Improving response to progestin treatment of low-grade endometrial cancer

Abstract

Objectives:

This review will examine how response rates to progestin treatment of low-grade endometrial cancer can be improved. In addition to providing a brief overview of the pathogenesis of low-grade endometrial cancer, we discuss limitations in the current classification of endometrial cancer and how stratification may be refined using molecular markers to reproducibly identify ‘low-risk’ cancers which may represent the best candidates for progestin therapy. We also discuss constraints in current approaches to progestin treatment of low-grade endometrial cancer and perform a systematic review of predictive biomarkers.

Methods:

PubMed, ClinicalTrials.gov and Cochrane Library were searched for studies reporting pre-treatment biomarkers associated with outcome in women with low-grade endometrial cancer or endometrial hyperplasia with an intact uterus who received progestin treatment. Studies of fewer than 50 women were excluded. The study protocol was registered in PROSPERO (ID 152374). A descriptive synthesis of pre-treatment predictive biomarkers reported in the included studies was conducted.

Results:

Of 1,908 records reviewed, 19 studies were included. Clinical features such as age or body mass index (BMI) cannot predict progestin response. Lesions defined as ‘low-risk’ by FIGO criteria (stage 1A, grade 1) can respond well, however the reproducibility and prognostic ability of the current histopathological classification system is sub-optimal. Molecular markers can be reproducibly assessed, have been validated as prognostic biomarkers and may inform patient selection for progestin treatment. POLE-ultramutated tumors and a subset of p53 wild-type or MMR-deficient tumors with ‘low-risk’ features (eg. progesterone and estrogen receptor-positive) may have improved response rates, though this needs to be validated.

Discussion:

Molecular markers can identify cases which may be candidates for progestin treatment. More work is needed to validate these biomarkers and potentially identify new ones. Predictive biomarkers are anticipated to inform future research into progestin treatment of low-grade endometrial cancer and ultimately improve patient outcomes.

Précis

Biomarkers can be used to identify women with low-grade endometrial cancer who may benefit from progestin treatment.

Introduction

Endometrial cancer is the most common gynecological cancer, and the fourth most common cancer among women in Western countries. There are approximately 382,000 new cases and 90,000 deaths annually worldwide ⁽¹⁾. Caucasian women have the highest incidence rates of endometrial cancer, though the majority of these tumors are low-grade and these patients generally have a favorable prognosis. Conversely, African-American women have the highest incidence rates of advanced disease with poorer survival ⁽²⁾. At least 41% of endometrial cancers have been attributed to obesity (BMI >30 kg/m²), with each 5 kg/m² increase in BMI being associated with a 62% increase in risk of endometrial cancer ⁽³⁾. Conversely, sustained weight loss reduces this risk ^(4–8).

Standard of care intervention for women with endometrial cancer involves a hysterectomy and bilateral salpingo-oophorectomy with or without surgical staging, as well as lymph node sampling and additional biopsies, although node dissection is not pursued for low-grade tumors in some areas of the world. Surgery is generally effective, however obesity increases the risk of surgical complications and patients often have concomitant comorbidities contributing to their perioperative risk ^(9–13). Reassessing therapeutic options in the increasingly common situation of medically complex, morbidly obese patients with endometrial cancer ⁽¹⁴⁾ and identifying conservative treatment options for these patients has been designated a research priority ⁽¹⁵⁾. Hysterectomy also results in irrevocable loss of fertility in young women who may wish to retain childbearing capacity. The estimated proportion of new cases of endometrial cancer in premenopausal women in 2018 varies worldwide, ranging from approximately 10% of all cases of endometrial cancer in North America, Europe and Oceania, to 20% in Africa and Latin America and 28% in Asia ⁽¹⁶⁾.

Progestins have been tested as a treatment option mostly in case series of women with low-grade endometrial cancer or hyperplasia who are high-risk surgical candidates due to obesity and/or medical comorbidities, or those who wish to retain fertility. To date, different types, doses and duration of progestins have been used, furthermore the patient selection process was often ad hoc. Meta-analyses indicate that 72–76% of tumors respond to progestins and 20–41% recur after an initial complete response ^{(17, 18)}. Reproducible stratification of tumors and biomarkers of progestin response are urgently required to identify tumors with intrinsic or emergent progestin resistance. Women who are unlikely to respond to progestins should have surgery and/or radiotherapy. This cohort also provides an opportunity to evaluate agents which might be employed to overcome endocrine therapy resistance. Identifying which patients will or will not benefit from progestin-based therapy was raised as one of the top ten unanswered research questions in a consensus engagement of endometrial cancer survivors, physicians and researchers ⁽¹⁹⁾.

This review will examine how response rates to progestin treatment of low-grade endometrial cancer may be improved. We will discuss how molecular markers can be used to reproducibly identify ‘low-risk’ tumors which may represent the best candidates for progestin treatment and perform a systematic review of pre-treatment biomarkers associated with progestin response.

Pathogenesis of low-grade endometrial cancer

The single biggest risk factors for endometrial cancer are obesity and metabolic dysfunction ^{(3, 20)}. In young women with endometrial cancer, 49–58% are obese and 8–18% have Lynch syndrome, another known risk factor for endometrial cancer ^(21–24). Young women are also frequently nulliparous and anovulatory and their tumors are typically considered to be in a hyperestrogenic state.

Obesity is particularly associated with low-grade endometrial cancer ^{(25, 26)}, however the mechanisms underlying this are poorly understood. A report from the International Agency for Research on Cancer concluded that there was strong evidence for sex hormone metabolism and chronic inflammation mediating the relationship between obesity and cancer, and the evidence for insulin and insulin-growth factor (IGF) signaling was moderate ⁽²⁶⁾. Non-steroidal anti-inflammatory drugs have been associated with a reduced risk of endometrial cancer, particularly in obese women, implying a causative role for inflammation in obesity-related endometrial cancer ^(27–29).

