Association Between Estrogen Receptors and GATA3 in Bladder Cancer: A Systematic Review and Meta-Analysis of Their Clinicopathological Significance

Background Estrogen receptors alpha (ERα) and beta (ERβ) and the cooperating protein GATA-binding factor 3 (GATA3) have been implicated in bladder carcinogenesis and tumour progression. GATA3 and ER have been functionally linked in the establishment of luminal fate in breast tissue, but to date their relationship in bladder cancer has not been established. This information will be useful to advance diagnostic and prognostic markers. Aim To determine the relationship between the expression of ERα, ERβ and GATA3 in bladder cancer, disclose their prognostic and diagnostic value and their association with clinicopathological characteristics. Methods A comprehensive literature search in PubMed database was performed for all immunohistochemical studies of ERα, ERβ and/or GATA3 in bladder cancer patients. We selected eligible studies in accordance with the PRISMA guidelines and evaluated methodological quality and risk of bias based on quality criteria from the reporting recommendations for tumour MARKer (REMARK) prognostic studies. Risk of bias assessment was performed using Review Manager 5. R software was used for all statistical analysis, the packages used were meta and dmetar for the standard meta-analysis, and netmeta for the network meta-analysis. Results Thirteen studies were eligible for ERα, 5 for ERβ and 58 for GATA3 meta-analysis. Low grade tumours showed significantly lower ERα expression. GATA3 was widely expressed in bladder tumours, especially urothelial carcinomas, with higher expression of GATA3 in low grade and low stage tumours. Data was insufficient to determine the prognostic value of either ERα or ERβ, but GATA3-positivity was associated with higher recurrence free survival. A negative correlation between ERα or ERβ positivity and GATA3 expression was disclosed. Additionally, several sources of heterogeneity were identified, which can be used to improve future studies. Conclusion The clinicopathological value of ERα and ERβ was inconclusive due to low availability of studies using validated antibodies. Still, this meta-analysis supports GATA3 as good prognostic marker. On the contrary, ERα-positivity was associated to higher grade tumours; while ERα and ERβ were inversely correlated with GATA3 expression. Considering that it has previously been shown that bladder cancer cell lines have functional ERs, this suggests that ERα could be activated in less differentiated cells and independently of GATA3. Therefore, a comprehensive analysis of ERα and ERβ expression in BlaCa supported by complete patient clinical history is required for the identification of BlaCa subtypes and subgroups of patients expressing ERα, to investigate if they could benefit from treatment with hormonal therapy. Systematic Review Registration Prospero, CRD42021226836.


