high-risk EC^86–89. Characterized as serial sectioning of the lymph node with subsequent staining (hematoxylin and eosin with or without cytokeratin), ultrastaging can be readily performed on fewer, high-yield lymph nodes (e.g. sentinel lymph nodes) to detect micro-metastases and isolated cancer cells with greater sensitivity than traditional nodal pathologic examination (e.g. single longitudinal cross-section with hematoxylin and eosin)⁹⁰. In contrast for cases with full lymphadenectomies, ultrastaging cannot be practically employed given its resource-intensive nature. Micrometastases detected using this approach should be denoted as node positivity (such as stage IIIC disease) whereas the prognostic significance of isolated cancer cells is unclear⁹¹. The use of adjuvant therapy in the presence of isolated cancer cells should be guided by uterine risk factors⁹¹.

EC with suspected or gross cervical involvement can be managed using several methods, although the choice of approach should be individualized based on the specific clinical scenario. Total extra-fascial hysterectomy followed by postoperative radiotherapy is the most widely used strategy. Alternatively, radical hysterectomy might be performed if obtaining negative resection margins is a priority or if distinguishing between stage II EC versus primary cervical cancer is challenging; use of postoperative radiotherapy in this scenario should be guided by pathological factors. However, this approach has not been shown to improve survival outcomes relative to simple hysterectomy for patients with stage II EC^92,93. A third strategy involves the use of neoadjuvant radiotherapy to shrink the cervical tumour and ‘sterilize’ the parametrial margins followed by interval total extra-fascial hysterectomy at approximately 4–12 weeks post-irradiation⁹⁴. Patients with stage IV EC that is grossly limited to the peritoneal cavity might also be eligible for surgical management if cytoreduction to no gross residual tumour is feasible, otherwise neoadjuvant systemic therapy followed by interval cytoreduction is advisable. Primary cytoreductive surgery should be avoided for patients with a high likelihood of unresectable disease given that gross residual disease is associated with inferior progression-free survival (PFS) and OS outcomes relative to complete cytoreduction⁹⁵.

Adjuvant therapy

Low-risk EC.

Low-risk ECs are defined by the following characteristics: low grade (grade 1 or 2), endometrioid or non-gastrointestinal mucinous-type histology, no lymphovascular invasion, an absence of deep myometrial invasion, and uterine-confined disease. Patients with low-risk EC typically have an excellent prognosis^96,97. Owing to the low risk of recurrence associated with these tumours (<5%) as well as the tendency to recur at the vaginal apex, adjuvant therapy can usually be safely omitted^97–100.

Intermediate-risk EC.

The category of intermediate-risk EC encompasses early stage low-grade endometrial cancer with at least one of the following: lymphovascular invasion, deep myometrial invasion or cervical stromal invasion. Stage I high-grade (grade 3 or 4) endometrioid tumours are also considered intermediate risk if no deep myometrial invasion is present. A high–intermediate risk category has been further defined by the Gynecologic Oncology Group (GOG) and PORTEC. On the basis of the GOG-99 criteria, high–intermediate risk is based on age category (<50, 50 – 69, ≥70) and the presence of 1–3 pathological risk factors (deep myometrial invasion, grade 2–3 endometrioid histology or lymphovascular invasion)¹⁰¹. According to the PORTEC criteria, high–intermediate risk is defined as age >60 years with either deep myometrial invasion or grade 3 histology¹⁰².

Several phase III randomized controlled trials (GOG-99, PORTEC-1 and PORTEC-2) have demonstrated benefit from adjuvant radiotherapy compared to surveillance in patients with high–intermediate risk disease^101–103. Although no significant improvement in OS was demonstrated in these trials, adjuvant radiotherapy reduces the risk of locoregional recurrence with significant reductions in risk after 10–15 years of follow-up monitoring^101–105 (TABLE 2). The risk of recurrence for women with low–intermediate risk EC is approximately 5% but can increase to 30% for women with high–intermediate risk EC who do not receive adjuvant radiotherapy^101–103. Given equivalent disease-free survival outcomes but lower toxicity risks, vaginal brachytherapy (VBT) is preferred over external beam radiotherapy (EBRT) for most patients with stage I high–intermediate risk EC¹⁰³. EBRT might be preferable for those with stage II EC given the risk of nodal recurrence even after surgical staging^39,101,106.

Table 2 |.