Endometrial hyperplasia is a common precursor of low-grade endometrial cancer and typically arises from chronic unopposed estrogen signaling. While hyperplasia without atypia is considered benign with a low risk of proceeding to carcinoma (RR 1.01–1.03), hyperplasia with atypia (also known as Endometrial Intraepithelial Neoplasia; EIN) has a high risk of proceeding to carcinoma (RR 14–45) ⁽³⁰⁾. Numerous driver mutations have been identified, the most frequently mutated genes in low-grade endometrial cancer are PTEN (phosphatase and tensin homolog), PIK3CA (phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha), CTNNB1 (catenin beta 1), ARID1A (AT-rich interaction domain 1A) and PIK3R1 (phosphoinositide-3-kinase regulatory subunit 1) ⁽³¹⁾. Mutations in PTEN are found in the majority of low-grade endometrial tumors as well as in premalignant lesions, leading to the assumption that they are an initiating event in tumorigenesis ^{(32, 33)}. Mutations in CTNNB1 exon 3 are particularly prevalent in young, obese women without Lynch syndrome ⁽³⁴⁾; however, the mechanism of action of these mutations is poorly understood.

Classification of endometrial cancer and its limitations

Endometrial cancers are classified according to histopathologic assessment of tumor type and grade, as well as surgical staging according to the International Federation of Gynecology and Obstetrics (FIGO) criteria ⁽³⁵⁾. Tumors that are stage I, grade 1 or 2 with no or superficial myometrial invasion are deemed ‘low-risk’ and are not routinely offered adjuvant therapy. Approximately 5% of all recurrences occur in these patients ⁽³⁶⁾, highlighting the need to reproducibly identify ‘low-risk’ tumors.

It is now recognized that the current pathological classification and grading system of endometrial carcinomas are limited in both reproducibility and prognostic ability. Lack of consensus on histologic subtype diagnosis is seen in at least one-third of cases ^(37–39). Furthermore, only a modest correlation between preoperative endometrial sampling and final pathology grading is seen with grade being upgraded in 15–20% of cases and high-risk pathology being identified in 19–29% of cases on final pathology ^(40–43). A new binary grading system that discriminates between low (grade 1–2) and high-grade (grade 3) tumors has been proposed which has superior prognostic significance for survival and greater interobserver reproducibility than current FIGO criteria ^(44–46). However, this may not be appropriate in a conservative therapeutic approach as only grade 1 tumors are generally considered suitable ⁽⁴⁷⁾.

In early-stage endometrial cancer, the European Society for Medical Oncology-modified classification, which includes uterine factors such as histological subtype, grade, myometrial invasion and lympho-vascular space invasion, has been demonstrated to have the highest power of discrimination for stratifying the risk of recurrence or nodal metastases, however it does not show high accuracy with a concordance index of only 0.73 ⁽⁴⁸⁾.

More recently, The Cancer Genome Atlas (TCGA) classified endometrial cancers into four prognostically distinct subtypes based on genomic features ⁽³¹⁾. Subsequently, other research teams sought to recapitulate these molecular subtypes using clinically-applicable methods on standard formalin-fixed paraffin-embedded material. POLE (DNA polymerase epsilon)-ultramutated tumors are associated with excellent prognosis, followed by p53 wild-type (also referred to as No Specific Molecular Profile; NSMP) and DNA mismatch repair (MMR)-deficient tumors with intermediate prognosis. p53-abnormal tumors have the worst prognosis ^(49–52). In young women (<50 years of age), p53 wild-type/NSMP tumors are the most frequent (64% of cases), followed by MMR-deficient (19%) and POLE-ultramutated (13%) tumors. p53-abnormal tumors are the least frequent (4%). The majority of obese women (82%) also have p53 wild-type/NSMP tumors ⁽⁵³⁾.

Approximately 3% of endometrial tumors have more than one of these four molecular features suggesting they are currently unclassifiable. Preliminary studies suggest that the POLE-ultramutated phenotype predominates in tumors with pathogenic POLE exonuclease domain mutations that are also p53-abnormal or MMR-deficient, and the MMR-deficient phenotype predominates in MMR-deficient tumors that are also p53-abnormal or have non-pathogenic POLE mutations, although these findings remain to be validated and standardized criteria developed for interpreting POLE variants ^{(49, 50, 54–56)}.

Marked inter-tumor and intra-tumor molecular heterogeneity have been reported in low-grade endometrial tumors ^(57–59). Intra-tumor heterogeneity may vary between molecular markers as one study reported >95% concordance between three tumor blocks for POLE and CTNNB1 mutation status and MMR protein expression, whilst concordance for p53 and L1CAM (L1 cell adhesion molecule) protein expression was 91–94%, supporting the use of select biomarkers in clinical decision-making ⁽⁶⁰⁾. Refinement of molecular classifiers that can reproducibly be assessed on diagnostic specimens is thus required to identify tumors that are ‘low-risk’ and may safely be managed conservatively. From a practical point of view, the assessment of molecular markers such as POLE mutation testing and immunohistochemistry for MMR proteins and p53 are not currently feasible in all facilities, spurring the need for the development of low-cost technologies that can easily be implemented within existing diagnostic workflows.

Molecular markers of ‘low-risk’ endometrial cancer

Improved endometrial cancer stratification is necessary to enable study of treatment efficacy within biologically similar tumors, ultimately improving patient outcomes. There is now increasing evidence that molecular markers will help achieve this, providing reproducible categorization, prognostic information and suggestion of predictive biomarkers for both conventional and targeted therapies. For example, women with MMR-deficient endometrial tumors have improved disease-specific survival after adjuvant radiotherapy compared to women with MMR-proficient tumors ⁽⁶¹⁾. MMR-deficiency also predicts clinical benefit of immune checkpoint blockade ^{(62, 63)}.