INTRODUCTION
Bladder cancer (BlaCa) arises and progresses along two distinct pathways with distinct behaviour and molecular profile (1)(2)(3). Low grade, non-muscle invasive cancers (NMIBC) account for 75% of the cases at diagnosis and are characterized by good prognosis. However, patients frequently develop local recurrences requiring lifelong cystoscopy surveillance, and around 25% of the cases will ultimately progress to invasive disease (4). In contrast, muscle invasive tumour (MIBC) progress rapidly and have a high propensity for metastasis with 5-year survival rate less than 15%, even after radical cystectomy and systemic treatment (5)(6)(7). Cisplatin based chemotherapy has been the standard of care for MIBC for the past three decades. Recently, immune check point inhibitors and erdafitinib, an FGFR antagonist, have been approved and show therapeutic benefit for a small group of patients (8,9). Still, the relative lack of molecular biomarkers and targeted therapies for BlaCa diagnosis and treatment (10,11), renders the pathological assessment currently used insufficient to predict disease progression and response to therapy (12).
BlaCa risk is mainly associated with cigarette smoking and gender (13). It is 3 to 4 times more frequent in men than in women, with the excess risk in males remaining even after adjustment for known risk factors (14). Gene expression studies identified intrinsic basal and luminal subtypes of BlaCa that closely resemble corresponding subtypes of breast cancer (BC) (15)(16)(17). Luminal BlaCa is characterized by high expression of PPARg and active estrogen receptor (ER) signalling pathway including expression of FOXA1, GATA3 and TRIM-24 (17). GATA3 is a marker of luminal cell differentiation in the breast and bladder (18) and together with FOXA1 are important mediators of PPARg signalling to drive luminal fate in BlaCa (19). GATA3 loss is associated with an invasive less differentiated phenotype (20) and is mutated in~5% of sporadic and~13% offamilial BC (21)(22)(23). It is unclear if estrogens have any protective effect because women are more likely to be diagnosed with invasive disease and have less favourable outcomes after treatments (24). However, ER activation requires both GATA3 and FOXA1 (25). Disclosing the functional connection between GATA3 and ER expression in BlaCa may improve the current tools for patient management, namely their eligibility for endocrine therapy used to inhibit ER-mediated proliferation.
The two ERs (ERa and ERb) are expressed in the normal urothelium of both sexes (26). Analysis of the TCGA urothelial cancer data set (n=406) showed that ERa and ERb mRNA expression is low (median FPKM 0.2 and FPKM 0.1, respectively) but detected in about 80% of the samples. Moreover, several independent studies showed that, BlaCa-derived cells lines are responsive to anti-estrogenic therapy (27,28). To date, few studies have assessed the association between ERa and ERb protein with the clinicopathological features of BlaCa. The reports are inconsistent, and the role of ERs in BlaCa development and progression remains controversial, partly because many of the studies dealt with small and heterogeneous patient cohorts and used antibodies that were not validated for clinical diagnosis of ERa, or anti-ERb antibodies that were proved to be unspecific at a later stage (29,30).
A previous meta-analysis of immunohistochemical studies correlated ERb expression with high grade (OR=2,169; p<0,001) and muscle-invasive (OR=3,104, p<0,001) tumours (31) and revealed associations between ERb expression and worse recurrence-free (HR=1,573; p=0,013) and progression-free (HR=4,148; p=0,089) survivals in patients with NMIBC. However, these results are compromised due to inclusion of studies that used anti-ERb antibodies that are unspecific (29). In the same study, incomplete information hampered conclusive evaluation of associations between ERa expression and patient's clinicopathological features. Regarding GATA3, much effort has been devoted into understanding its prognostic value as immunohistochemical marker, but to date there is no systematic evaluation and meta-analysis of such findings. Additionally, there is no study assessing the relationship between these functionally related proteins. In this work, we present a systematic review of the literature and meta-analysis to investigate the associations between immunohistochemical detection of ERa, ERb and GATA3 with clinicopathological features such as patient's gender, age, tumour stage, grade and survival and explore the relationship between the expression of these three makers.

METHODS
This study was submitted to PROSPERO on January 7, 2021 and registered on February 7, 2021 (CRD42021226836).

Search Strategy
The aim was to identify all primary literature that reported immunohistochemical detection of ERa, ERb and GATA3 in BlaCa. All potentially relevant articles were identified by a search in PubMed/Medline database using both Medical Subject Headings (MeSH) terms and free text words in the search queries. Singular