Trials testing adjuvant radiotherapy in patients with EC

Trial	Population characteristics and intervention	Recurrence outcomes	OS
GOG-99101 (phase III)	392 patients with intermediate-riska EC who underwent abdominal staging surgery including total hysterectomy and routine lymphadenectomy were randomized (1:1) to pelvic EBRT (total 50.48 Gy with >28 180-cGy fractions) vs observation	Overall: 2-year CIR 3% vs 12%; HR 0.42, 90% CI 0.25–0.73; P = 0.007	Overall: estimated 4-year OS 92% vs 86%; HR 0.86, 90% CI 0.57–1.29; P = 0.55
		High–intermediate risk: 2-year CIR 26% vs 6%; HR 0.42, 90% CI 0.21–0.83	High–intermediate risk: estimated 4-year OS 88% vs74%; HR 0.73, 90% CI 0.43–1.26
		Low–intermediate risk: 2-year CIR 5% vs 2%; HR 0.46, 90% CI 0.19–1.11	Low–intermediate risk: estimated 4-year OS 94% vs 92% HR 1.04; 90% CI 0.56–1.93
PORTEC-1102,104, 115 (phase III)	715 patients with stage 1 (grade 1–3) EC who underwent total abdominal hysterectomy and bilateral salpingo–oophorectomy, without lymphadenectomy were randomized (1:1) to pelvic EBRT (total of 46 Gy delivered as 2-Gy daily fractions) vs observation	5-year LRR 4% vs 14%, P <0.001; 5-year LRR in patients with aberrant p53 3.8% vs 27.8%, P = 0.15b, 5-year LRR in patients with NSMP 1.7% vs 12.3%, P = 4.6 × 10−4b; DRR 7.9% vs 7%	5-year OS 81% vs 85%, P = 0.31
PORTEC-2103,105, 115 (phase III)	427 patients with stage I–IIA high–intermediate risk EC who underwent total abdominal hysterectomy and bilateral salpingo-oophorectomy without lymphadenectomy were randomized (1:1) to VBT (21-Gy high-dose rate or 30 Gy low-dose rate) vs pelvic EBRT (46 Gy delivered as 23 fractions)	5-year LRR 5.1% vs 2.1%, HR 2.08, 95% CI 0.71–6.09, P = 0.17; p53abn 0% vs 31.4%b, P = 0.06, NSMP 1.8% vs 2.3%, P = 0.38b 5-year pelvic recurrence 1.5% vs 0.5%, HR 3.1, 95% CI 0.32–29.9, P = 0.30; 5-year vaginal recurrence 1.8% vs 1.6%, HR 0.78, 95% CI 0.17–3.49, P = 0.74; 5-year distant recurrence 8.3% vs 5.7%, HR 1.32, 95% CI 0.62–2.74, P = 0.46	5-year OS 84.8% vs 79.6%, HR 1.17, 95% CI 0.69–1.98, P = 0.57
PORTEC-3 109 (phase III)	660 patients with stage I–III EC who underwent total abdominal or laparoscopic hysterectomy and bilateral salpingo-oophorectomy, with optional lymphadenectomy were randomized (1:1) to CRT (48.6 Gy EBRT delivered in 1.8-Gy fractions plus 2 cycles of cisplatin on week 1 and week 4 of RT followed by 4 cycles of carboplatin–paclitaxel vs pelvic EBRT (48.6 Gy delivered in 1.8-Gy fractions) only	5-year PFS 76.5% vs 69.1%, HR 0.70, 95% CI 0.52–0.94, P = 0.02; MMRd 68% vs 75.5%, HR 1.29, 95% CI 0.68–2.45, P = 0.43b; p53abn 58.6% vs 36.2%, HR 0.52, 95% CI 0.30–0.91, P = 0.02b; NSMP 79.7% vs 67.7%, HR 0.68, 95% CI 0.36–1.30, P = 0.25b	5-year OS 81.4% vs 76.1%, HR 0.70, 95% CI 0.51–0.97, P = 0.03; MMRd 78.6% vs 84%, HR 1.33, 95% CI 0.64–2.75, P = 0.45a; p53abn 64.9% vs 41.8%, HR 0.55, 95% CI 0.30–1.0, P = 0.05b; NSMP 89.3% vs 87.6%, HR 0.68, 95% CI 0.26–1.77, P = 0.44b
		Stage I–II: 5-year PFS 81.3% vs 77.3%; HR 0.87, 95% CI 0.56–1.36; P = 0.54	Stage I–II: 5-year OS 83.8% vs 82.0%; HR 0.84, 95% CI 0.52–1.38; P = 0.50
		Stage III: 5-year PFS 70.9% vs 58.4%; HR 0.61, 95% CI 0.42–0.89; P = 0.01	Stage III: 5-year OS 78.5% vs 68.5%; HR 0.63, 95% CI 0.41–0.99; P = 0.04
		Serous only: 5-year PFS 59.7% vs 47.9%; HR 0.42, 95% CI 0.22–0.80; P = 0.008	Serous only: 5-year OS 71.4% vs 52.8%; HR 0.48, 95% CI 0.24–0.96; P = 0.04
GOG-258110, 111 (phase III)	707 patients with surgically staged III–IVA (stage I/II serous or clear cell permitted if positive peritoneal washings) EC who underwent hysterectomy and bilateral salpingo-oophorectomy with optional lymphadenectomy were randomized (1:1) to CRT (45 Gy EBRT delivered in 1.8-Gy fractions plus 2 cycles of cisplatin on days 1 and 29 of RT followed by 4 cycles of carboplatin–paclitaxel vs 6 cycles of carboplatin–paclitaxel	5-year PFS 59% vs 58%, HR 0.90, 90% CI 0.74–1.10; P = 0.20; 5-year vaginal recurrence 2% vs 7% HR 0.36 (95% CI 0.16–0.82); 5-year nodal recurrence 11% vs 20%, HR 0.43, 95% CI 0.28–0.66; 5-year distant recurrence 27% vs 21%, HR 1.36, 95% CI 1.0–1.86	mOS NR vs NR; HR 1.05, 95% CI 0.82–1.34; P = 0.72
GOG-249106 (phase III)	601 patients with stage I–II high–intermediate risk EC who underwent hysterectomy with bilateral pelvic and para-aortic lymphadenectomy also recommended were randomized (1:1) to EBRT (45–50.4 Gy delivered in 1.8-Gy fractions) vs VBT (high dose-rate or low dose-rate) followed by 3 cycles of carboplatin–paclitaxel	5-year RFS HR 0.92, 95% CI 0.65 – 1.30; P = 0.31	5-year OS HR 1.04, 95% CI 0.66 – 1.63; P = 0.57
		Stage I endometrioid: 5-year RFS HR 0.90, 95% CI 0.56–1.45	Stage I endometrioid: 5-year OS HR 1.12, 95% CI 0.60–2.06
		Stage II endometrioid: 5-year RFS HR 1.21, 95% CI 0.52–2.79	Stage II endometrioid: 5-year OS HR 2.04, 95% CI 0.51–8.15
		Stage I serous/clear cell: 5-year RFS HR 1.17, 95% CI 0.54–2.53	Stage I serous/clear cell: 5-year OS HR 0.93, 95% CI 0.34–2.57
		Stage II serous/clear cell: 5-year RFS HR 0.39, 95% CI 0.12–1.28	Stage II serous/clear cell: 5-year OS HR 0.67, 95% CI 0.19–2.38

abn, abnormal; CI, confidence interval; CIR, cumulative incidence of recurrence; DMI, depth of myometrial invasion; DRR, durable response rate; EC, endometrial carcinoma; EBRT, external beam radiation therapy; HR, hazard ratio; LRR, locoregional recurrence rate; LVSI, lymphovascular space invasion; MMRd, mismatch-repair deficient; NR, not reported; NSMP, no specific molecular profile; OS overall survival; PFS, progression-free survival; mut, mutation; RFS, recurrence-free survival; RT, radiotherapy; VBT, vaginal brachytherapy.

^aGOG-99 high–intermediate risk criteria: ≥70 years of age with one uterine risk factor, ≥50 years of age with two uterine risk factors, or <50 years of age with 3 uterine risk factors. Uterine risk factors include grade 2–3 tumours, DMI ≥50%, and lymphovascular invasion.

^bData are from a post-hoc analysis.

High-risk EC.

High-risk EC is a heterogeneous group that encompasses tumours of a high-grade non-endometrioid histology of any stage, deeply invasive stage I grade 3 endometrioid histology, or any stage III/IV histology and is associated with inferior survival outcomes^{98,100,107,108}. Several phase III trials have evaluated adjuvant therapy strategies in these patients^{39,106,109,110} (TABLE 2). In GOG-249, patients with high–intermediate risk, stage II, and stage I–II serous or clear-cell tumours were randomized to EBRT versus VBT plus three cycles of carboplatin and paclitaxel¹⁰⁶. Despite no significant differences in OS and recurrence-free survival (RFS), patients in the VBT plus chemotherapy arm had a greater incidence of nodal recurrence and increased acute toxicities¹⁰⁶. The trial investigators recommended EBRT over VBT plus chemotherapy based on these findings, although whether oncological outcomes would differ if six cycles of adjuvant chemotherapy plus VBT were administered remains unclear. Furthermore, the trial was not powered to evaluate differences in oncological outcomes among patients with serous or clear-cell subtypes, which have a greater propensity for distant recurrence.

Two phase III trials compared chemoradiotherapy (CRT) with either radiotherapy or chemotherapy alone^39,109,110. In PORTEC-3, a heterogeneous group of patients with high-risk EC (TABLE 2) were randomized to either CRT (with two cycles of concurrent cisplatin) followed by four cycles of carboplatin plus paclitaxel versus EBRT alone^39,109. In an updated analysis, patients in the CRT arm had greater 5-year RFS (76.5% versus 69.1%; P = 0.016) and OS (81.4% versus 76.1%; P = 0.034), with patients with stage III or serous tumours deriving the highest level of benefit¹⁰⁹. In GOG-258, patients with high-risk EC were randomized to CRT (two cycles of concurrent cisplatin) followed by four cycles of carboplatin plus paclitaxel versus six cycles of carboplatin plus paclitaxel alone¹¹⁰. No significant differences in 5-year RFS were observed (59% versus 58%). However, patients in the CRT arm had a fewer locoregional recurrences (2% versus 7% and 11% versus 20% for vaginal or para-aortic lymph node recurrences, respectively), albeit with an increase in distant recurrences compared to the chemotherapy only arm (27% versus 21%)¹¹⁰. Median OS was not reached for either group on updated survival analyses¹¹¹. Various studies have also explored the utility of other chemotherapy and/or radiotherapy strategies (such as sequential or ‘sandwich’ regimens) in patients with high-risk EC^112–114. In a phase III trial (Lunchbox), the investigators evaluated CRT followed by chemotherapy versus sandwich therapy (chemotherapy followed by radiotherapy then chemotherapy) in patients with advanced-stage endometrial cancer or stage I serous or clear-cell histology with positive cytology¹¹⁴. Although the trial closed prematurity owing to low accrual (n = 48) and was therefore underpowered, no significant differences were observed between arms in terms of adverse events or survival outcomes¹¹⁴. Owing to heterogeneous study populations and treatment protocols, the optimal adjuvant treatment strategy for patients with high-risk EC has yet to be determined and should be individualized based on histopathological and molecular features.