Progestins can be offered to women with low-grade tumors, although as discussed earlier, reproducible identification of these tumors can be problematic. Stratification using molecular markers, possibly in combination with histopathological features, is predicted to reproducibly identify ‘low-risk’ tumors that may represent women who will benefit from progestins. Low-grade endometrioid tumors are largely p53 wild-type/NSMP (60%), although some are MMR-deficient (29%) and a minority are POLE-ultramutated (6%) or p53-abnormal (5%) ⁽³¹⁾. Molecular features thus do not entirely correlate with grade. It has been postulated that FIGO grading is most appropriate in p53 wild-type/NSMP and MMR-deficient tumors, as these mostly correspond to endometrioid subtype ⁽⁶⁴⁾. Molecular markers could also be used to refine stratification of these subtypes in order to reproducibly identify ‘low-risk’ tumors.

Both estrogen (ER) and progesterone (PR) receptors have been recognized as independent prognostic biomarkers in early-stage endometrial cancer for many decades ^{(65, 66)}. ER is generally expressed in p53 wild-type/NSMP, POLE-ultramutated and MMR-deficient tumors, whilst PR expression is increased only in p53 wild-type/NSMP tumors ^{(31, 67)}. Within p53 wild-type/NSMP tumors, CCND1 (cyclin D1) C-terminal mutation, CTNNB1 exon 3 mutation, 1q32.1 amplification, L1CAM overexpression, loss of ER and PR and high DNA damage have all been identified as poor prognostic markers ^{(34, 54, 68–72)}, indicating that further molecular stratification within this subtype is possible.

Although MMR-deficient tumors represent a significant proportion of low-grade endometrioid tumors, they have clinical features associated with poor outcomes ^{(49–51, 53, 73)}. A recent study of stage 1, grade 1 endometrioid tumors indicated that MMR-deficiency was associated with increased risk of recurrence ⁽⁷⁴⁾, questioning whether this subtype can be considered ‘low-risk’ and therefore may not benefit from progestin therapy. Further stratification within MMR-deficient tumors could potentially be applied as tumors with CCND1 C-terminal mutation ⁽⁶⁹⁾ and methylated PTEN ⁽⁷⁵⁾ have been associated with worse prognosis. Furthermore, up to one-quarter of young women with Lynch syndrome who have endometrial cancer have synchronous ovarian cancer ⁽²⁴⁾, suggesting that women with MMR-deficient tumors, and particularly those with Lynch syndrome, require careful evaluation by both molecular and imaging methods for improved risk assignment and may require close monitoring if offered progestin therapy.

The excellent prognosis of POLE-ultramutated tumors appears to be irrespective of adjuvant treatment ^{(76, 77)}, suggesting that early-stage POLE-ultramutated tumors could benefit from conservative management ⁽⁶⁷⁾. Conversely, p53-abnormal tumors have the worst prognosis and low-grade tumors with overexpression of p53 have increased risk of relapse and decreased survival ^{(78, 79)}, suggesting that women with these tumors should not be offered conservative treatment. However, it should not be excluded that TP53 variants may be passenger events, as evidenced by subclonal p53 overexpression in tumors with concomitant pathogenic POLE exonuclease domain mutations or MMR-deficiency, as was discussed earlier ⁽⁵⁵⁾.

It thus appears that three molecular subtypes potentially represent tumors that are ‘low-risk’: 1) POLE-ultramutated tumors; 2) p53 wild-type/NSMP tumors with wild-type CCND1 and CTNNB1, are ER and PR-positive, lack 1q32.1 amplification and with low L1CAM expression and DNA damage; and 3) MMR-deficient tumors with wild-type CCND1, are ER- and PR-positive, lack PTEN methylation and without Lynch syndrome (Figure 1). Further studies are required to validate these molecular markers of ‘low-risk’ endometrial cancer, compare them to conventional criteria for risk assignment in terms of both patient outcomes and cost-effectiveness, and evaluate whether they represent the best candidates for conservative therapy and specifically progestin treatment.

Figure 1.

Evolution of the classification of endometrial cancer. Since TCGA classified endometrial cancers into four prognostically distinct molecular subtypes in 2013, stratification of tumors into risk groups using molecular markers has been and continues to be improved.

*NSMP = No Specific Molecular Profile.

A risk prediction model that identifies individuals at high risk of endometrial cancer was recently proposed ⁽⁸⁰⁾. The model is based on genetic, insulin, reproductive and obesity risk scores. Inflammation is not currently directly incorporated as it is not known which inflammatory factors should be assessed. The model remains to be validated but could potentially be adapted to identify ‘low-risk’ cases of endometrial cancer that could benefit from conservative treatment. Furthermore, a recent study concluded that L1CAM <1% and nuclear PR >85% assessed by immunohistochemistry on presurgical samples and myometrial invasion <50% correctly determined ‘low-risk’ patients in 80% (56/70) of cases ⁽⁸¹⁾, highlighting the need to combine clinical and molecular features in diagnostics.