Eligibility and Data Collection
All retrieved references were screened for eligibility based on the title and abstract analysis by two of the authors. Potentially eligible full-text articles were retrieved for full-text assessment. The articles were reviewed against the following inclusion criteria: (1) expression level of ERa, ERb or GATA3 analysed in human BlaCa samples by immunohistochemistry (IHC); (2) reports with sufficient data to evaluate the methodological quality of the trial and to carry out a meta-analysis, including a clear description of the study population and IHC methods (i.e. tissue handling, antibodies used, positive controls), and description of the methodology and cut-off used to assign expression status; (3) Correlation between ERa, ERb and/or GATA3 expression and clinicopathological data discussed; (4) when different papers reported ERa, ERb and GATA3 expression from the same patient cohort, the most recent or the most complete study was included. Only original reports were considered. Letters, reviews, case reports, editorials and comments were excluded. Selected references for which a fulltext report was not available after contact with dedicated libraries and with corresponding authors were also excluded. For ERs, only published studies using validated antibodies were included. A flowchart depicting the literature search and selection process is represented in Figure 1.
For ERs, a total of 331 articles were identified, 298 were excluded after title and abstract screening for relevance. Of the 33 studies included in the qualitative analysis, 17 were excluded after full text analysis due to insufficient data, duplicated report of the same cohort, or use of non-validated or non-specific antibodies (Table S1). This resulted in 16 studies included of which 2 included information on ERa and ERb (27,32), 11 on ERa (33-41) and 3 on ERb (42)(43)(44). For GATA3, 211 articles were retrieved, of which 129 were excluded after title and abstract screening. Of the 83 studies included in the qualitative analysis, 24 were excluded after full-text analysis due to insufficient data, duplicated report of same sample cohort, contradictory data between text and tables, and lack of information about antibody used (Table S1), resulting in 58 studies included. Three studies reported ERa and GATA3 in the same tumour sample cohort (35,45,46).
Data was extracted from all relevant articles independently by two authors using a predefined data collection template which included identification details (surname of first author, year of publication), number of cases (total number and number of positive cases), primary antibody and dilution used, cut-off for positivity, subcellular localization of the staining (cytoplasmic or nuclear), tissue used for analysis [whole section or tissue microarray (TMA)], tissue collection method [transurethral resection (TUR) and/or cystectomy (CYS)], expression levels according to clinicopathological features such as age, gender, tumour grade, stage, lymph node metastasis and histology. Tumour histology was grouped as pure urothelial carcinomas (UC), UC with divergent differentiation (UCDD) and variant histologies (VH) such as adenocarcinomas and pure squamous cell carcinomas. Prognostic data (duration of follow-up after surgery or treatment, endpoint, overall survival (OS), recurrence and progression-free survival) and the statistical analysis used in each study (type of statistical test, P-value, hazard or risk ratio, 95% confidence interval (CI), univariate or multivariate analysis) were also collected.
The methodological quality and the risk of bias of each study were assessed independently by two of the authors using a list of quality criteria derived from the reporting recommendations for tumour MARKer (REMARK) prognostic studies and any disagreement was resolved by consensus. Four areas of potential bias were assessed: study design, assay methodology, results reporting and methods for statistical analysis. Risk of bias assessment was performed using Review Manager 5 (RevMan 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). The overall risk of bias for an individual study was categorized as low (green: risk of bias low in all domains), unclear (yellow: risk of bias is unclear in at least one domain, but no domains with high risk) or high (red: high risk of bias in at least one domain) as shown in Figure S1. The weight of all studies on the overall risk of bias for each specific domain is shown in Figure 2.

Data Analysis
All analysis were performed using R software (version 3.6.2) and the packages meta, dmetar (47) for statistics and netmeta for network meta-analysis (48). The prevalence, odds ratio (OR), Cohen's d and relative risk (RR) were calculated as point estimates of the association between expression of ERa, ERb or GATA3 and the patients' clinicopathological characteristics. Pooled prevalence indicates the proportion of positive staining for each marker. Pooled OR was used to evaluate differences in the proportion of positive cases between pre-defined groups. Cohen's d effect size was calculated relative to differences between the average age of the patients reported to be positive or negative where d = 0 means that distribution of ages in one group overlaps the distribution of ages in the other group. The effect size can further be interpreted as small (0.1), medium (0.5) and large (0.8), with higher values indicating less overlap between the groups (49). Pooled RR was calculated for differences in GATA3 positivity regarding Relapse-Free Survival (RFS). Between-studies heterogeneity was estimated using heterogeneity index (I 2 ) statistics (50). In case of substantial heterogeneity between studies (I 2 >50%), only the results from random effects model were considered for further analysis; otherwise, a fixed effect model was used for the pooled statistical analysis and a meta-regression analysis (mixed-effects model) was performed using an 'adjusted effect' to potential moderators. All results were considered statistically significant at the level of 5% (p <0.05). Sensitivity analysis was carried out to assess the robustness of the results by removing individual studies from the meta-analysis and assessing the effect on the pooled results. The publication bias was evaluated using funnel plots and two-sided Egger's tests ( Figure S2).
Subgroup meta-analysis and/or meta-regression were performed to explore sources of heterogeneity using five factors: 1) antibody used, 2) cut-off for positivity, 3) tumour histology, 4) sample type and 5) sample collection. Meta-regression was also used to assess the influence of the following seven factors in the ERs and GATA3 proportion of positive cases: 1) gender, 2) tumour stage, 3) lymph node metastases, 4) tumour grade, 5) tumour histology, 6) therapy pre-collection, and 7) deaths by cancer.
To assess a possible relationship between the expression of ERa, ERb and GATA3, we estimated OR using pairwise and network meta-analysis with random effects using frequentist methods. Moreover, we evaluated the inconsistencies between direct and indirect comparison using the z-value of test for disagreement (direct versus indirect) in network meta-analysis.