Adjuvant therapy in the era of molecular classification.

Given the clinical significance of the molecular subtypes outlined in the TCGA analysis, the PORTEC investigators sought to evaluate the efficacy of adjuvant treatment regimens in patients with specific EC molecular subtypes participating in the PORTEC1–3 trials^115,116 (Table 2). In an analysis of data from patients with high–intermediate risk disease in the PORTEC-1 and PORTEC-2 trials, investigators observed no locoregional recurrence in any patient with a POLE-mutant tumour, regardless of treatment modality (observation, VBT, or EBRT)¹¹⁵. Among patients with p53 abnormal tumours, those who received EBRT had fewer locoregional recurrences compared to those treated with VBT or observation¹¹⁵. Receiving any radiotherapy seemed to provide limited benefit for patients with MMRd tumours but provided significantly improved locoregional control for patients with NSMP tumours (96.2%–98.3% versus 87.7%; P <0.0001)¹¹⁵. Among patients with high-risk EC in the PORTEC-3 trial, investigators observed significantly improved 5-year RFS with CRT versus the radiotherapy-only arm (59% versus 36%; P = 0.019) among patients with TP53-mutated tumours but no significant benefit from CRT in patients with MMRd tumours or those with NSMP tumours¹¹⁶. As anticipated, patients with POLE-mutant tumours had excellent RFS outcomes regardless of treatment received (100% versus 97%), suggesting that chemotherapy could be safely omitted in this population¹¹⁶. Thus, analysis of data from the PORTEC trials highlights how molecular classification and/or specific biomarkers might be utilized as tools to triage patients to receive the most appropriate adjuvant therapy.

Several clinical trials attempting to prospectively integrate molecular markers into adjuvant therapy decision-making are currently under way. PORTEC-4A is a multicentre phase III trial comparing adjuvant therapy regimens comprising either molecularly-integrated risk-based recommendations (such as observation, VBT or EBRT) with standard adjuvant VBT in patients with high–intermediate risk EC (NCT03469674). Other trials testing molecularly guided therapies include TAPER (NCT04705649), CAN-STAMP (NCT04159155) and RAINBO (NCT05255653).

Circulating tumour DNA (ctDNA) is a novel, minimally invasive blood-based biomarker that has demonstrated prognostic potential in patients with multiple solid tumours, including those with EC^117,118. ctDNA might be released into circulation passively (following cancer cell apoptosis or necrosis) or actively (as extracellular vesicles containing DNA)¹¹⁹. Methods of detecting and quantifying ctDNA are constantly evolving¹¹⁷ and are beyond the scope of this Review. However, a multitude of potential clinical applications for ctDNA exist including early diagnosis, detection of minimal residual disease following first-line therapy, monitoring for disease recurrence, triage of patients for adjuvant treatment, real-time identification of resistance mutations, and monitoring response to therapy. Data from a prospective study involving 44 patients with EC indicate that the presence of ctDNA following surgical staging and/or post-adjuvant therapy is associated with inferior PFS^118,120. Longitudinal monitoring of changes in ctDNA might be a useful method of de-escalating or escalating the intensity of adjuvant therapy^117,119. Further prospective data, ideally associated with clinical trials, are urgently needed to assess the utility of incorporating data from ctDNA-based assays into clinical decision making.

Nonsurgical management

Surgical management is the standard of care for patients with early stage EC, although other non-surgical alternatives might be reasonable in certain patient populations. For women of reproductive age who desire future childbearing, endocrine therapy might be offered in select cases after extensive counselling. Ideal candidates include those with FIGO grade 1 histology ECs confirmed on dilatation and curettage, or those with tumours confined to the endometrium (confirmed on imaging, preferably MRI), aged <45 years with otherwise good reproductive potential, no contraindications to hormone therapy, and who are likely to adhere to both treatment and follow-up monitoring. Progestin-based therapy is recommended and can include oral options (such as megestrol acetate, medroxyprogesterone acetate) and a progestin-releasing intrauterine device (IUD). No consensus exists regarding the optimal regimen and dosage, and complete response rates (CRRs) among patients managed using this approach range from 48–96%^121–123. Given the lower risk of systemic adverse effects and elimination of any concerns regarding compliance, a progestin-releasing IUD is an attractive strategy; addition of oral progestins might be considered based on physician preference and treatment response on follow-up biopsy sampling. Endometrial sampling (such as office endometrial biopsy sampling) can be safely performed with an IUD in situ and should be conducted every 3–6 months to monitor treatment response. Approximately 35% of patients managed using this approach will have disease recurrence and those with persistent or progressive disease (for 9–12 months since initiation of endocrine therapy) should be considered for hysterectomy¹²³. Most women attempting to conceive in this scenario will require assisted reproductive technology owing to the presence of co-existing risk factors for infertility (such as obesity, polycystic ovarian syndrome and chronic anovulation) and consultation with an onco-fertility specialist is recommended; the live birth rate is approximately 28%^123,124. Following completion of family planning, total hysterectomy should be performed.

Patients with clinically significant medical comorbidities or morbid obesity with stage I low-grade EC might also benefit from progestin-based therapy. Primary radiotherapy is another viable alternative that has resulted in 5-year disease-free survival rates of 57%¹²⁵.

Management of recurrent or unresectable advanced-stage disease

Locoregional recurrence in select patients

Patients with isolated, locoregionally recurrent tumours might be amenable to radiotherapy or surgery depending on anatomical location of the tumour. The vaginal apex is the most frequent location of locoregional recurrence and salvage radiotherapy provides excellent outcomes in this scenario, particularly in radiotherapy-naive patients (CRR 87%)¹²⁶. The prospective GOG-238 study compared the efficacy of CRT to that of radiotherapy alone in 165 patients with localized disease recurrence¹²⁷. The majority of patients had grade 1 or 2 EC confined to the vaginal apex. Patients who received CRT had similar outcomes to those treated with radiotherapy alone suggesting that, at least for patients with low-risk recurrent disease, radiotherapy alone might be appropriate¹²⁷. However, those with extravaginal or regional lymph node involvement have an inferior response to salvage radiotherapy (CRR 40%)¹²⁶. Patients with central, locoregional recurrence who are not amenable to radiotherapy with a curative intent might be candidates for total pelvic exenteration. Eligible candidates should undergo extensive evaluation to rule out distant metastases, evaluate the likelihood of complete gross tumour resection, and evaluate and/or manage any comorbidities in order to reduce the risks of perioperative morbidity and mortality to acceptable levels. Given pelvic exenterations require the creation of a permanent ileal conduit and end colostomy, comprehensive counselling and psychological evaluation is recommended to ensure complete comprehension of the effects of surgery on both physical function and quality of life. The cure rate of patients undergoing total pelvic exenteration is approximately 50% and 5-year OS is 56%¹²⁸.

