The effect of progestin therapy in advanced and recurrent endometrial cancer: A systematic review and meta-analysis.

Study was conceived and designed by WW and JP. Data were acquired by WW and PB who also performed the quality control of data and algorithms. Data analysis and interpretation were performed by WW, PB, RL, RK, AR and JP, and statistical analysis was performed by WW, PB and JI. The manuscript was prepared by WW and PB, edited and reviewed by WW, PB, RL, JI, RK, AR and JP.

This study was performed in accordance with the declaration of Helsinki. No individual patient data was used, therefore institutional review board was not indicated.

There was no funding for the execution of this research.

This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) and Meta‐analysis Of Observational Studies in Epidemiology (MOOSE) guidelines and registered with PROSPERO (CRD42018089801) in March 2018. 29 , 30 , 31 The research question was drafted according to the PICO framework: patients with advanced stage (FIGO Stage III–IV) or recurrent EC were eligible, the intervention was progestin therapy, the comparison was any other treatment or no treatment and the primary outcomes were overall response rate, clinical benefit rate and toxicity. Exclusion criteria included studies consisting of case series with fewer than ten participants, studies on endometrial stroma sarcoma and studies on adjuvant or combined treatment of different types of hormones or hormonal treatment with other therapy. Further exclusion criteria were languages other than English, Dutch or German, systematic reviews and non‐full text availability. Studies that did not report all outcome measures of interest were included in this review for the reported outcomes only. An electronic literature search for eligible studies was performed in the databases PubMed, Embase and the Cochrane database of Systematic Reviews from inception to 31 December 2020. The citation lists of included studies were checked to identify further reports of trials. The search terms were ‘Endometrial cancer’, ‘Progestins’ and outcome measures ‘Disease progression’, ‘Survival’ and ‘Recurrence’ and related terms. The complete search string is shown in Table  S1 . The primary outcomes were overall response rate (ORR), clinical benefit rate (CBR) and toxicity. The ORR was defined as the proportion of patients with a complete or partial response. The CBR was defined as the proportion of patients with a response and with stable disease. Toxicity was defined as any adverse reaction that could be related to the applied hormonal drug in accordance with the protocol of the included studies. The secondary outcomes were overall survival (OS) and progression‐free survival (PFS). For the selection of eligible studies, the systematic review tool rayyan was used. 32 After exclusion of duplicates, studies were screened based on title and abstract by two of three reviewers (WW, PB and JP). Potential eligible studies were screened based on the complete text of the study. Discrepancies were resolved in consensus meetings of the three reviewers. The following patient characteristics were extracted from the studies using a standardised protocol: age, tumour stage and grade, PR status (PR‐positive or PR‐negative), histology, (previous) treatment, duration of study intervention, ORR, CBR, OS, PFS and toxicity. Authors of included articles were contacted if additional data were required. The quality of a study was appraised using the Revised Cochrane risk‐of‐bias tool for randomised trials (RoB2) for randomised studies and the ROBINS‐I risk‐of‐bias tool for non‐randomised trials. 33 , 34 Two authors (WW and PB) applied the risk‐of‐bias tools independently, followed by a consensus meeting to discuss and solve differences. In case of discrepancies, consensus was reached after discussion with the third reviewer (JP). The meta‐analyses for ORR and CBR were conducted with a random effects model. Between‐study heterogeneity was assessed with the restricted maximum likelihood estimator for τ 2 , and expressed as I 2 . 29 Publication bias was explored with funnel plots and Egger's test. 35 Sensitivity analyses were performed to evaluate any possible biases in the results caused by studies with high risk of bias, studies published before the year 2000, studies that did not apply standardised evaluation criteria and studies that applied other progestin therapies than the currently used megestrol acetate (MA) and medroxyprogesterone acetate (MPA). Subgroup analysis was performed based on the presence of PR and tumour grade (grade 1–2 versus grade 3 and grade 1–2 versus grade 3–4 for studies that applied the tumour grading system as described by Broder). 36 Response was not stratified by ER status because progestin interacts mainly with PR to exert its biological function and a recent report showed that PR is superior to ER in predicting response to hormonal therapy. 37 The PR‐status definition was adopted from the cutoff values defined by the individual studies. The association between response and biomarker presence or absence was expressed as risk difference and analysed with Mantel–Haenszel method with Hartung–Knapp adjustment for a random effects model. Average scores were used to describe the prevalence of overall and severe toxicity. OS and PFS were expressed in ranges stratified by response and PR status if available. The analyses were conducted with the statistical software R, version 3.6.3 (2020), and packages meta (version 4.11‐0), metafor (version 2.1‐0), and mixmeta (version 1.0.8). 38 , 39 , 40 Patients were not involved in the execution of this review. No core outcome set was used in this research.