Characteristics of the Eligible Studies for the Systematic Review
All studies were retrospective, 13 were eligible for ERa comprising a total of 1616 tumour samples (1998-2020; 20 -  Methodological Quality and Risk of Bias REMARK (106) based risk of bias assessment is shown Figure 2 and Figure S1. The most common factor in the bias analysis was lack of information on pre-operative treatment status (high risk for ERa and GATA3 in over 50% of studies). Followed by positive controls (unclear in 25% of ERa studies and in over 50% of ERb or GATA3 studies), incomplete description of clinicopathological characteristics of the cohort (specially for GATA3, with nearly 50% studies unclear or not reporting) and no information about the quality controls including positivity criteria (above 50% of ERb or GATA3). The main differences in the methodology between studies included: use of different antibodies and antibody dilutions or different scoring systems.

Meta-Analysis of ERa Expression in BlaCa
The pooled proportion of ERa-positive cases was 7%, (0-38%; Figure S3). Despite the high level of variation between the 13 studies (I2 = 93%), the sensitivity analysis did not identify any study as having a significant influence in the overall heterogeneity ( Figure  S4A). However, it is worth mentioning that data from Imai (2019) stood out and influenced pooled results, most likely due to the use of a lower cut-off (1%) and inclusion of UCDD tumours. Subgroup analysis could not explain the heterogeneity between studies (Table  S2); Meta-regression disclosed lymph node metastases as a significant source of variation associated with ERa expression (p-value = 0.0275; mixed-effects model; Table S3).

Correlation of ERa Immunostaining With Clinicopathological Parameters
We conducted a binary meta-analysis to establish the correlation of ERa-positive cases with clinicopathological parameters: gender, age, tumour grade, tumour stage and histology ( Table 4).     Figure 3). For stage analysis, data from 4 studies (I 2 = 5%) was divided as Ta+T1 (218 cases) and >=T2 (136 cases) and no significant association was found (CI= [0.31-1.04], although there was tendency for higher ERa-positivity in late-stage tumours. The association with histology could only be inferred from a single study (n=125) with low number of VH cases and showed that the proportion of ERa-positive cases was lower in UC tumours when compared with either VH or UCDD ( Table 4).

Meta-Analysis of ERb Expression in BlaCa
Four hundred and thirty samples pooled from 5 studies were ERb-positive ( Figure S5), corresponding to 69% of the cases (range: 49-91%; I2 = 94%). Neither subgroup analysis nor meta- regression could explain the source of heterogeneity (Tables S2 and  S3) and a sensitivity analysis showed that selective omission of each study did not influence the overall heterogeneity ( Figure S4B).

Correlation of ERb Immunostaining With Clinicopathological Parameters
A binary meta-analysis was conducted to evaluate the association of ERb positivity with patients' gender, tumour stage, grade, presence of lymph node metastasis and patients' pre-operative treatment ( Table 4). Variation between studies was high and no significant association was found between ERb expression and the clinicopathological parameters evaluated except for lymph node metastases. ERb-positive cases were significantly correlated with the presence of lymph node metastasis [n=309 from 2 studies (32, 44)] (Figure 4).

Meta-Analysis of GATA3 Expression in BlaCa
GATA3 was expressed in 85% of the 4275 pooled cases from 58 studies (range: 3-100%; Figure S6). Despite the high level of   heterogeneity (I2 = 97%), the sensitivity analysis did not identify any study as having a large influence in the overall results ( Figure S4C). However, data from Liang (2014) stands out and influences pooled results possibly due to the higher number of UCDD cases analysed (79). However, this trend did not reach significance in the subgroup meta-analysis (Tables S2, S3) indicating that technical variations or cohort composition were not the main drivers of heterogeneity. Meta regression was used to estimate whether the heterogeneity between studies was explained by clinicopathological covariates (Table S3).

Association Between ERa, ERb and GATA3
Network meta-analysis was performed to assess a possible relationship between the expression of ERa, ERb and GATA3 ( Table 5). The model was based on direct evidence pooled from studies evaluating at least two of the proteins in the same study: 3 studies for ERa and GATA3 (35,45,46), 2 studies for ERa and ERb (27,32) and 1 study for ERb and GATA3 (86). The model showed that both ERb (0.014; 95%; CI: 0.007-0.030) and GATA3 (0.002; 95%CI: 0.001-0.005) positive cases negatively correlate with ERa-positivity. GATA3 positivity was also negatively associated with ERb positive cases (0.168; 95%; CI: 0.098 -0.290), even though the association wasn't as strong as for ERa. Still, this associations should be interpreted with extreme caution as even though the studies evaluating ERb used antibodies that to date were not found to be unspecific, there is still great controversy as to how to best detect ERb by IHC and the number of studies is low. No disagreement/inconsistency between direct and indirect comparison were detected as significant (p = 0.936).