First-line therapy based on tumour biology

Patients with recurrent EC or unresectable advanced-stage EC at diagnosis are managed with systemic therapy. Historically, chemotherapy has been the standard-of-care systemic therapy in these patients. Various chemotherapies have antitumour activity in patients with EC, including platinum-based agents (such as carboplatin and cisplatin), taxanes (such as paclitaxel) and doxorubicin. On the basis of data from a range of phase III trials comparing combination regimens, paclitaxel, doxorubicin and cisplatin (TAP) is viewed as the most effective regimen, although toxicities and dose scheduling can both be problematic¹²⁹. GOG-209, an open-label, randomized, non-inferiority phase III trial, compared carboplatin plus paclitaxel versus TAP in patients with chemotherapy-naive unresectable stage III/IV or recurrent EC¹³⁰. Carboplatin plus paclitaxel demonstrated noninferior median OS (37 months versus 41 months) and PFS (13 months versus 14 months) with reductions in the incidence of certain adverse events with carboplatin–paclitaxel, including grade ≥2 sensory neuropathy (in 26% versus 20% of patients) ¹³⁰. Another phase III trial (NRG-GOG 0261) testing carboplatin–paclitaxel versus paclitaxel–ifosfamide in patients with uterine carcinosarcoma demonstrated improved median PFS in patients with stage III/IV disease, as well as noninferior median OS¹³¹. Although the final results of both GOG-209 and NRG-GOG 0261 were only reported in 2022, carboplatin–paclitaxel has been used for at least a decade as a standard first-line systemic therapy option for patients with advanced-stage or recurrent EC regardless of tumour histology or biology. However, owing to a dramatic increase in the number trials evaluating novel therapeutic agents and biomarker-driven management, the treatment approach for patients with advanced-stage EC is continuously evolving and being refined.

MMRd/MSI-H tumours

The efficacy of chemotherapy plus anti-PD-1 or anti-PD-L1 antibodies in patients with advanced-stage or recurrent EC has been evaluated in three randomized phase III trials (KEYNOTE-868/NRG-GY018, RUBY and DUO-E) with similar cohort characteristics^132–134 (Table 3). All of these trials contained subgroups with mismatch repair proficient/microsatellite stable (MMRp/MSS) and MMRd/MSI-H tumours^132–134. In KEYNOTE-868/NRG-GY018, patients were randomized 1:1 to carboplatin–paclitaxel and either pembrolizumab or placebo for 6 cycles followed by maintenance therapy with pembrolizumab or placebo maintenance therapy every 6 weeks for up to 2 years¹³². Among those with MMRd tumours, median PFS was significantly longer in the pembrolizumab group than in the placebo group (not reached versus 7.6 months, HR 0.30, 95% CI 0.19–0.48)¹³². Similarly, patients in RUBY were randomized 1:1 to receive carboplatin–paclitaxel along with either dostarlimab or placebo for 6 cycles followed by maintenance dostarlimab or placebo every 6 weeks for up to 3 years¹³³. PFS outcomes were significantly improved in the dostarlimab group (24-month PFS 61.4% versus 15.7%, HR 0.28, 95% CI 0.16–0.50; P <0.001) with substantially improved OS also observed (83.3% versus 58.7%, HR 0.30, 95% CI 0.13 – 0.70)¹³³. In DUO-E, patients were randomized 1:1:1 to carboplatin–paclitaxel plus durvalumab followed by maintenance durvalumab and olaparib (durvalumab plus olaparib arm), carboplatin–paclitaxel plus durvalumab followed by maintenance durvalumab plus placebo (durvalumab-only arm), or carboplatin–paclitaxel plus placebo followed by maintenance placebo (the control arm)¹³⁴. Median PFS was significantly longer In both the durvalumab plus olaparib arm and in the durvalumab-only arm (31.8 months versus not reached versus 7 months, respectively) although no significant difference was observed between the two experimental arms (HR 0.97, 95% CI 0.49–1.98)¹³⁴ Overall, grade ≥3 adverse events were 67.2%, 54.9% and 56.4% among the treatment arms, respectively, and were mainly hematologic (e.g neutropenia, anemia, or thrombocytopenia)¹³⁴. These findings are further supported by data from the MMRd subgroup of the randomized phase II MITO END-3 trial, in which patients with receiving carboplatin–paclitaxel plus avelumab had an improved median PFS (not reached versus 8 months, respectively) compared to patients who received carboplatin–paclitaxel only; a similarly favourable trend was observed for median OS (not reached versus 26.1 months, respectively)¹³⁵. Interim results from the phase III AtTEnd/ENGOT-en7 trial presented at the 2023 ESMO annual meeting demonstrated significantly improved median PFS with carboplatin–paclitaxel plus atezolizumab compared with carboplatin–paclitaxel only (not reached versus 6.9 months, HR 0.36, 95% CI 0.23–0.57; P = 0.0005); OS data are currently immature¹³⁶. Given these impressive oncological outcomes, carboplatin–paclitaxel plus an anti-PD-1 or anti-PD-L1 antibody should be the standard-of care first-line therapy for patients with advanced-stage or recurrent MMRd/MSI-H EC. It should be noted at this time of this Review, pembrolizumab, dostarlimab, and durvalumab have all received FDA approval to utilized in combination with chemotherapy for MMRd/MSI-H EC.

Table 3 |.

Chemotherapy and/or immunotherapy combinations in patients with advanced-stage or recurrent EC