L1CAM overexpression has been demonstrated to be an independent poor prognostic marker ^(82–84), others include overexpression of HER-2/neu (human epidermal growth factor receptor 2) ⁽⁸⁵⁾, STMN1 (stathmin 1) ⁽⁸⁶⁾, CD133 ⁽⁸⁷⁾ or MCT1 (monocarboxylate transporter 1) ⁽⁸⁸⁾; loss of ASRGL1 (asparaginase and isoaspartyl peptidase 1) ^{(89, 90)} or E-cadherin ⁽⁹¹⁾; aneuploidy ⁽⁹²⁾ or few intraepithelial CD8⁺ T lymphocytes at the invasive border ⁽⁹³⁾. Blood-based biomarkers such as CA-125 (cancer antigen 125), CA 15–3 (cancer antigen 15–3), HE4 (human epididymis protein 4) and more recently, metabolites and steroids, have also been reported to identify endometrial cancers at high risk of recurrence ^(94–99). High visceral fat percentage, as quantified by computed tomography, has also been associated with poor outcome in endometrial cancer ^{(97, 100)}. Finally, genetic polymorphisms, notably the G allele in rs13222385 in EGFR (epidermal growth factor receptor), have also been associated with worse overall survival ⁽¹⁰¹⁾. The prevalence of these markers and their utility in stratification within the four prognostic molecular subtypes described earlier remain to be assessed.

Progestin treatment of endometrial cancer

The progestins megestrol acetate and medroxyprogesterone acetate are approved by the US Food and Drug Administration as adjunctive or palliative treatment of advanced, recurrent or metastatic endometrial cancer. Various randomized and non-randomized clinical trials have offered progestins to young women with low-grade, early-stage disease who desire to retain childbearing capacity, as well as obese women and women with comorbidities at high risk of surgical complications. For young women who are successfully managed with progestins, subsequent pregnancy is not uncommon (12–83% live birth rate) though assisted reproduction technology is advised to maximize chances of a live birth ^{(18, 102–106)} and hysterectomy is often recommended once childbearing has been completed ^{(47, 107)}. Most studies completed to date used the oral progestins megestrol acetate or medroxyprogesterone acetate at various doses, whilst intrauterine progestins are now increasingly utilized, sometimes in combination with oral progestins, though treatment duration varies. It has been reported that intrauterine progestins achieve a higher rate of pathological complete response than oral progestins ^{(17, 108)}, possibly due to improved patient compliance and increased progestin concentration in the endometrium ⁽¹⁰⁹⁾.

Meta-analyses have indicated that in women with early-stage endometrial cancer, progestins are associated with a 72% to 76% response rate; however, 20% to 41% of patients relapse after having developed a complete pathological response ^{(17, 18)}. The age range in these meta-analyses varies considerably, including women up to 88 years of age, although the mean age was under 40 years. A meta-analysis including only studies with women under 44 years of age with atypical hyperplasia (EIN) or early-stage endometrial cancer who desired fertility, reported that remission reached a plateau of approximately 80% 12 months after commencing treatment; however, recurrence probability increased continually with time, being 17% at 12 months and 29% at 24 months ⁽¹¹⁰⁾. Prospective studies of Asian women under 40 years of age with early-stage endometrial cancer, most of whom were nulliparous, have reported much lower response rates after six months treatment. A Japanese study of 45 women reported a complete response rate of 55% and a recurrence rate of 57% with oral progestins and low-dose aspirin ⁽¹⁰³⁾, whilst a recent Korean study of 35 women reported a complete response rate of only 37% with combined oral and intrauterine progestins ⁽¹¹¹⁾. These studies raise the question of whether ethnicity affects response to progestins. Asian women present younger at diagnosis and with higher stage disease than Caucasians, suggesting differences in risk factors such as obesity ⁽¹¹²⁾. Asian women reportedly have a higher body fat percentage with greater abdominal adiposity and higher rates of metabolic syndrome than Caucasian women ⁽¹¹³⁾. The Cancer Genome Atlas (TCGA) data indicated tumors from Asian women have an increased mutation load and frequency of somatic MMR mutations versus tumors from Caucasian women ⁽¹¹⁴⁾. It should be noted that there were only 20 tumors from Asian women in TCGA, highlighting the need for more extensive molecular and clinical profiling of tumors from non-Caucasian women to better understand potential confounding factors.

Current prospective trials exploring progestin treatment of low-grade endometrial cancer are reviewed in Supplementary File 1. Inclusion criteria are based on clinicopathological features with three trials limiting inclusion to PR-positive tumors (NCT02990728, NCT03463252 and NCT03538704). Only one trial involves a follow-up time exceeding 36 months (NCT02397083), limiting the ability to comprehensively assess women whose tumors may recur. Three trials have a formal aim of identifying predictive biomarkers (NCT01686126, NCT02990728 and NCT03567655), two of which include the addition of either weight loss or metformin to intrauterine progestin, either of which are proposed to increase pathological complete response. Sustained weight loss, either by surgical ^{(7, 115–118)} or non-surgical methods ^(4–6), is associated with reduced risk for endometrial cancer, highlighting that the relationship between obesity and endometrial cancer is reversible. Multiple meta-analyses have indicated that prior metformin use is associated with improved survival in endometrial cancer patients ^(119–122).

Current guidelines stipulate that conservative treatment of endometrial cancer should only be considered in women desiring to retain fertility and patients should be counselled for hysterectomy as definitive treatment once childbearing has been completed, or those with persistent or progressive disease. The National Comprehensive Cancer Network ⁽⁴⁷⁾ and European Society of Gynaecological Oncology (ESGO) ⁽¹⁰⁷⁾ guidelines both state that stage IA, grade 1 endometrioid adenocarcinomas can be considered for fertility-sparing treatment. Formal dilatation and curettage (D&C) instead of pipelle biopsy is the preferred method to obtain histology, demonstrating a higher correlation with the final histological results, and specimens should be examined by at least one pathologist. Pelvic MRI scan is the preferred method to establish myometrial invasion, though transvaginal ultrasound scan can be used if MRI is contraindicated or not available. ESGO guidelines also stipulate that hysteroscopy may be performed in combination with D&C and there is no need to routinely assess PR status, although the authors acknowledged that the recommendations should be interpreted with caution due to the lack of prospective high-quality studies ⁽¹⁰⁷⁾. A recent survey of European clinicians indicated that, despite the majority believing that grade 1 endometrial cancer without myometrial invasion could be offered progestins, most centers treated few patients conservatively. There was no consensus on whether PR status should be examined prior to commencing conservative treatment, or whether patients with Lynch syndrome could be considered ⁽¹²³⁾, highlighting the need for predictive biomarkers that are validated in large, prospective studies.