DISCUSSION
BlaCa is a heterogeneous disease for which to date, limited histopathological markers and therapeutic options exist. Gene expression signatures with GATA3 and active ER signalling characterize luminal BlaCa (15,17) and disclose some similarities between luminal BlaCa and BC (107). In the breast, GATA3 is a necessary transcriptional coactivator of ERamediated proliferation (25,108), both proteins cooperate to maintain the epithelial lineage and are diagnostic tools for luminal BC (109). However, it is unclear whether these proteins collaborate or have a role in luminal BlaCa pathophysiology. Since ERa is the gold standard for indication of hormonal therapy and both ERs can be targeted with hormonal therapy (110), disclosing the relationship between ERs and GATA3 is a necessary step to advance BlaCa diagnostics and therapeutics. To date, this is the first systematic review and meta-analysis addressing a potential relationship between ERs and GATA3 in BlaCa. Moreover,  based on recent findings that disclosed a vast amount of anti-ERb antibodies as unspecific (29,30), we restricted the inclusion criteria to only include validated anti-ERb antibodies.
To improve our understanding ERa, ERb and GATA3 roles in BlaCa pathophysiology we defined the following questions a priori: 1) What is their prognostic value? 2) What is their diagnostic value? 3) How do clinicopathological parameters impact their expression? 4) What are the sources of heterogeneity and which can be controlled in future studies? and 5) Is expression of these three markers associated in any way?

Prognostic Value
Six studies analysed the association between ERa and patients' prognosis, with no significant association between ERa expression and tumour recurrence/progression or survival observed in each individual study (27,33,34,39,40). An exception was Pena et al. that showed less likelihood for tumour recurrence in ERa -positive cases using unadjusted logistic regression (46). Due to differences in the methodology and information reported, it was not possible to carry out a metaanalysis. Therefore, current data is still insufficient to determine the prognostic value of ERa. However, higher ERa-positivity was observed in late-stage and high-grade tumours not only in the present meta-analysis but also in individual studies (33,(36)(37)(38), which support the hypothesis that ERa positivity may be a marker of poor prognosis. Our analysis disclosed an association between ERa expression and higher-grade tumours. Moreover, cell line studies showed that blocking ERa signalling with antiestrogens reduces cancer cell viability (27,28), and a case study reported regression of metastatic transitional cell carcinoma in response to tamoxifen (111). Aromatase expression in the tumour parenchyma and stroma has been found significantly associated with more than a 2-fold risk of bladder cancer recurrence and may be associated with advanced tumour stage and poorer survival outcomes (112), while aromatase in the tumour stroma was significantly associated to adverse pathologic variables and poorer overall survival (113). On the other hand, the predictive value of ERb is debatable, one study found ERb-positivity to be associated with worse prognosis for low-grade tumours and lower CSS in high-stage tumours (114), while another study didn't find any correlation between ERb positivity and tumour recurrence (42). Additionally, Kauffman et al. found that higher ERb levels were predictive of worse RF and OS following cystectomy (115). In the current meta-analysis data was insufficient to determine the prognostic value of ERb due to differences in the methodology and data reported among individual studies. Regarding GATA3, pooled    analysis indicated that positive expression was significantly associated with lower risk of recurrence, which is in agreement with the results of the individual studies included in this analysis (112,113) and others that didn't meet including criteria (73). This result independently confirms the prognostic value of GATA3 immunohistochemical determination in BlaCa.