Study	Trial design and population characteristics	Recurrence outcomes	OS
Chemotherapy only
GOG-209130 (phase III)	1,304 patients with stage III/IV or recurrent EC were randomized to up to 7 cycles of paclitaxel–carboplatin vs up to 7 cycles of paclitaxel–doxorubicin–cisplatin	mPFS 13.2 vs 13.9 months; HR 1.032, 90% CI 0.93–1.15	mOS 37 vs 41.1 months; HR 1.0, 90% CI 0.90–1.12
NRG-GOG 0261131 (phase III)	539 patients with ovarian or uterine carcinosarcoma were randomized 1:1 to received 6–10 cycles of paclitaxel–carboplatin vs 6–10 cycles of paclitaxel–ifosfamide	mPFS 16 vs 12 months, HR 0.74, 95% CI 0.58–0.93, P <0.001; stage I/II: HR 1.10, 95% CI 0.72–1.68; stage III/IV: HR 0.65, 95% CI 0.48–0.88; recurrent: HR 0.86, 95% CI 0.44–1.67	mOS 37.3 vs 29.0 months, HR 0.87, 90% CI 0.70–1.08; stage I/II: HR 1.29, 95% CI 0.80–2.08; stage III/IV: HR 0.74, 95% CI 0.54–1.01; recurrent: HR 0.79, 95% CI 0.40–1.58
Chemotherapy plus immunotherapy
NRG-GY018132 (phase III)	813 patients with advanced-stage or recurrent EC were randomized (1:1) to 6 cycles of carboplatin–paclitaxel plus pembrolizumab followed by pembrolizumab maintenance vs 6 cycles of carboplatin–paclitaxel plus placebo × 6 followed by placebo maintenance	MMRd: mPFS: NR vs 7.6 months; HR 0.30, 95% CI 0.19–0.48; P <0.001	Immature
		MMRp: mPFS 13.1 vs 8.7 months; HR 0.54, 95% CI 0.41– 0.71; P <0.001
RUBY133 (phase III)	487 patients with advanced-stage or recurrent EC were randomized (1:1) to 6 cycles of carboplatin–paclitaxel plus dostarlimab followed by dostarlimab maintenance vs 6 cycles of carboplatin–paclitaxel plus placebo × 6 followed by placebo maintenance	MMRd/MSI-H: 24-month PFS 61.4% vs 15.7%; HR 0.28, 95% CI 0.16– 0.50; P <0.001	MMRd/MSI-H: 24-month OS 83.3% vs 58.7%; HR 0.30, 95% CI 0.13–0.70
		MMRp/MSS: 24-month PFS 28.4% vs 18.8%; HR 0.76 95% CI 0.59–0.98	MMRp/MSS: 24-month OS 67.7% vs 55.1%; HR 0.73, 95% CI 0.52–1.20
DUO-E134 (phase III)	718 patients with advanced-stage or recurrent EC were randomized (1:1:1) to 6 cycles of carboplatin–paclitaxel plus durvalumab followed by durvalumab plus olaparib maintenance vs 6 cycles of carboplatin–paclitaxel plus durvalumab followed by durvalumab maintenance vs carboplatin–paclitaxel plus placebo followed by placebo maintenance	MMRd: mPFS 31.8 months (HR vs chemotherapy alone 0.41, 95% CI 0.21–0.75) vs NR (HR vs chemotherapy alone 0.42, 95% CI 0.22–0.80) vs 7 months	Immature
		MMRp: mPFS 15.0 months (HR vs chemotherapy 0.57, 95% CI 0.44–0.73) vs 9.9 months (HR vs chemotherapy alone 0.77, 95% CI 0.60 – 0.97) vs 9.7 months
		PD-L1+: mPFS 20.8 months (HR vs chemotherapy 0.42, 95% CI 0.31–0.57) vs 11.3 months (HR 0.63, 95% CI 0.48–0.83) vs 9.5 months
		PD-L1–: mPFS 10.1 months (HR vs chemotherapy 0.80, 95% CI 0.55 – 1.16) vs 9.7 months (HR vs chemotherapy 0.89, 95% CI 0.59–1.34) vs 9.9 months
MITO END-3135, 137 (phase II)	125 patients with advanced-stage or recurrent EC were randomized (1:1) to 6–8 cycles of carboplatin–paclitaxel plus avelumab followed by avelumab maintenance vs 6–8 cycles of carboplatin–paclitaxel	MMRd: mPFS NR vs 8 months; HR 0.46, 95% CI 0.22–0.94	MMRd: mOS NR vs 26.1 months; HR 0.41, 95% CI 0.14–1.18
		MMRp: 8.3 vs 10.8 months; HR 1.17, 95% CI 0.65–2.10	MMRp: mOS 22.3 months vs NR; HR 2.21, 95% CI 0.86–5.26
		MMRp/TP53-mutated: HR 2.24 (95% CI 1.06 – 4.72); ARID1A-mutated: HR 0.48 (95% CI 0.22 – 1.01); PTEN-mutated: HR 0.36, 95% CI 0.17–0.76	MMRp/TP53-mutated: HR 2.08, 95% CI 0.82–5.26; ARID1A-mutated: HR 0.82, 95% CI 0.32–2.10; PTEN-mutated: HR 0.39, 95% CI 0.13–1.20
AtTEnd/ENGOT-en7 Trial.136 (phase III)	549 patients with advanced stage or recurrent EC were randomized (2:1) to 6–8 cycles of carboplatin–paclitaxel plus atezolizumab followed by atezolizumab maintenance vs 6–8 cycles of carboplatin–paclitaxel plus placebo followed by placebo maintenance	MMRd: mPFS NR vs 6.9 months, HR 0.36, 95% CI 0.23– 0.57, 12-month PFS 62.7% vs 23.3%, 24-month PFS 50.4% vs 16.0%	MMRd: mOS NR vs 25.7 months, HR 0.41, 95% CI 0.22–0.76, 12-month OS 86.8% vs 66.8%, 24-month OS 75% vs 54.2%
		MMRp: mPFS 9.5 vs 9.2 months, HR 0.92, 95% CI 0.73–1.16; 12-month PFS 39.5% vs 30.2%; 24-month PFS 21.3% vs 16.4%	MMRp: mOS: 31.5 vs 28.6 months, HR 1.0, 95% CI 0.74–1.35; 12-month OS 77.8% vs 77.3%; 24-month OS 57.4% vs 58.3%
Immunotherapy plus oral tyrosine kinase inhibitor
KEYNOTE-775153 (phase III)	827 patients with advanced-stage or recurrent EC were randomized 1:1 to lenvatinib plus pembrolizumab vs physician’s choice of chemotherapy (doxorubicin or paclitaxel)	Overall: mPFS 7.2 vs 3.8 months; HR 0.56, 95% CI 0.47–0.66; P <0.001	Overall: mOS 18.3 vs 11.4 months; HR 0.62, 95% CI 0.51–0.75; P <0.001)
		MMRp: mPFS 6.6 vs 3.8 months; HR 0.60, 95% CI 0.50–0.72; P <0.001	MMRp mOS: 17.4 vs 12 months; HR 0.68, 95% CI 0.56–0.84; P <0.001
		MMRd: mPFS 10.7 vs 3.7 months; HR 0.36, 95% CI 0.23–0.57; P <0.001	MMRd: mOS NR vs 8.6 months; HR 0.37, 95% CI 0.22–0.62; P <0.001

CI, confidence interval; EC, endometrial carcinoma; HR, hazard ratio; MMRd, DNA mismatch repair deficient; MMRp, DNA mismatch repair proficient; mOS, median overall survival; mPFS, median progression-free survival; MSI, microsatellite instability; MSS, microsatellite stable; NR, not reported; OS, overall survival; PFS, progression-free survival; T-DXd, trastuzumab deruxtecan.

MMRp/MSS tumours

In comparison to those with MMRd/MSI-H EC, responses of patients with advanced-stage or recurrent pMMR EC to regimens comprising carboplatin–paclitaxel plus an anti-PD-1 or anti-PD-L1 antibody have been modest^132–134. In KEYNOTE-868/NRG-GY018, patients in the pembrolizumab group had an improved median PFS compared to the placebo group (13.1 months versus 8.7 months, HR 0.54, 95% CI 0.41–0.71; P <0.001)¹³². In June 2024, the FDA approved the combination of pembrolizumab plus chemotherapy in advanced or recurrent EC, regardless of MMR/MSI status. Although the OS data is still immature, the final results of KEYNOTE-868/NRG-GY018 are eagerly anticipated. In RUBY, the dostarlimab group had numerically longer PFS than the placebo group (24-months PFS 28.4% versus 18.8%, HR 0.76, 95% CI 0.59–0.98) albeit with a <10% difference in PFS at 24 months¹³³. In DUO-E, patients in the durvalumab-only arm had a modest improvement in median PFS of 0.2 months (9.9 versus 9.7 months; HR 0.77, 95% CI 0.60 – 0.97)¹³⁴. Similar to the RUBY trial results, the 18-month PFS rates were higher in the durvalumab-only arm compared to the control arm (31.3% vs 20%, respectively)¹³⁴. In interim data from the AtTEnd/ENGOT-en7 trial, investigators observed similar median PFS durations (9.5 versus 9.2 months) in the avelumab plus chemotherapy versus chemotherapy-only arms¹³⁶. Although OS data from these trials remain immature, observable differences in OS between the experimental and control arms have yet to become apparent^{132–134,136}. In light of these data, the magnitude of OS benefit from routinely adding anti-PD-1 or PD-L1 antibodies to chemotherapy in patients with MMRp/MSS ECs in the absence of other known biomarkers remains to be determined. Interestingly, although not statistically significant, the MITO END-3 investigators observed improvements in median PFS (8.3 months versus 10.8 months, HR 1.17, 95% CI 0.65–2.10) and OS (22.3 months versus not reached, HR 2.21, 95% CI 0.86–5.26) favouring the control arm over the immunotherapy arm¹³⁵. A preplanned exploratory translational analyses of data from MITO END-3 indicates that chemotherapy plus avelumab is associated with inferior PFS outcomes in the presence of a TP53 mutation (P = 0.003) and improved PFS in the presence of ARID1A (P = 0.01) and PTEN (P = 0.002) mutations; these observations highlight the importance of establishing biomarkers capable of guiding the use of ICIs¹³⁷. The availability of mature OS data from these landmark trials is eagerly anticipated and should be followed by attempts to identify biomarkers enabling the identification of patients who derive the greatest level of benefit from these regimens. Additionally, further clarification on the impact of combination vs sequential ICI strategies (e.g. chemotherapy plus pembrolizumab vs chemotherapy only followed by pembrolizumab plus lenvatinib in second-line setting) on overall OS should be investigated in future strategies.