Systematic review of biomarkers of progestin response

The objective of this systematic review was to identify pre-treatment biomarkers of progestin response in low-grade endometrial cancer. Endometrial hyperplasia was also included as it is a precursor lesion and many studies include both endometrial cancer and hyperplasia. Previous systematic reviews of predictive biomarkers have only assessed immunohistochemical markers and included all studies regardless of participant numbers, resulting in the inclusion of some very small sample sizes ^(124–126). We sought to provide an assessment of clinical, histopathological and molecular markers associated with progestin response in larger studies (≥50 women) in order to focus on predictive biomarkers with higher quality evidence for future validation. The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO ID 152374).

Sources:

PubMed, ClinicalTrials.gov and Cochrane Library were searched for studies reporting pre-treatment biomarkers of progestin response in women with low-grade endometrial cancer or endometrial hyperplasia and with an intact uterus. Search terms included: “endometrial cancer”, “endometrioid adenocarcinoma”, “uterine cancer”, “uterine adenocarcinoma” or “endometrial hyperplasia” AND “progest*”, “levonorgestrel”, “LNG”, “IUD”, “MPA”, “medroxyprogesterone”, “megestrol” or “gestagen” AND “predictive”, “*marker” or “response”. All studies published in English until 1^st October 2019 were included.

Study Selection:

Titles and/or abstracts were retrieved and screened against the inclusion and exclusion criteria. Full-text articles of potentially eligible studies were assessed. Additional studies manually curated were also considered. Only studies assessing pre-treatment biomarkers associated with outcome in women with low-grade endometrial cancer or hyperplasia treated with progestins were included. Studies had to include at least 50 women. Progestin treatment could be of any type, dose or duration and could be administered in combination with another form of conservative therapy. Treatment outcomes were evaluated as disease regression or recurrence. Studies reporting predictive biomarkers in advanced or recurrent endometrial cancer or women without an intact uterus were excluded. Reviews, editorials, commentaries and conference abstracts were also excluded. Risk of bias was not assessed. For each included study, data extracted included the study type, population, treatment, outcome and biomarker assessed. A descriptive synthesis of predictive biomarkers reported in the included studies was conducted.

Results:

A total of 1,908 unique records were reviewed and 19 studies were included (Figure 2). Details of all the included studies can be seen in Supplementary File 2. 12 of these studies were retrospective and 7 were prospective. Age, BMI, ethnicity, menopause status, progestin type, dose and duration as well as outcome measured varied between studies.

Figure 2.

Flow diagram outlining study selection.

Reports on clinical factors associated with outcome are conflicting (Table 1). Many studies investigating BMI reported that obesity was associated with failure to achieve disease regression and increased recurrence ^{(105, 127–130)}. However, a recent study of Japanese women reported that lower BMI was associated with increased recurrence ⁽¹³¹⁾, whilst numerous studies have reported no association between BMI and outcome ^(132–137). The association between age or menopause status and outcome are also conflicting, with two studies reporting younger age or premenopausal status were associated with disease regression or reduced recurrence ^{(127, 132)}, whilst another reported younger age was associated with increased recurrence ⁽¹³⁶⁾. Multiple studies have reported no association between age or menopause status and outcome ^{(105, 128–131, 133–135, 137)}. A thinner endometrium has been associated with disease regression or reduced recurrence in one study each of women with endometrial hyperplasia ^{(128, 134)}. Diabetes has been associated with increased recurrence in one study ⁽¹²⁸⁾ but not in other studies ^{(127, 129, 130, 137)}. Numerous other clinical factors including gravidity ^{(127, 129, 134)}, parity ^{(105, 127–130, 134–137)}, polycystic ovarian syndrome ^{(105, 127, 129, 131)}, smoking ^{(129, 133, 134)}, family history of cancer ^{(127, 133)} and hypertension ^{(128, 130, 137)} have been investigated in multiple studies, but none has shown an association with outcome.

Table 1.

Pre-treatment clinical features investigated for their association with disease regression and/or recurrence.

Clinical feature	Regression	Recurrence	No association
BMI	Non-obese (105, 127, 128, 130)	Non-obese (131) Obese (105, 127–129)	(132–137)
Age	Younger (132)	Younger (136) Older (127)	(105, 128–131, 133–135, 137)
Menopause status		Premenopausal (136)	(128)
Endometrial thickness	Thin (134)	Thick (128)
Diabetic status		Diabetic (128)	(127, 129, 130, 137)
Gravidity			(127, 129, 134)
Parity			(105, 127–130, 134–137)
Polycystic ovarian syndrome			(105, 127, 129, 131)
Smoking			(129, 133, 134)
Family history of cancer			(127, 133)
Hypertension			(128, 130, 137)

Studies on histopathological features as predictors of progestin response are generally in agreement with each other (Table 2). Lower nuclear or histological grade have been associated with improved histological response or survival respectively ^{(138, 139)}. Lesion type has been associated with disease regression and reduced recurrence as hyperplastic lesions without atypia have improved outcome compared to hyperplastic lesions with atypia (EIN), which in turn have improved outcome compared to cancer ^{(128, 129, 132, 136)}. However, numerous studies have reported similar outcomes between lesion types ^{(127, 131, 133, 135)}. Low mitotic index and tumor volume have also been associated with improved histological response and survival respectively in one study each ^{(138, 139)}.