The literature search yielded 7865 records based on application of the search string and screening of citations of eligible studies (Figure  1 ). After removal of duplicates and screening for title and abstract, 181 records were assessed for eligibility. A total of 155 records were excluded because of: case series with fewer than ten cases ( n  = 20), review/comment or guideline ( n  = 33) and wrong population or treatment ( n  = 102) including studies on fertility preservation and adjuvant progestin therapy mainly. Twenty‐six studies were included in this review. PRISMA flow diagram of selection process of included studies The 26 included studies comprised a total of 1639 patients. Four studies (15.3% of included studies) were randomised controlled trials. 17 , 41 , 42 , 43 One trial compared low‐dose and high‐dose MPA, the second trial compared MA and irosustat (a steroid sulphatase inhibitor), the third compared MA and combined treatment with MA and tamoxifen and the fourth trial compared MPA and tamoxifen. Out of the other studies, ten were prospective (38.5%) and 12 were retrospective (46.2%). The studies were published between the years 1961 and 2017 and 21 (81%) of them were published before the year 2000. The characteristics of the included studies are described in Table  S2 . Patients in the included studies had a mean age that ranged from 60 to 79 years. Different types and dosages of progestins were applied: treatment with either MA or MPA was applied exclusively in 13 studies (50.0%). The quality assessment showed a moderate risk of bias for 17 studies (65.3%) and a high risk of bias for nine studies (34.7%, Table  S3 ). None of the studies had a low risk of bias. All studies with the exception of one, reported ORR to progestin therapy, resulting in a total number of 1611 included patients. 44 The ORR was 30% (95% CI 25–36%) with substantial heterogeneity ( I 2 79%, Figure  2A ). The funnel plot can be found in Figure  S1 (A). The reported durations of progestin therapy ranged from 18 days to 108 months and the reported duration of response was between 3 and 95 months. CBR was available in 14 studies and 1017 patients and was 52% (95% CI 42–61, I 2 81%, Figure  2B ), For the funnel plot see Figure  S1 (B). Sensitivity analyses for ORR and CBR are shown in Figures  S2 and S3 , respectively. Among moderate‐risk studies, the ORR was 29% (95% CI 22–36%) and the CBR was 50% (95% CI 40–59%). ORR and CBR were similar for studies published before and after the year 2000. In 19 of 25 studies that reported ORR or CBR, a standardised response scale was applied: in two studies RECIST criteria were used, in three studies GOG response criteria were applied and in one study ECOG response criteria were used. In the other studies, the response criteria ranged from any decrease in tumour size to a complete response, or were not specified ( n  = 3). The minimally required duration of effect for response was given by six other studies ranging from 4 weeks to 3 months. ORR was lower in studies with standardised response criteria (27% 95% CI 21–33%) compared with studies without defined response criteria (45%) (95% CI 34–56%). There was no difference between studies that applied exclusively MA or MPA compared with other progestins. Meta‐analysis of overall response rate (A) and clinical benefit rate (B) to progestin therapy. The PR expression status was reported by seven studies. In two studies, ORR was not reported stratified by PR status leaving five studies and 201 patients. 17 , 24 , 42 , 45 , 46 , 47 , 48 PR status was based on primary tumour tissue in two studies and metastatic tumour lesions in three studies (Figure  3 ). The method for PR analysis was the ligand‐binding assay with different cutoff values in all included studies. As shown in Figure  3 (A), the pooled difference in ORR between PR‐positive and PR‐negative tumours was 43% (95% CI 15–71%, I 2 53%). The ORR for PR‐positive tumours was 55.4%, while PR‐negative tumours had an ORR of 12.2%. CBR stratified by PR status was not reported in any of the included studies. The ORR to progestin therapy stratified by tumour grade was reported by 12 studies including 915 patients (Figure  3B ). The pooled difference in ORR between low‐grade and high‐grade tumours was 19% (95% CI 15–24%, I 2  = 0%). Low tumour grade was associated with an ORR of 29.1% whereas high tumour grade yielded an ORR of 9.7%. Meta‐analysis of overall response rate in subgroup stratified by progesterone receptor status (A) and tumour grade (B). *All available studies used ligand‐binding assays instead of presently used immunohistochemical analysis methods. In the study of Yunokawa and Pautier, response was stratified by PR status, but raw data were not published nor available on request. Abbreviations: fmol, femtomole; high grade, tumour grade 3 (and grade 4 of Broder's grading system); low grade, tumour grade 1–2; PR, progesterone receptor. Treatment‐related toxicity was reported by 11 studies (Table  1 ). Adverse effects to progestin treatment occurred in 28.8% of patients and these effects were mainly mild. Severe toxicity was reported by 6.5% of all patients (range between individual studies: 0–31.0%) with the highest severe toxicities reported in the study by Lentz et al., in which megestrol acetate was used at the high dosage of 800 mg/day. 18 Thromboembolic processes were the most common severe adverse effect, occurring in 2.5% of patients. Toxicity for progestin therapy MA 160 mg/day MPA 500 mg/day 40 (69.0) MPA 250 mg/week MA 40 mg/day Medrogestone 600 mg/day Abbreviations: HPC, hydroxyprogesterone‐caproate; MA, megestrol‐acetate; MPA, medroxyprogesterone‐acetate. Studies that did not report total toxicity were excluded from this analysis. Follow up was reported in 4 of the 26 included studies. 44 , 48 , 49 , 50 Data on OS was available in 14 studies. The OS ranged between 2.8 and 57.0 months, depending on initial response to progestins. In responding patients, the range in OS was between 7.0 and 59.8 months, in patients with stable disease, the range in OS was 6.3–42.0 months and for patients with progressive disease, OS ranged from 2.8 to 9.1 months. OS stratified by PR status was investigated in two studies. 17 , 45 Patients with PR‐positive tumours had OS that ranged from 9.0 to 12.1 months, whereas patients with PR‐negative tumours had OS ranging from 6.8 to 7.3 months. Data on PFS was available from seven studies and was stratified by neither response status nor PR status. The PFS ranged from 2.5 to 10 months.