Several situations exist in which ICIs can be considered for patients with MMRp/MSS ECs, although longer-term survival data are required. In DUO-E, such patients in the durvalumab plus olaparib arm had significantly longer PFS than those in the control arm (15.0 versus 9.7 months, HR 0.55, 95% CI 0.43 to 0.69; P <0.0001)¹³⁴. Interestingly, among the subgroup of patients with PD-L1-positive tumours (defined as a tumour area positivity score >1%, 483 of 701 patients; 68.9%), those receiving durvalumab plus olaparib had the greatest median PFS compared to the durvalumab-only and control arms (20.8 versus 11.3 versus 9.5 months)¹³⁴. The proportion of MMRd/MSI-H tumours among the PD-L1-positive EC subgroup in DUO-E is unclear, although the majority are likely to have MMRp/MSS tumours given that 19.9% of the entire study population had MMRd/MSI-H disease. Furthermore, interim OS data (28% maturity) from DUO-E suggest an OS advantage in the durvalumab plus olaparib arm compared to the control arm (HR 0.59, 95% CI 0.42–0.83; P <0.003)¹³⁴. In the second part of the ongoing phase III RUBY trial (NCT03981796), investigators are comparing carboplatin–paclitaxel plus dostarlimab followed by dostarlimab plus niraparib maintenance versus carboplatin–paclitaxel plus placebo followed by placebo maintenance in a study population with nearly identical characteristics to those of the part 1 cohort¹³⁸. The RUBY part 2 investigators presented interim results at the 2024 Annual Meeting of the Society of Gynecologic Oncology, which indicate a 6-month improvement in median PFS in favour of the experimental arm (14.3 months versus 8.3 months) among patients with MMRp/MSS tumours; OS data remain immature¹³⁸. Like DUO-E, data from RUBY part 2 indicate possible synergy between ICIs and poly ADP-ribose polymerase (PARP) inhibitors, and mature OS data from both trials are therefore eagerly anticipated. However, both DUO-E and RUBY part 2 lack a chemotherapy plus PARP inhibitor-only arm, making it difficult to determine the relative contributions of each agent in the ICI plus PARP inhibitor arms^134,138. Regardless, the efficacy of ICI–PARP inhibitor combinations should be further investigated in the MMRp/MSS EC population. Lastly, although not prospectively evaluated, patients with MMRp/MSS EC tumours with a high TMB (≥10 mut/Mb) might also benefit from the combination of chemotherapy and an anti-PD-1 antibody given the established associations between this biomarker and response to anti-PD-1 antibodies as monotherapy^139,140.

HER2 overexpression and p53 status

HER2 overexpression, denoted as HER2 IHC3+ or IHC2+ with fluorescence in situ hybridization (FISH) – positivity, has been reported in 1% of patients with endometrioid ECs, although this incidence might be as high as 44% in those with serous EC^33,141. Nonetheless, most HER2-positive ECs have a non-serous histology and, therefore, universal HER2 testing might be considered, particularly in the advanced-stage or recurrent disease setting¹⁴². In a randomized phase II trial, investigators randomized patients with HER2-positive stage III/IV or recurrent serous EC to either carboplatin–paclitaxel plus trastuzumab for 6 cycles followed by trastuzumab maintenance therapy until progression or unacceptable toxicity or carboplatin–paclitaxel¹⁴³ (Table 4). The addition of trastuzumab to chemotherapy significantly improved median PFS in patients with advanced-stage disease (17.7 months versus 9.3 months; HR 0.44, 90% CI 0.23–0.83; P = 0.015)¹⁴³. Median OS was also significantly longer in the in the trastuzumab arm (31.9 months versus 21.3 months; HR 0.44, 90% CI 0.22–0.88; P = 0.02)¹⁴³. Thus, trastuzumab should be added to carboplatin and paclitaxel for first-line systemic therapy for patients with advanced-stage HER2-positive serous EC. Additionally, consideration should also be made for the addition of trastuzumab to chemotherapy in non-serous HER2-positive tumours as well.

Table 4 |.

Targeted therapy in patients with advanced-stage or recurrent EC

Study	Trial design and population characteristics	Recurrence outcomes	OS
Chemotherapy plus targeted therapy
Fader et al143 (phase II)	58 patients with advanced-stage or recurrent EC with HER2-amplified serous histology (IHC3+ or IHC2+ and FISH-positive) were randomized (1:1) to 6 cycles of carboplatin–paclitaxel plus trastuzumab followed by trastuzumab maintenance vs carboplatin–paclitaxel	Overall: mPFS 12.9 vs 8 months; HR 0.46, 90% CI 0.28–0.77; P = 0.005	Overall: mOS 29.6 vs 24.4 months; HR 0.58, 90% CI 0.34–0.99; P = 0.05
		Advanced-stage disease: mPFS 17.7 vs 9.3 months; HR 0.44, 90% CI 0.23–0.83; P = 0.02	Advanced-stage disease: mOS NR vs 25.4 months; HR 0.49, 90% CI 0.25–0.97; P = 0.04
		Recurrent disease: mPFS 9.2 vs 7 months; HR 0.13, 90% CI 0.03–0.48; P = 0.004	Recurrent disease: mOS 25 vs 22.5 months; HR 0.86, 90% CI 0.36– 2.1; P = 0.39
GOG-86P145,146 (phase II)	349 patients with measurable stage III/IVA–IVB or recurrent EC were randomized (1:1:1) to carboplatin–paclitaxel plus bevacizumab (CPB) vs carboplatin–paclitaxel plus temsirolimus (CPT) vs carboplatin–ixabepilone plus bevacizumab (CIB)	CPB HR 0.81, 92.2% CI 0.63–10.2b; CPT HR 1.22, 92.2% CI 0.96–1.55b; CIB HR 0.87, 92.2% CI 0.68–1.11a	mOS 34 vs 25 vs 25.2 vs 22.7a months
		TP53 wildtype: mPFS 11.3 vs 9.6 months; HR 0.74, 95% CI 0.49 – 1.12b	TP53 wildtype: mOS 33.6 vs 36.7 months; HR 1.05, 95% CI 0.64–1.77b
		TP53 null: mPFS 18.8 vs 10.5 months; HR 0.71, 95% CI 0.24–2.34b	P53 null: mOS 33.6 months vs NR; HR 1.25, 95% CI 0.35–5.87b
		p53 overexpression: mPFS 12.8 vs 7.8 months, HR 0.46, 95% CI 0.26–0.88b	p53 overexpression: mOS 29.1 vs 14.4 months, HR 0.31, 95% CI 0.16–0.62b
MITO END-2147 (phase II)	106 patients with advanced-stage or recurrent EC were randomized 1:1 to 6–8 cycles of carboplatin–paclitaxel plus bevacizumab followed by bevacizumab maintenance vs 6–8 cycles of carboplatin–paclitaxel	mPFS 13.7 vs 10.5 months; HR 0.85, 95% CI 0.5–1.3; P = 0.44	mOS 40 vs 29.7 months; HR 0.71, 95% CI 0.31–1.36
ENGOT-EN5/GOG-3055/SIENDO148 (phase III)	263 patients with stage IV ECs or with initial recurrence with a partial or complete response to at least 12 weeks of platinum–taxane chemotherapy were randomized 2:1 to selinexor following at least 12 weeks of platinum–taxane chemotherapy vs placebo maintenance following at least 12 weeks of platinum–taxane chemotherapy	Overall: mPFS 5.7 vs 3.8 months; HR 0.71, 95% CI 0.50–0.99; P = 0.05	mDSS NR vs NR; HR 0.92, 95% CI 0.54–1.59; P = 0.77
		TP53-wildtype: mPFS 13.7 vs 3.7 months; HR 0.38, 95% CI 0.21–0.67; P = 0.001
		TP53-mutant: mPFS 3.7 vs 5.6 months; HR 1.31, 95% CI 0.79–2.15; P = 0.29
		TP53 unknown: mPFS 3.8 vs 3.8 months; HR 0.69, 95% CI 0.25–1.89; P = 0.50
Antibody-drug conjugate
DESTINY-PanTumor02156 (phase II)	267 patients with HER2 IHC2+ or IHC3+ locally advanced unresectable or metastatic disease solid tumours (including 40 with EC) received T-DXd	mPFS 11.1 months in all patients with EC; IHC2+ 8.5 months; IHC3+ NR	mOS 26 months in all patients; IHC2+ 16.4 months; IHC 3+ 26 months
STATICE157 (phase II)	32 patients with HER2 IHC ≥1+ carcinosarcoma received T-DXd	mPFS 6.7 months; IHC 1+ 6.7 months; IHC 2–3+ 6.2 months	mOS 15.8 months; IHC1+ NR; IHC 2–3+ 13.3 months