Diagnostic Value
Out of the 13 eligible studies for ERa, 8 analysed the association between ERa and tumour stage, grade, histological type and/or presence of lymph node metastasis. Two studies evaluated ERa expression in cohorts with multiple histologies and none found significant differences among the different histological types (37,40). In the current meta-analysis, the association between ERa and tumour histology type was inferred from a single study (37) due to mathematical limitations. However, ERa-positivity was found to be higher in VH and UCDD histological types and less frequent in UC. From the 6 studies analysing ERa among different grade and stage, four found ERa expression to be significantly associated with high grade and muscle invasive tumours (33,(36)(37)(38) as in this meta-analysis. These results suggest that ERa-positivity is associated with more advanced tumours. No significant associations were found between ERa positivity and lymph node metastasis in individual studies (34,40), and the available data was not suitable for meta-analysis. In the case of ERb, previous studies based on evaluation of mRNA showed that ERb expression was associated with tumours of the luminal subtype (107). None of the eligible studies provided data to allow the investigation ERb expression among different histological types. The correlation between ERb expression and tumour grade was inconsistent among the 5 eligible studies. Two studies found significant association between positive expression and tumour grade but in opposing directions. One noted higher expression in low grade tumours (44) an another in high grade tumours (43). A trend association between positive expression and high grade (32) and no relationship found in the remaining 2 (27,42). Out of the 5 studies, 4 investigated the relationship between ERb expression and tumour stage. ERb expression was found significantly associated with high stage tumours in 2 of them (32,44). In the same cohorts, ERb-positivity was also associated with lymph node metastasis as also observed in this meta-analysis. The association between ERb expression and other clinicopathological variables remains to be investigated. GATA3 is an established marker of luminal papillary bladder tumours which are the least aggressive tumours and still retain some of the features of the urothelial differentiation (98). Similar to the results of this meta-analysis, four studies reported significant association between GATA3 expression and low grade (51,59,71,86) and low stage (51,59,71,86) tumours. GATA3 expression showed a significant association with tumour histology, with higher expression in UC as opposed to VH and UCDD, both in individual studies (57,60,70,77,79,84,86,90) and in our meta-analysis. This is not surprising given its role in urothelial differentiation. Contradictorily, individual studies found significant correlation between increased GATA3positivity and cases with lymph node metastasis (86) while in another data set it was associated with lymph node negative cases (104). Only one of these studies had data suitable for our pooled analysis (86) which found no significant correlation between GATA3-positive cases and lymph node metastases.

Impact of Age and Sex
The relationship between age and sex and ERa (27, 33, 36-39,) and ERb-positivity was investigated in seven and two independent studies, respectively, revealing no significant associations. The meta-analysis of the pooled data didn't reach statistical significance but, suggest that both ERa and ERb are more frequent in tumours from female patients as compared to males, and no differences were observed regarding age. Considering the higher estrogen levels in females, even after menopause, this observation is aligned with epidemiological data showing less frequent but more aggressive BlaCa in females (24). In the case of GATA3, 10 studies reported expression levels by age and sex (57,65,73,81,86,93,96,98,101), but no significant associations were reported. In the current meta-analysis, GATA3 expression was more frequent in tumours from older patients and, although not significant, there was a trend for higher GATA3 expression in males. These results are in line with a recent study that identified differences in BlaCa molecular subtypes based on sex, with tumours from females expressing higher levels of basal genes and more frequently from the basal/ squamous subtype, while tumours from male patients expressed higher levels of luminal markers (116). A reduction of estrogen levels, as observed in menopause, causes urogenital side-effects (117,118) and may participate in the carcinogenic process by promoting an inflammatory environment (119,120). Therefore, it can be argued that antiestrogen therapy as used in BC treatment, which report similar urinary side-effects as menopausal and post-menopausal women would result in higher risk of developing BlaCa. We found a case study reporting a 65-year-old woman who developed non-muscle invasive low-grade papillary urothelial carcinoma grade 1, one and a half year after starting on endocrine therapy with aromatase inhibitors (121). However, this would not be related to ERa signalling directly initiating urothelial cell transformation, but to the inflammatory environment resulting from lower estrogen levels or higher androgen/estrogen ratio. Moreover, a large cohort prospective study found no overall associations of HRT use and oral contraceptive use with reduced risk of BlaCa (122).