In this systematic review and meta‐analysis we have found an overall response rate to progestin treatment of 30%, and the clinical benefit rate was 52%. Progesterone receptor expression was associated with an ORR of 55.4% in PR‐positive EC and 12.2% in PR‐negative EC. The impact of PR expression on PFS and OS is unknown. Severe toxicity was reported in 6.5% of patients, among which thromboembolic processes occurred in 2.5% of patients. The overall quality of the included studies was moderate to low and the type and duration of progestin treatment as well as response evaluation varied substantially between studies. Progestin therapy has been the subject of three earlier systematic reviews. In the Cochrane review performed by Kokka et al., 7 progestin therapy for advanced and recurrent EC was found to have no statistically significant impact on OS. However, the Cochrane review included studies on adjuvant progestin therapy without reported ORR, hence results cannot be compared to our review. The review by Ethier et al. 28 reported an ORR of 23.3% and CBR of 45.8%, which is slightly lower when compared with our findings. This is probably explained by the inclusion of more recent studies with higher ORR in our review. In the review by Decruze et al., 25 the ORR was 23% in line with results of Ethier et al. 28 Compared with the reviews of Decruze and Ethier and their colleagues, our review provides updated information on the efficacy of progestin therapy and the predictive value of PR, tumour grade and toxicity data. The presence both of ER and of PR can be used as a predictive biomarker. As PR expression is more relevant, we focused on PR expression in this review. 37 , 51 Currently, immunohistochemical (IHC) staining analysis is the reference standard to determine expression of ER and PR. Only two studies in this review used IHC analysis for PR and in these studies, response was not stratified by PR status. 42 , 48 Upon request, these data could unfortunately not be made available for this review. As a result, only studies with an older ligand‐binding assay for PR expression were included. 17 , 24 , 45 , 46 , 47 Ligand‐binding assays were used as the standard test for presence of ER and PR until the 1990s. After that, monoclonal antibodies used in IHC analyses became available and quickly replaced ligand‐binding assays because they are easier and cheaper to use and better reflect PR status in the tissue and predict response to hormonal treatment in breast cancer. 52 , 53 Unfortunately, there are still a limited number of studies that related PR‐IHC expression to response to hormonal treatment in EC. Nevertheless, PR positivity was associated with a higher ORR to progestin therapy compared with PR‐negative patients. Among the PR‐negative group, we observed an ORR of just 12.2%, which questions the application of progestin therapy in these patients. Of note, in the study with the highest cutoff value for PR positivity, the highest ORR in PR‐negative tumours was observed. 24 This implies that further research into identification of the optimal cutoff is needed to differentiate between response and no response and improve the predictive value of PR expression. In a recent retrospective study, a novel cutoff value of 50% for PR expression was proposed as all patients with a response had PR expressions greater than 50%. 37 Tumour grade was also related to response to progestin treatment with a higher observed ORR in low‐grade tumours. Tumour grade and ER/PR expression are strongly correlated and ER/PR expression is significantly higher in low‐grade compared with high‐grade tumours. 54 As a result, most responses to hormonal therapy can be found among low‐grade tumours. Nevertheless, a subset of high‐grade tumours have positive PR expression and might therefore be responsive to hormones. 54 In our review, data stratified by both grade and PR status were not reported by individual studies. Further study is indicated to determine the response to progestin treatment in PR‐positive high‐grade ECs. The incidence of severe toxicity to progestin treatment was low with a comparable risk of thromboembolic processes to treatment with first‐line chemotherapy with carboplatin and paclitaxel. 16 Progestin therapy compares favourably with chemotherapy as severe toxicity occurs in around 35% of chemotherapy‐treated patients. 