+Note that TP53 mutational status was also obtained by next-generation sequencing testing with 128 wildtype and 85 mutated tumors; 92% concordance between TP53 next-generation sequencing and IHC p53; similar results with the integrated TP53/p53 analysis. CI, confidence interval; EC, endometrial carcinoma; FISH, fluorescence in situ hybridization; HR, hazard ratio; IHC, immunohistochemistry; mDSS, median disease-specific survival; mOS, median overall survival; mPFS, median progression-free survival; NR, not reported; OS, overall survival; PFS, progression-free survival; T-DXd, trastuzumab deruxtecan.

^aCompared with historical data from the carboplatin–paclitaxel group of GOG-209.

^bData from a post-hoc translational analysis.

Knowledge of TP53 mutation status might be useful as a method of identifying patients who are unlikely to respond well to carboplatin–paclitaxel only and thus probably require adjunctive therapy. Although not perfectly concordant with respect to functional status, overexpression of p53 in endometrial tumours is a surrogate marker of TP53 missense mutations and may function as biomarker for response to targeted therapy¹⁴⁴. GOG-86P was a three-arm randomized phase II trial testing carboplatin–paclitaxel plus bevacizumab versus carboplatin–paclitaxel plus temsirolimus versus carboplatin–ixabepilone plus bevacizumab as first-line therapy for patients with advanced-stage or recurrent EC^145,146. Although the initial trial analysis did not demonstrate differences in PFS or OS between arms, data on p53 status were retrospectively obtained from patients’ tumour samples¹⁴⁵. In an unplanned analysis following retrospective investigations of p53 status, patients with p53-overexpressing tumours receiving chemotherapy plus bevacizumab arm had improved median PFS (12.8 months versus 7.8 months; HR 0.48, 95% CI 0.31–0.75) and median OS (29.1 months versus 14.4 months HR 0.61, 95% CI 0.38–0.98) compared to those in the chemotherapy plus temsirolimus arm¹⁴⁵. Conversely, patients with TP53-wildtype tumours did not derive benefit from bevacizumab^145,147. In the phase III ENGOT-EN5/GOG-3055/SIENDO trial, patients with stage IV and/or recurrent EC and a partial or complete response after at least 12 weeks of carboplatin–paclitaxel were randomly assigned (2:1) to receive maintenance therapy with the selective XPO1 nuclear export transporter inhibitor selinexor or placebo¹⁴⁸. Although analysis of the intention-to-treat population demonstrated modest improvements in PFS with selinexor (5.7 versus 3.8 months), a prespecified exploratory subgroup analysis demonstrated that all of this benefit was driven by the TP53-wildtype subgroup (median PFS 13.7 months versus 3.7 months; HR 0.41, 95% CI 0.23–0.72; P = 0.002)¹⁴⁸. Given these early signals, future studies should seek to further evaluate these, and potentially other p53-triaged treatment strategies.

The confounding effects of permissive crossovers in randomized trials

The explosion in the number of potentially effective second-line, or later line therapies, including ICI, requires consideration of the confounding effects of crossover in the first-line setting¹⁴⁹. At first consideration, it may seem that permissive crossover (for example, a patient on the placebo arm of a randomized trial of frontline chemotherapy vs chemotherapy plus an ICI who subsequently receives an ICI in the second line setting) would indicate that the impact on OS of the combination arm might be even greater. However, it is possible that not all eligible patients will receive this effective agent in the second line, leading to potential overestimation of benefit from the first-line combination. Similarly, allowing crossover of patients in the control arm to receive an experimental drug that has not been proven in later lines of therapy might be problematic given the reasonable chance that the drug is completely ineffective or even harmful — another scenario that might lead to overestimation of benefit of the experimental arm. This has led to a recommendation to either explicitly mandate or prohibit crossover in randomized trials testing first-line therapies (depending on the availability of effective later-line therapies), as opposed to the permissive approach used in most trials to date¹⁴⁹.

Second and later line therapy

Decision-making on the use of subsequent lines of therapy after progression on first-line systemic therapy will depend on the previously received agents, the length of any treatment-free interval, and the results of tumour molecular profiling. If not performed already, all tumours should undergo testing for MMR proteins or MSI, oestrogen/progesterone receptor expression, somatic mutations, and PD-L1 and HER2 expression. Disease progression following second-line or later-line therapy is challenging and established chemotherapy options include monotherapy, such as with doxorubicin, paclitaxel or pegylated liposomal doxorubicin, or combination therapy such as cisplatin–gemcitabine. Given the poor prognosis of many patients, clinical trial enrollment should be strongly considered for those who progress on chemotherapy and/or ICIs.

Among patients with ICI-naive MMRd/MSI-H or TMB-high ECs, carboplatin–paclitaxel plus anti-PD-1 or anti-PD-L1 antibodies should be used in patients with a platinum-free interval ≥6 months without disease progression. Patients with a platinum-free interval <6 months would likely benefit from anti-PD-1 or anti-PD-L1 antibodies. Various trials, such as KEYNOTE-158 and GARNET, have demonstrated excellent objective response rates (ORRs) of 42–48% with this approach in these patients^150,151.

Among ICI-naive patients with MMRp/MSS EC with a platinum-free interval ≥6 months, regimens comprising carboplatin, paclitaxel, and an anti-PD-1 or anti-PD-L1 antibody can be used. Irrespective of platinum-free interval, the combination of the oral multi-tyrosine kinase inhibitor lenvatinib with pembrolizumab was regarded as the first choice second-line therapy. However with the approval of pembrolizumab to chemotherapy for first-line systemic therapy for any ECC based on PFS benefit, it is unclear whether combination (e.g. first-line chemotherapy plus pembrolizumab followed by non-ICI second line treatment) vs sequential (e.g. first-line chemotherapy followed by pembrolizumab plus lenvatinib) ICI therapies will result in greater OS benefit. Additionally, there is little data on treatment response to pembrolizumab plus lenvatinib following progression on ICI and the overall impact on OS. Nonetheless, pembrolizumab plus lenvatinib should be considered as a potential treatment option as it has been shown to confer a higher ORR over pembrolizumab monotherapy in patients with MRRp/MSS tumours (30.3% versus 13%)^152–154. In the phase III KEYNOTE-775 trial, patients previously treated with platinum-based chemotherapy who received lenvatinib plus pembrolizumab had significantly longer median PFS (6.6 months versus 3.8 months, HR 0.60; 95% CI 0.50–0.72; P <0.001) and OS (17.4 months versus 12.0 months, HR 0.68, 95% CI 0.56–0.84; P <0.001) compared to those receiving physician’s choice of chemotherapy (doxorubicin or paclitaxel)¹⁵³. However, patients in the lenvatinib–pembrolizumab group frequently required lenvatinib dose reductions (66.5%), interruptions (69.2%) or discontinuations (33%) owing to treatment-related toxicities¹⁵³. Thus, concerns exist regarding the tolerability of lenvatinib and pembrolizumab in clinical practice. However, retrospective data indicate that commencing lenvatinib at a reduced starting dose (14 mg rather than 20 mg daily) results in fewer toxicities with equivalent response and survival outcomes compared with standard dosage¹⁵⁵. These data also indicate that this combination has activity in patients with carcinosarcoma (ORR 25%), a histology excluded from KEYNOTE-775¹⁵⁵. Given these retrospective data as well as the median dose intensity of lenvatinib in KEYNOTE-775 of 13.8 mg daily, the strategy of starting at the 14 mg daily dose level should be considered and might lead to fewer dose interruptions and discontinuations¹⁵⁵.