Limitations of the Meta-Analysis
We found several sources of heterogeneity common to evaluation of ERa, ERb or GATA3, which may limit this meta-analysis but will certainly elucidate variables to consider in future research studies. These involve the inclusion of tumour samples from patients previously submitted to local or systemic therapy, which varied across different studies and most of the times it was not possible to stratify results by therapy. This might contribute to protein expression fluctuations in response to treatment. Another source of heterogeneity might be the publication bias related to lower number of non-statistically significant results, which can be explained by lack of reporting or less detailed description of results (123). Heterogeneity in stage and grade may be explained by inclusion of recurrent tumours which may have the same stage and grade of a firstly diagnosed tumour but distinct expression profile (124). Differences between primary and recurrences were not taken into consideration for analysis as this information is not available. The exclusion of cases with preoperative treatment and carcinoma in situ was also observed in some studies, which affect the stage and grade of the tumours under analysis. Similarly, the source of tissue contributes to variations as observed by a lower proportion of positive cases in TMA cohorts than in studies using whole-tissue sections. Regarding GATA3 and ERb, differences in tumour histology explained part of the heterogeneity, with more GATA3 positive cases among UC and UCDD than in VH tumours. Furthermore, for ER detection, the antibodies are also an important source of variability as, even though they are validated for IHC and clinical use, come may detect more than one ER isoform. For ERa, the use of the clones 6F11 and SP1 provided higher dispersion in ERa positive cases, while the 1D5 gave more consistent results. These monoclonal antibodies recognize different epitopes, 6F11 was raised against the full length ERa, 1D5 recognizes the N-terminus, while SP1 antibody recognizes the C-terminus of human ERa. Others have shown that 6F11 and 1D5 antibodies only bind the full-length protein (66kDa) and SP1 could in principle also detect splice variants of smaller size (36 kDa, 46 kDa) (116). In the case of ERb, all studies using non-specific antibodies were excluded, limiting the metaanalysis to the clone 14C8, which has been independently validated by different groups (29,117), and the polyclonal MYEB which, to date, has not been probed unspecific. However, clone 14C8 detects ERb isoforms 1 and 2 which, at least in vitro, have different biological effects (118) and may be differentially expressed. For GATA3, antibody usage does not seem to have much influence in the results obtained as evidenced by Kandalaft et al that used two different antibodies (119). Additional sources of heterogeneity that weren't explored in this meta-analysis might also be at play such as different technologies to perform IHC, sensitivity to recognize positivity by different pathologists, among others. Finally, we were not able to include absolute positive/ negative proportions, leaving some studies out of the pooled analysis for individual clinicopathological parameters.

Correlation Between ER and GATA3 Positivity
The network meta-analysis model showed that there is a negative correlation between ERa or ERb positivity with GATA3 expression, being the effect stronger for ERa. This agrees with Miyamoto et al (60) that showed a negative correlation between GATA3 and ERb expression. Furthermore, both individual studies and our meta-analysis, propose that ERa and ERb are markers of bad prognosis (33, 36-38, 60, 120), while GATA3 is associated to lower risk of recurrence and more differentiated tumours (60,73,98,115,121). Moreover, while GATA3 is higher in males, ERa appears to be higher in females. Therefore, in BlaCa, ERs and GATA3 do not appear to cooperate as observed in BC. Interestingly, ERa and ERb expression were also negatively correlated. ERb has been shown to counteract ERa activation, at least in some contexts (118,122), so it is possible that lower ERb contributes to an even more aggressive phenotype in ERa-positive BlaCa. Notably, this small subset of tumours may be eligible for hormonal treatment.

CONCLUDING REMARKS
This systematic review confirmed that ERa is expressed in a small proportion of bladder tumours (3 -13%) and is associated with higher tumour grade and stage independently of tumour histological type. Even if the % of positive cases is low, the possibility of benefiting these subgroup of worse prognosis patients with endocrine therapy should be further explored.
Our analysis and evidence from cell lines and aromatase expression points to a role of ERa in the progression of the disease. Functional studies are needed to identify if ERa is in fact a driver of proliferation in this subgroup of high-grade tumours and the relationship with aromatase expression in order to understand if these patients can benefit from antiestrogen therapy. No conclusion could be reached regarding ERb even though it is a signature marker for luminal BlaCa and detected by IHC in 69% cases. On the other hand, GATA3 is expressed in about 80% cases and associated with low grade and low risk of recurrence. Therefore, while we were able to confirm the prognostic value of GATA3 using data from two studies, more studies correlating these biomarkers with time to event endpoints are needed to establish their prognostic value. Interestingly, this meta-analysis highlighted that ERa expression is dissociated from GATA3. In fact, higher positivity for each protein was identified in different groups of tumours with GATA3 positive expression associated with well differentiated tumours and ERa with loss of urothelial differentiation. Therefore, these two proteins do not collaborate to maintain epithelial luminal differentiation as observed in BC (109) and instead, they either participate in different stages of tumour progression or may be required for growth of different cancer cell types. This should be further confirmed in prospective studies considering both markers in advanced tumours and preresection treatment.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.