Table 2.

Pre-treatment histopathological features investigated for their association with disease regression and/or recurrence.

Histopathological feature	Regression	Recurrence	No association
Grade	Low (138)		(139)
Lesion type	Hyperplastic (128, 132)	Cancer (128, 129, 136)	(127, 131, 133, 135)
Tumor volume	Low (138)
Mitotic index	Low (139)
Nuclear grade	Low (139)

PR is the most studied molecular marker associated with progestin response (Table 3). Multiple studies have shown that PR expression is associated with disease regression, though PR-negative lesions can benefit from progestins ^{(129, 137, 139, 140)}. Isoform-specific studies are conflicting: high PRβ has been associated with disease regression in one study ⁽¹⁴¹⁾, whilst other studies have reported no association with outcome ^{(139, 142)}. PRα has not been associated with disease regression in any study ^{(141, 142)}. PR location has also not been associated with disease regression ⁽¹⁴²⁾, but low stromal PRα and high glandular PRβ have been associated with increased recurrence ^{(135, 136)}.

Table 3.

Pre-treatment molecular markers investigated for their association with disease regression and/or recurrence.

Molecular marker	Regression	Recurrence	No association
PR	High (137, 140)		(129, 139)
PRβ	High (141)		(139, 142)
Glandular PRβ		High (135, 136)	(142)
Stromal PRβ			(142)
Glandular PRα		Low (136)	(142)
Stromal PRα		Low (135, 136)	(142)
Glandular PRα:PRβ		≤1 (136)
Stromal PRα:PRβ		≤1 (136)
ERα	High (140)		(129, 135, 139, 142)
ERβ			(135, 142)
MMR status	Proficient (132)
HSPA5/GRP78	Low (143)
Ki67	Low (139)		(129)
p53	Low (137)
AR			(142)
BAX			(135)
BCL2			(135, 137, 139–141)
Cleaved caspase			(139)
COX2			(140)
MLH1			(140)
PAX2			(135, 141, 144)
PTEN			(135, 141, 144)
CA-125			(134)
Estradiol			(136)

ER expression has also been associated with disease regression, though similar to PR, ER-negative lesions can benefit from progestins ^{(129, 139, 140, 142)}.

Conversely, biomarkers of resistance to progestin treatment are relatively understudied with small numbers of cases. MMR-deficient lesions have been associated with failure to achieve disease regression in one study ⁽¹³²⁾. Overexpression of HSPA5/GRP78 (heat shock protein family A member 5) ⁽¹⁴³⁾ and p53 ⁽¹³⁷⁾ have also been associated with failure to achieve disease regression in one study each of women with endometrial hyperplasia. One study also reported that high Ki67 was associated with failure to achieve disease regression ⁽¹³⁹⁾, though another study reported no association with outcome ⁽¹²⁹⁾.

Other molecular markers that have no association with outcome are AR (androgen receptor) ⁽¹⁴²⁾, BAX (BCL2 associated X) ⁽¹³⁵⁾, BCL2 (B-cell lymphoma 2) ^{(135, 137, 139–141)}, cleaved caspase ⁽¹³⁹⁾, COX2 (cytochrome c oxidase subunit II) ⁽¹⁴⁰⁾, MLH1 (mutL homolog 1) ⁽¹⁴⁰⁾, PAX2 (paired box 2) and PTEN ^{(135, 141, 144)}. Finally, only two studies have investigated blood-based biomarkers and neither levels of CA-125 ⁽¹³⁴⁾ nor estradiol ⁽¹³⁶⁾ were associated with outcome.

Discussion:

Multiple factors have been investigated as potential markers of progestin response in endometrial hyperplasia and low-grade endometrial cancer. Many of the studies conducted include small sample sizes with either few cases or numbers of non-responders, potentially resulting in biased conclusions. Systematic reviews of predictive biomarkers conducted to date have only assessed immunohistochemical markers and included all studies regardless of participant numbers ^(124–126). We included all predictive biomarkers in this systematic review regardless of how they were assessed, but were more selective by only including studies with a minimum of 50 women. The reason for this was to focus on markers with higher quality evidence for future validation, though this did result in the exclusion of multiple studies which either explored novel predictive biomarkers or provided further evidence supporting biomarkers reviewed here (predominantly PR). Many studies include both endometrial cancer and precursor lesions, and the inability to separate between lesion types is a limitation of this study.

Reports on clinical factors associated with progestin response are conflicting. More studies have reported the lack of an association between BMI and outcome ^(132–137) than the number of studies that have reported an association ^{(105, 127–130)}, with one conflicting study ⁽¹³¹⁾. Similarly, for age or menopause status, more studies have reported the lack of an association with outcome ^{(105, 128–131, 133–135, 137)} than the number of studies that have reported an association and even then, results are conflicting ^{(127, 132, 136)}. A thinner endometrium has been associated with disease regression and decreased recurrence in one study each ^{(128, 134)}, however the cut-off values for assessing endometrial thickness used in either study varied. Diabetes has been associated with increased recurrence in one study ⁽¹²⁸⁾ but not in other studies ^{(127, 129, 130, 137)}. Therefore, there do not appear to be any clinical factors that could be used to select women who could benefit from progestin treatment.

Reports on histopathological features associated with progestin response are relatively consistent with less aggressive, lower grade lesions being more likely to respond. Whether hyperplastic lesions, either with (EIN) or without atypia, have improved outcomes to cancer is conflicting, indicating that lesion type is not a basis for offering progestin treatment.