16 To our knowledge this is the first systematic review and meta‐analysis that specifically investigates the efficacy and toxicity of the role of PR and grade as predictive biomarkers for progestin therapy in advanced and recurrent EC. There are also some limitations to be addressed. The included studies were mostly retrospective, had moderate or high risk of bias and between‐study heterogeneity was greater than 60% for the main outcome measures. In addition, none of the studies included PR‐IHC data and therefore, available data on the predictive value of PR is solely based on ligand‐binding assays. This shows that high‐quality research into the efficacy of progestin treatment in the population with advanced and recurrent EC is still lacking. The definition for response was defined by 76% (19 of 25 studies reporting ORR) of included studies supporting validity of this outcome measure. However, length of follow up was formalised by just 25% (4 of 16 studies reporting OS or PFS), impacting the validity of OS and PFS outcome measures. There is an urgent need for a prospective study with standardised response criteria, regular follow‐up visits including radiological evaluations and PR‐IHC expression data to study the relation between predictive biomarkers and ORR, PFS and OS. Similarly, toxicity was defined by a subset of 11 studies and a standardised toxicity scale was applied in 27% (3 of 11) of these studies, although most studies did differentiate between severe and non‐severe toxicity. Grade 3 or 4 toxicity in studies with a standardised toxicity scale did not differ from prevalence of severe toxicity in studies without a standardised toxicity scale. Progestin treatment can be an effective palliative systemic treatment and has the advantage of limited toxicity. Women with advanced and recurrent EC for a large part consist of elderly women with comorbidities so it is essential to have low‐toxicity treatments available. However, the ORR to progestin treatment is still limited and further stratification based on predictive biomarkers is needed. Based on the current study, PR expression appears to be the most promising predictive biomarker for response to progestins. However, more research is needed to optimise the predictive value of PR expression. The genomic characterisation of EC has increased possibilities to apply targeted treatment to molecular alterations. 55 For example a novel ER pathway activity test that infers the activity of the ER pathway based on the level of mRNAs of established ER target genes has recently been related to progestin therapy in EC. 56 The response rate to progestin therapy in EC with an active ER pathway was 62.1% in this study. 37 In the same study a high correlation between ER pathway activity and PR‐IHC expression was found, suggesting that PR‐IHC expression might be used as a cheap and easy‐to‐use surrogate marker for estrogen‐driven tumour growth. A PR expression cutoff value of more than 50% was suggested in this study as the most relevant cutoff value showing an ORR of 56.8% and a sensitivity of 100%. 37 Genomic characterisation of EC has also resulted in the identification of a novel molecular classification in The Cancer Genome Atlas (TCGA) project. The classification involes three markers (POLE, mismatch repair deficiency, P53 mutation) that are used to classificay cases into four groups with distinct prognosis. There is growing evidence that these markers have not only prognostic but also predictive value for targeted therapy. For progestin therapy however, there is no clear association between TCGA subgroups and response. To date, two studies have found mismatch‐repair status to be potentially predictive for progestin therapy, but the PR status was not incorporated in both studies and the population included atypical hyperplasia and low‐grade, stage I EC instead of advanced stage and recurrent disease. 57 , 58 Therefore, ER and especially PR remain the best investigated biomarkers for the progestin therapy response in this population. In this systematic review and meta‐analysis, the efficacy and toxicity of progestin therapy in advanced and recurrent EC is reported. In unselected populations, the ORR to progestin treatment was 30% and the CBR was 52%. PR expression was the most discriminative biomarker with an ORR of 55.4% in PR‐positive EC. The toxicity profile of progestins is favourable. Prospective studies are needed to confirm the predictive value of PR‐IHC expression on response to progestins and investigate the impact on PFS and OS.

