Endometrial cancer (EC) is the most common gynecological malignancy with increased incidence rate in developed countries [1]. Almost 80% ECs are of endometrioid endometrial adenocarcinoma (EEA) histology and diagnosed early with good prognosis [2]. However, there is still a subset of patients failed to reach 5-year overall survival even for those with low-grade or early-stage EEA [3456]. After the initial diagnosis 15% of women in grade I and 25% in grade II EEA died from the disease in 5 years [35]. Also, among patients with grade I–II endometrioid ECs, approximately 3% in stage I and 20% in stage II did not survive over 5 years [4]. The current risk assessment and therapeutic decision-making majorly rely on the International Federation of Gynecology and Obstetrics (FIGO) grade, stage and histology [7]. Thus, this assessment is still suboptimal that some patients with poor prognosis might be left undertreated. More improved means are urgently needed, to precisely identify high-risk patients with grade I–II EEA.

Estrogen receptor (ER) and progesterone receptor (PR) are most validated prognostic biomarkers for endometrial cancer [891011]. Gene expression of ESR1 and PGR were found significantly correlated to the protein expression of ERα and PR by immunohistochemistry (IHC), respectively [12131415]. A prospective study evaluating ER and PR on curettage specimens from 832 endometrial carcinoma patients reported that ER and PR double loss independently predicted lymph node (LN) metastasis and shorter disease-specific survival (DSS) [16]. Other studies also associated the negativity of ER and PR with poor survival in endometrial carcinoma even in low-grade subtypes [891011171819], However, no clinical risk-stratified analysis has been done for grade I–II ECs patients with ER/PR loss. It is still unclear whether patients of this subgroup may need more intervention after surgery. Currently ER and PR status have not been implemented into any clinical guideline [202122]. It has not been investigated, whether integrating ER and PR into clinical risk stratification can assist to select high-risk patients with grade I–II EEA.

Integrating genomic classification into clinicopathological risk assessment has shown improved prognostic value [8232425]. The Cancer Genome Atlas (TCGA) research network had significantly improved risk stratification by discriminating ECs into four molecular categories: POLE ultra-mutated tumors, microsatellite instability tumors, copy-number low tumors and copy-number high tumors [26]. Nevertheless, whether TCGA classification has prognostic value in grade I–II EEA cases, and whether it correlates to expression of ER and PR are unclear.

Our study was to investigate whether ER and PR double negativity (ER/PR loss) could improve the risk stratification for patients with grade I–II EEA. We aimed at 1) the clinical characteristics for patients with ER/PR loss in grade I–II EEA; 2) the prognostic value of ER/PR loss in grade I–II EEA and in currently clinical risk groups; and 3) the prognostic value of TCGA genomic subgroups in grade I–II EEA and its association between ER/PR loss and TCGA classification.

The clinical and pathological characteristics of 903 patients were summarized in Table 1. After a median follow-up of 59.0 months (0-109.0) of follow-up, eighty-two (9.1%) patients were lost contact with us. Among 821 patients with complete follow-up, in total 41 (5.0%) patients experienced disease recurrence, of whom 17 (41.5%) died eventually due to EEA.

In grade I–II EEA, we found the ER/PR loss was a strong and independent risk factor for recurrence and death. These prognostic values were further observed in early and late stages and might be independent of TCGA classification. These findings are clinically important, because ER/PR loss may help identify high-risk patients from presumed low-risk group for in-time and tailored treatment. Also, immunohistochemistry of ER and PR are low-cost and easily applied for most medical centers. To our knowledge, this is the first large-scale study on the prognostic impact of ER/PR loss in grade I–II EEA integrated with currently clinical risk stratification.

The strengths of our study are including a large-scale homogenous population of grade I–II EEA and integrating molecular biomarkers with standard risk stratification. The prognostic values of ER and PR in ECs were mostly evaluated across all histological subtypes [1617193334]. Due to the tumor heterogeneity and the histological-subtype specific of ER and PR, possible prognostic effects should be analyzed in a subtype-specific manner [19]. A few studies with low-grade subgroup analyses among patients with endometrioid histology presented independently prognostic value of ER and PR, or their combination as a panel [16171835]. However, these subgroup-analyses were based on relatively small amounts of patients or curettage specimens, and no stratification analyses were followed. It is still unclear whether ER/PR can help identify patients that currently left undertreated in grade I–II EEA. In our study, we observed ER/PR loss had independently prognostic value in both presumed low-risk and high-risk subgroups. Grade I–II EEA patients in stage IA, IB, and II–IV were commonly considered with low-, intermediate- and high-risk, and provided with different therapeutic regimens after surgery. According to our findings, women with ER/PR loss should be shifted into a higher risk group or at least be paid on more clinical attention: those in stage IA may need adjuvant therapies after surgery rather than simple observation; and those in stage II–IV may need more frequent follow-up or more effective therapeutic protocols. Besides, we observed ER/PR loss predicted poor PFS and OS even in grade I and stage IA EEA patients, suggesting fertility preserving therapy requested by young patients with ER/PR loss should be considered carefully and re-assessed by clinical specialists. Thus, we suggest the cost-effective immunohistochemistry of ER/PR should be considered to apply routinely as prognostic biomarker panel to meet clinical needs, particularly for patients with grade I–II EEA.