Numerous predictive molecular markers have been proposed and ER, and especially PR, are the most reported to date, though most studies have only been conducted in women with endometrial hyperplasia. There are numerous sources of evidence for PR being the best biomarker for progestin response to date, however it is not required for response as PR-negative lesions can benefit from progestins ^{(129, 137, 139, 140)}. A recent meta-analysis of immunohistochemical biomarkers for progestin response in women with endometrial hyperplasia or early endometrial cancer concluded that PR was a predictive biomarker only when intrauterine and not oral progestins were used, although the accuracy of intrauterine progestins was too low to be considered determining for clinical practice ⁽¹²⁴⁾. It should be noted that only two studies of intrauterine progestins were included in this meta-analysis. Large studies assessing PR isoforms are limited with one study indicating PRβ was associated with disease regression ⁽¹⁴¹⁾. A recent systematic review of immunohistochemical markers concluded that PRβ was the most promising predictive biomarker, however this was based on only two studies reporting a significant association, whilst a third study reported no association ⁽¹²⁶⁾. However, glandular PRβ, as well as PRα, have also been associated with increased recurrence ^{(135, 136)}. PRβ expression correlates with activated PR, which has been proposed to reflect active PR signaling ⁽¹⁴⁵⁾. The PR antagonist onapristone has demonstrated clinical benefit in recurrent or metastatic endometrial tumors expressing activated PR ⁽¹⁴⁶⁾, though whether activated PR is also a predictive biomarker for PR agonists remains to be seen. As most low-grade endometrioid tumors are PR-positive, the clinical utility of PR as a predictive biomarker needs to be validated; furthermore, the role of the activated form of the receptor as well as expression levels and location remain to be clarified. Studies in mice indicated that stromal PR was required for response to progestins ^{(147, 148)}, however this remains to be validated in humans ⁽¹⁴²⁾.

Expression of PTEN ^{(135, 141, 144)} has not been associated with outcome in multiple studies. A recent meta-analysis of seven studies, only two of which included at least 40 women, indicated that loss of PTEN had no significant impact on response to progestins, though the authors suggested that combined assessment of PTEN with other markers may be useful ⁽¹²⁵⁾. MMR-deficiency has been associated with failure to achieve disease regression in one study, however, this study only had six cases with abnormal MMR staining, three of which had germline MMR mutations. These women were older, had lower BMI and a higher incidence of endometrial cancer than women with tumors with normal MMR staining ⁽¹³²⁾. Overexpression of HSPA5/GRP78 ⁽¹⁴³⁾ or p53 ⁽¹³⁷⁾ have also been associated with failure to achieve disease regression in one study each, though cut-off values for either biomarker were not established. More studies with larger numbers of cases are needed to independently assess and validate these potential biomarkers of resistance to progestin therapy.

Current guidelines state that conservative management of endometrial cancer should only be considered in women with stage IA, grade 1 endometrioid adenocarcinomas who desire to retain fertility ^{(47, 107)}. However, progestins have also successfully been given to women at high risk of surgical complications due to obesity and/or comorbidities. Clinical and pathological phenotypes vary between these populations and establishing which women will respond to progestins is essential to improve patient outcomes and reduce healthcare costs. As discussed here, there is some evidence that molecular markers may assist in reproducibly identifying these women, though none have yet been validated. Of the four prognostic molecular subtypes described earlier, progestin therapy has been documented as conservative management in a subset of young women with predominantly p53 wild-type/NSMP tumors and a small proportion of MMR-deficient or POLE-ultramutated tumors, but not in p53-abnormal tumors; however the outcomes of these women in unclear due to missing data ⁽⁵³⁾. p53 wild-type/NSMP tumors are the most frequent subtype amongst young and obese women ⁽⁵³⁾ and are predicted to respond best to progestins ^{(31, 67)}; however, no study to date has assessed whether this molecular subtype has improved response rates. A small retrospective study in women <40 years of age undergoing hysteroscopic resection followed by progestin therapy indicated that 7/7 PR-positive grade 1 endometrioid tumors that were p53 wild-type/NSMP had complete response at 6 months; however, two women were subsequently diagnosed with ovarian cancer ⁽⁵⁹⁾. This same study also reported that 5/7 MMR-deficient tumors had complete response at 6 months; however, two women, both of whom had germline MMR mutations, were subsequently diagnosed with a second cancer. Two tumors were POLE-ultramutated, one of which had concomitant MMR-deficiency; only the tumor that was MMR-proficient had a complete response at 6 months and this woman continued to do well after 86 months follow-up. Finally, one tumor was p53-abnormal but it also had concomitant germline MMR-deficiency. Although this woman had a complete response at 6 months, she was subsequently diagnosed with a second cancer. Although this study by Falcone et al. ⁽⁵⁹⁾ was small, it supports the hypothesis that molecular subtypes could inform patient selection for progestin therapy, though further stratification is required to identify ‘low-risk’ tumors. Whether POLE-ultramutated tumors and a subset of p53 wild-type/NSMP or MMR-deficient tumors with ‘low-risk’ features (summarized in Figure 1), have improved response rates versus current histopathological selection methods needs to be assessed. Larger studies, including women with a range of ages and BMI and different ethnicities, are required to validate this hypothesis, as well as establish which markers further refine stratification to a level that both improves patient outcomes and is clinically feasible.