The aim of this study was to systematically review the efficacy and toxicity of progestin therapy in advanced stage and recurrent EC and investigate the value of predictive biomarkers.

Progestins have a multitude of applications in gynaecological practice. 1 , 2 , 3 , 4 , 5 Aside from their use for menstrual disorders, endometriosis and prevention of preterm birth, progestin therapy can be applied in women with endometrial cancer (EC) for fertility preservation and as palliative treatment in advanced stage or recurrent disease. 6 Approximately 15% of EC patients present either with advanced stage disease or recurrence; for these patients prognosis is poor and curative treatment options are limited. 7 , 8 , 9 , 10 , 11 Hormonal therapy and chemotherapy are the most frequently used palliative systemic treatment options. 12 First‐line chemotherapy consists of carboplatin and paclitaxel and is beneficial in 50–60% of patients. 13 , 14 , 15 , 16 However, this therapy is associated with substantial toxicity including sensory neuropathy in 20% of patients. 13 , 16 Hormonal therapy on the other hand yields response rates of 20–40% in unselected populations, but with limited toxicity. 17 , 18 Hormonal therapy was introduced in the 1960s when progestins were shown to have antiproliferative properties that counteract the effects of estrogens in EC. 19 Today, hormonal therapy is increasingly used, mostly as progestins, with other drugs like selective estrogen receptor modulators and degraders (i.e. tamoxifen) and aromatase inhibitors being reserved as second‐line treatment options. 8 , 20 The response to hormonal therapy can be predicted by the presence of estrogen receptors (ER) and progesterone receptors (PR) in tumour cells. The ER and PR serve as mediators of intracellular estrogen and progestin action, respectively, and are expressed by endometrial cells. 21 In advanced or recurrent EC, PR expression is frequently lost and, as a result, sensitivity to progestins might be lower. 22 , 23 However, progestin therapy efficacy has mainly been studied in unselected populations. A limited number of studies have evaluated response to progestin therapy stratified by PR status, but response rates vary substantially and implications for clinical practice are therefore unclear. 17 , 24 In addition, previously performed reviews have summarised disease outcome with hormonal therapy, but not focused specifically on progestin therapy. 7 , 25 Inclusion of less active hormonal therapies in these reviews might underestimate the effect of progestin therapy. 26 , 27 Finally, toxicity has not yet been systematically reviewed. 28 Because patients with palliative treatment of EC are often elderly with significant comorbidities, it is essential to have insight into toxicity to ascertain optimal application of progestins in this group of patients. Performance of a systematic review and meta‐analysis might provide more robust evidence that facilitates improved treatment selection for systemic palliative therapy in EC.

None declared. Completed disclosure of interests form available to view online as supporting information.

The authors wish to acknowledge the help of Radboud university medical center to publish this article open access.

Figure S1 Click here for additional data file. Figure S2 Click here for additional data file. Figure S3 Click here for additional data file. Table S1 Click here for additional data file. Table S2 Click here for additional data file. Table S3 Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file. ICMJE Click here for additional data file.