The optimal cut-off for prognostic biomarkers would ideally be clinically validated against patient outcome in prospective clinical trials, or in systematically collected archived specimens from randomized clinical trials [36]. According to this, the optimal cut-offs for ER and PR staining have not been defined in ECs, hindering the evaluation on their prognostic values between studies [837]. Historically, many investigators had defined 10% or greater nuclear staining as the threshold for positive ER and PR [1824383940]. However, extended studies using ≥1% as such cut-off showed significantly better survival for patients with ≥1% ER-positive cells compared with patients with complete negative cells [29303141]. Also 67% of the exactly 10% ER-positive cancers and 24% of the 1%–9% ER-positive cancers showed ESR1 mRNA levels and gene signatures that were consistent with >10% ER-positive cancers, suggesting IHC might have underestimate ER positivity in 1%–9% ER-positive cancers [37]. Hence, we use ≥1% as the definition of positive ER and PR to provide more precise and robust evidence for clinical use.

In our study, ER/PR loss was found associated with poorer survivals in grade I–II EEA. Herein, we suspected if this subtype might include unidentified serous ECs that usually led to unfavorable prognosis. Some ambiguous endometrial carcinomas have overlapping histological features of serous and endometrioid carcinoma [274243]. On these challenging cases, lack of consensus on diagnosing histotype and grade between expert pathologists leads to vast difference in assessing prognosis and treatment for the same patient. These cases usually involve uterine serous versus FIGO grade II endometrioid carcinoma [27]. Using the FIGO grading system for endometrial endometrioid carcinomas, a tumor composed predominantly of glands or papillae (architecture grade I), even in the presence of high nuclear grade (nuclear grade III), would be classified as FIGO grade II endometrioid carcinoma [27]. p53 immunohistochemistry was showed with differential value in these ambiguous ECs that “favor serous” ECs have more p53 over-expression than “favor endometrioid” ECs [2732]. Ambiguous cases with p53 over-expression had a more aggressive clinical course than cases without p53 over-expression, which had similar 3-year PFS (94%) and DSS (100%) to that of FIGO grade II endometrioid carcinomas (85%) [27]. Nevertheless, no difference in p53 expression was observed between “ER/PR loss” and “ER or PR positive” groups in grade I–II EEA, suggesting the poor prognosis in ER/PR loss group may not result from unidentified serous ECs.

Improved risk assessment in ECs by integrating TCGA classification with clinicopathological factors has been addressed [232425]. In TCGA cohort, we found no prognostic value for TCGA classification in grade I–II EEA. This might due to the relatively small (n=140) series of grade I–II EEA had assigned TCGA genomic subgroups. In addition, to analyze the potential association between TCGA classification and ER/PR loss, we defined the low ESR1/PGR group in TCGA cohort. Studies showed mRNA expression of ESR1 and PGR were significantly related to ER and PR protein expression by IHC [121315]. We defined the lower quartile of mRNA expression as the cut-off for low ESR1/PGR [14], which achieved a closest HR (2.40; 95% CI=1.08–5.32) as that of ER/PR loss group in our cohort (HR=3.77; 95% CI=1.48–9.56), with sufficient samples (n=36) for analyses. Other cut-off values used previously were the median [1444], the first tertile [45] or the log2-converted value 10.18 [37]. However, none of them achieved statistical appropriateness in our study. We subsequently found no association between TCGA subgroups and ESR1/PGR status, suggesting ER/PR loss may predict poor prognosis independent of TCGA classification in grade I–II EEA. Nevertheless, expression of ESR1/PGR mRNA and ER/PR protein were not identified in same cohort. Whether the low ESR1/PGR group in TCGA cohort could represent for ER/PR loss in our cohort needs further validation.

Studies showed ER loss correlated with alternated phosphoinositide 3 kinase/mammalian target of Rapamycin (PI3K/mTOR) pathway 19. ER or PR loss might enable epithelia cells to become more like mesenchymal cells with increased ability to invade and resist apoptosis [1446]. Given PI3K/mTOR inhibitors today are in phase I/II trial in advanced ECs (www.clinialtrials.gov, November 2017), PI3K/mTOR inhibitors might also have potential as the target therapy for grade I–II EEA with ER/PR loss.

Despite the strength of the large and homogenous population, a robust cut-off of ER/PR and improved risk stratification, this study has some limitations. Firstly, our study is a retrospective single-center study. Moreover, the mRNA expression of ESR1/ PGR and the IHC of ER/PR were not performed in the same cohort that might generate bias. The cut-off for low ESR1/PGR expression corresponding to ER/PR loss also needs further validation. Currently we have no tissue for all the patients to run RNA-sequencing or polymerase chain reaction (PCR), although no study yet reports “real world” TCGA classification in grade I–II EEA, either. Given that RNA-sequencing is expressive and large-scale tissue collection is time-consuming, thus, our study still has merits for using public TCGA data to explore potential correlation. Since we found ER/PR status may not correlate with TCGA subgroups in grade I–II EEA patients, further studies may shift their priorities to other biomarkers when funding was limited, instead of beginning with high-cost investigations in coordinate with TCGA classification.

In conclusion, based on a large-scale and homogenous population, we found ER/PR loss strongly and independently predicted poor survival for patients with grade I–II EEA, in both early and late stages. This prognostic value might be independent of TCGA molecular classification. Grade I–II EEA patients with ER/PR loss may need more post-surgery intervention to improve their outcome. Fertility preserving therapy for these women may also need re-consideration. Herein, integrating cost-effective immunohistochemistry of ER/PR into future randomized clinical trials and clinical algorithms may improve risk stratification for tailored therapeutic management.