Patients with HER2-positive EC should receive biomarker-directed therapy. HER2 expression is initially measured using IHC with a score of 0–1+ traditionally indicating HER2 negativity and 3+ indicating overexpression. IHC 2+ denotes equivocal expression, which if followed by FISH to evaluate possible ERBB2 amplification (with HER2-positivity contingent on a ratio of total HER2 signals: total CEP17 signals ≥2). For patients with HER2 IHC3+ or IHC2+ and ERBB2 FISH-positive EC, carboplatin–paclitaxel plus trastuzumab followed by trastuzumab maintenance can be used. This regimen has been demonstrated to confer an improved median PFS compared to chemotherapy alone in the advanced-stage or recurrent disease setting (9.2 months versus 7 months, HR 0.12, 90% CI 0.03–0.48; P = 0.004)¹⁴³. No significant difference in median OS was observed between the chemotherapy plus trastuzumab and chemotherapy alone arms, although this might reflect the small size of the recurrent disease subgroup (n = 17)¹⁴³. HER2-targeted therapies have generally been administered only to patients with HER2 overexpressing (IHC3+ of IHC2+ and FISH-positive) tumours; however, data from the phase II DESTINY-PanTumor02 trial demonstrated clinical benefit in patients receiving trastuzumab deruxtecan (T-DXd), a HER2-directed antibody–drug conjugate equipped with a potent topoisomerase I inhibitor payload, among patients with lower levels of HER2 expression¹⁵⁶. In this single-arm phase II basket trial, investigators evaluated T-DXd in patients with metastatic or recurrent HER2-positive (IHC 3+ or IHC 2+ without confirmatory FISH tumours¹⁵⁶. Patients in the EC subgroup (n = 40) had an ORR of 57.5% with responses in 84.6% (11 of 13) of patients with IHC3+ tumours and 47.1% (8 of 17) for patients with IHC 2+¹⁵⁶. Median PFS in the EC cohort was 11.1 months, including not reached and 8.5 months in the IHC3+ and IHC2+ subgroups, respectively¹⁵⁶. Given the potential benefit, T-DXd received agnostic approval by the FDA for HER2 IHC3+ metastatic solid tumors following progression on a prior line of systemic therapy in April 2024. Interestingly, evidence suggests that T-DXd might provide benefit for patients with carcinosarcomas, even with low levels of HER2 expression¹⁵⁷. In the STATICE trial conducted in Japan, patients with recurrent uterine carcinosarcoma with high (IHC ≥2+; n = 22) and low (IHC 1+; n = 11) levels of HER2 expression received T-DXd¹⁵⁷. Interestingly, ORRs in these subgroups were 54.5% and 70%, with median PFS (6.2 and 6.7 months) and OS (13.3 months and not reached) also reported; these observations should be evaluated further in a larger cohort¹⁵⁷. Given the impressive responses and clinical benefit reported thus far, T-DXd should be considered for patients with IHC≥2+ EC and also potentially for those with IHC≥1+ carcinosarcomas.

Patients with ER⁺ EC might benefit from endocrine therapy, with various regimens providing ORRs of 15–30%, with benefit observed particularly in patients with low-grade tumours^158–160. Given that endocrine therapy is generally well-tolerated relative to other systemic therapies, this approach might be suitable for patients with a low-burden disease and/or those who wish to avoid treatment-related toxicities. Specific regimens include megestrol acetate alternating with tamoxifen, single-agent progestins, and single-agent tamoxifen^158–160. Single-agent aromatase inhibitors (such as letrozole or anastrozole) might have limited activity, although synergistic effects have been reported when these are combined with other agents^161–163. Combining letrozole with the CDK4/6 inhibitor abemaciclib, resulted in an ORR of 30% in patients with ER⁺ EC predominantly of an endometrioid histology¹⁶³. Other letrozole and CDK4/6 inhibitor combinations (such as ribociclib or palbociclib) have also been tested for this purpose^164,165. An interim results from the phase II NSGO-PALEO/ENGOT-EN3 trial indicated improved median PFS (8.3 months versus 3 months) and disease control (64% versus 38%) with palbociclib–letrozole compared with placebo–letrozole in patients with recurrent ER⁺ EC¹⁶⁵. Similarly, letrozole plus the mTOR inhibitor everolimus has demonstrated an ORR of 32% in patients with recurrent EC of endometrioid and non-endometrioid histological subtypes¹⁶². The ORR increased to 37.5% among those with endometrioid ECs¹⁶². Given this evidence of synergy, everolimus plus letrozole with or without ribicoclib is being evaluated in an ongoing trial enrolling patients with advanced-stage or recurrent EC with an endometrioid or mixed endometrioid histology (NCT03008408).

Conclusions

Advances in our understanding of EC have provided valuable insight and demonstrated that ECs are a highly heterogeneous group of cancers, beyond the dualistic histology model. Detailed characterization of the molecular features of EC has enabled a more-refined approach to prognostication, as well as individualized treatment with conventional and novel therapeutic approaches. However, with the rising prevalence and mortality among patients, a greater understanding of the biology of EC, and especially that of biomarkers related to response to specific therapies, is essential. A plethora of exciting trials testing novel biomarker-based strategies are ongoing and this list will continue to grow. Future research should seek to unravel the underlying causes of racial disparities in incidence and mortality and to improve the diversity of patients enrolling in clinical trials. For both early stage and advanced-stage cancers, future studies should investigate the incorporation of ctDNA and/or other molecular markers (such as p53 status or POLE mutations) into treatment algorithms to guide the escalation or de-escalation of first-line therapy. Greater attention should also be dedicated to the management of rare EC subtypes (such as carcinosarcoma and mesonephric-like adenocarcinoma) ideally with expansion of the inclusion criteria of prospective studies. Finally, with greater use of ICIs in the first-line and recurrent disease settings, dedicated translational research efforts are needed to decipher mechanisms of innate and/or adaptive resistance and to determine the most-effective subsequent therapy combinations following disease progression on ICIs.

Key points.

• With both the prevalence and mortality of endometrial cancer (EC) increasing, interventions should focus on cancer prevention and management strategies to improve oncological outcomes while reducing the effects of racial disparities.

• The molecular characterization of EC has greatly improved our understanding of the tumour biology as well as highlighting the extent of tumour heterogeneity, resulting in better prognostication as well as the opportunity for treatment personalization.

• The landscape of adjuvant therapy is complex, although molecular profiling and the use of circulating tumour DNA both have the potential to refine treatment strategies and thus reduce the extent of both undertreatment and overtreatment.

• Immune-checkpoint inhibitors (ICIs) provide profound levels of benefit for patients with mismatch repair (MMR)-deficient tumours and should be combined with chemotherapy as first-line therapy for patients with advanced-stage and/or recurrent disease.

• Investigations to identify key biomarkers and other adjunctive therapies in order to maximize the level of benefit of ICIs in patients with MMR-proficient tumours are ongoing and results are eagerly anticipated.

• Antibody–drug conjugates (such as trastuzumab deruxtecan) have demonstrated considerable potential in patients with previously treated advanced-stage and/or recurrent EC and should be further investigated.