Taken together, these studies indicate that patients and lesions with certain features may exhibit the best progestin response and prognosis. Importantly, there are no clinical features associated with progestin response. Whilst reports on histopathological features associated with progestin response are relatively consistent, reproducibly classifying lesions is problematic as was discussed earlier. Molecular markers can be identified and have been validated as prognostic biomarkers, though none has been validated as a predictive biomarker for progestin response. PR is the most studied predictive biomarker to date; however its clinical utility remains to be validated and a standardized scoring system needs to be developed if it is to be implemented into clinical practice. Combined assessment of PR with other biomarkers may have improved predictive ability. The association between MMR status and progestin response is unclear with only a small number of cases studied to date, as is the importance of the mechanism of MMR-deficiency, though the International Society of Gynecological Pathologists has proposed that universal MMR testing be performed in young women desiring fertility-sparing treatment ⁽⁶⁴⁾. Whether germline MMR-deficient women should be excluded from receiving progestins or monitored more closely remains to be determined. What is clear from available evidence is that women with p53-abnormal tumors should be excluded from receiving conservative treatment, though concomitant pathogenic POLE exonuclease domain mutations or MMR-deficiency need to be excluded as TP53 variants occurring in these contexts are likely passenger and not driver mutations ⁽⁵⁵⁾.

The heterogeneity in the type, dose and duration of progestin used, study type, population, number of participants, outcomes measured and cut-off values used for hormone receptor expression in studies to date highlight the need for large, prospective trials with consistent parameters in order to provide high-quality evidence. Longer studies are also required in order to monitor recurrences and subsequent pregnancies and correlate these with pre-treatment biomarkers. Of note, successful pregnancy after progestin treatment has been associated with reduced recurrence in two studies with long-term follow-up ^{(105, 127)}. All assessments of molecular markers to date have been targeted, no study has performed an unbiased genome-wide assessment of the molecular features of endometrial lesions that do or do not respond to progestin therapy. Increasing the range of predictive biomarkers to include mutations in other genes, epigenetic modifications, gene and protein expression signatures and post-translational modifications is anticipated to identify novel predictive biomarkers as well as improve specificity and sensitivity. Ideally, separate analysis of both tumor and stroma would be conducted with the recognition that stromal factors may be predictive of response. Only two studies to date have investigated blood-based biomarkers (CA-125 and estradiol) and neither reported an association with outcome ^{(134, 136)}, though other circulatory factors remain to be assessed. Other factors such as fat localization remain to be assessed. Finally, integrating molecular biomarkers with clinical and histopathological features as well as quality-of-life assessments will provide a more comprehensive assessment, enabling clinicians to provide their patients with options on whether they can safely delay or avoid standard of care without adversely affecting their cancer-related outcomes or quality-of-life.

To date, only three prospective trials have a formal aim of identifying biomarkers of progestin response and although they are important resources of samples, patient numbers are clearly insufficient to validate the biomarkers proposed to date. Samples collected in other trials, such as those reviewed in Supplementary File 1, could potentially be aggregated into an international biobank to obtain the statistical power needed to validate and potentially identify new predictive biomarkers as endometrial biopsies are typically collected at baseline as part of standard of care. A drawback of all cohorts to date is the lack of a control arm of women not treated with progestin for comparison, though the ethics of including untreated women with a formal diagnosis of endometrial cancer is highly questionable. Most prospective studies also collect tumor and blood specimens every three months throughout treatment, potentially enabling a comprehensive picture to be established of how the molecular features of tumors and spectrum of circulating factors change as tumors respond or not to progestins, providing insights into tumorigenic processes and their reversibility. Understanding the mechanisms of the pathogenesis of low-grade endometrial cancer and its relationship with obesity will most likely result in the identification of novel targets for treatment as well as preventative strategies.

Conclusions

Molecular markers have been validated as prognostic factors in endometrial cancer in numerous studies, as well as predictive biomarkers for select treatments, paving the way for biomarkers to replace current histopathological grading and staging of tumors, reproducibly stratify tumors into risk groups and direct patients towards the optimal treatment strategy. However, there are currently no validated markers of response to progestin therapy, which would be of significant benefit to young women who wish to retain fertility as well as obese women and/or those with comorbidities who are at high risk of surgical complications. Key unanswered questions for women considering progestin therapy for their endometrial cancer are summarised in Table 4. To date, only three prospective studies include a formal outcome of identifying predictive biomarkers, however samples from other trials may be used for discovery and validation. Prospective clinical trials provide consistent progestin type, dose, duration, sampling times and assessment of response, enabling a high level of evidence-based recommendations to be generated. Predictive biomarkers are anticipated to improve response rates and guide further research into progestin treatment of endometrial cancer. Outcomes such as weight loss and subsequent pregnancy will also need to be considered in future trials as they can contribute to improved response and reduced recurrence respectively.

Table 4.

Key unanswered questions for women considering progestin treatment for their endometrial cancer.

Which tumors will have a complete pathological response? What are the optimum type, dose and duration of progestin treatment? What are the optimum duration and frequency of follow-up after achieving a pathological complete response? Should progestin treatment be continued after achieving a pathological complete response and if so, for how long? Should progestin treatment be continued after a partial or failed response and if so, for how long? What are the criteria for stopping progestin treatment? Is a hysterectomy necessary after completing childbearing? Should dual-agent therapy be administered (eg. metformin, weight loss, targeted therapy) and if so, which patients would benefit? Can MMR-deficient tumors due to germline MMR mutation(s) be treated similarly to tumors with somatic MMR modifications? How can emerging molecular data best be incorporated into patient management?

Supplementary Material

Supp1

Supplementary File 1. Current prospective trials exploring progestin treatment of low-grade endometrial cancer. LNG-IUD = levonorgestrel-releasing intrauterine device. MA = megestrol acetate. MPA = medroxyprogesterone acetate.

Supp2

Supplementary File 2. Overview of the included studies. EC = endometrial cancer. EH = endometrial hyperplasia. EIN = endometrial intraepithelial neoplasia. GnRH = gonadotrophin-releasing hormone. LNG-IUD = levonorgestrel-releasing intrauterine device. MA = megestrol acetate. MPA = medroxyprogesterone acetate.