Endometrial cancer (EC) is the 6^th most prevalent neoplasm among women worldwide, representing 4.5% of cancer diagnoses in females. There are an estimated 420,000 new cases per year, and just over 97,000 women will die from this neoplasm in the same period (1). In Brazil, EC is the 7^th most prevalent tumor among women, with approximately 8,000 new diagnoses per year (2). Although the absolute numbers are not that impactful, a significant increase in new cases was observed in Brazil, probably related to the rise in obesity rates (3).

For years, the understanding of the pathophysiology of EC was based on the dualistic model proposed in the early 1980s, classifying it into type I (endometrioid) and type II (non-endometrioid) (4). In 2013, with the publication of The Cancer Genome Atlas (TCGA), this neoplasm was molecularly classified into four groups with different prognoses: POLE ultramutated (POLE), microsatellite instability hypermutated (MSI), copy-number low (endometrioid), and copy-number high (serous-like) (5). Due to a complex multi-omics methodology, it was difficult to reproduce and implement in daily practice.

Subsequently, two independent groups developed more straightforward methods to classify EC molecularly. The Proactive Molecular Risk Classifier for Endometrial Cancer (ProMisE) and Leiden/TransPORTEC classification used immunohistochemistry to detect the presence/absence of mismatch repair (MMR) proteins and to evaluate p53 expression, using, in both methodologies, the molecular analysis of POLE hotspot mutations. Based on this classification, EC patients were classified into one of four molecular subtypes: POLE mutated (POLEmut), mismatch repair deficient (MMRd), p53 abnormal (p53abn), and no specific molecular profile (NSMP) (6, 7). To simplify the analysis of molecular subgroups of EC in a single exam with a single sample, the Canadian group that created the ProMisE methodology carried out a study comparing the standard method with next-generation sequencing methodology (ProMisE NGS). Of the 164 samples tested in this study, there was disagreement in only 5, with an overall accuracy of 0.97 (8).

In 2021, the European Societies of Gynaecological Oncology, Radiotherapy, and Pathology (ESGO/ESTRO/ESP) published a consensus where, for the first time, the ProMisE molecular classification began to be considered for the therapeutic management of patients with EC (9). Based on this, patients in the p53abn subgroup, even in less advanced stages, are considered high risk and undergo an upgrade in their risk stratification. In contrast, patients in the POLEmut subgroup, who have suffered downgrades in their risk stratification, may have a de-escalation in their adjuvant treatment (9). Moreover, ProMisE was included in the new FIGO staging published in 2023 (10). Therefore, the molecular classification will become part of the therapeutic definition of patients with EC, even more so with the future results of phase 3 studies that base their treatment protocols on this novel stratification (11, 12). Recently, the lack of diversity in genomics studies has been highlighted, leading to unmet scientific needs and health disparities (13). In this context, the knowledge of endometrial cancer molecular profile in the Latin-American population, particularly the Brazilian, is scarce.

The present study aimed to describe the implementation of the ProMisE classification for routine endometrial cancer evaluation in a Brazilian tertiary public cancer center hospital, to compare it with a molecular approach of MSI and TP53 mutation, and to evaluate ProMisE impact on patients’ prognosis.

A total of 114 EC samples were analyzed prospectively. Clinicopathologic characteristics related to ProMisE classification are shown in Table 1 . Approximately 50% of patients were stage I, while 10.5% were metastatic at diagnosis. Sixty-eight (59.6%) patients had histological grade 3, 30 (26.3%) of which had non-endometrioid histology. Considering ESGO risk stratification, 63 (55.3%) patients were considered high-risk, supporting adjuvant treatment ( Table 1 ).

Evaluating ProMisE classification ( Table 1 ), we observed 18 (15.8%) POLEmut ( Supplementary Figure 2 ), 32 (28.1%) MMRd ( Supplementary Figure 3 ), 31 (27.2%) p53abn ( Supplementary Figure 4 ), and 33 (28.9%) NSMP EC. There was a statistical difference between the groups in age, FIGO stage, tumor grade, histological subtype, ESGO risk classification, lymph node status, recurrence, and disease status. Namely, the recurrence rate, a significant unfavorable association was found for the p53abn and NSMP subgroup (p=0.013) ( Table 1 ). Furthermore, the cancer mortality rate was more prevalent in the p53abn subgroup, with 29.0% of deaths (p=0.006) ( Table 1 ).

Next, we compared the classification of the traditional ProMisE with the molecular approach ( Table 2 ). Among the 114, we had 10 (8.7%) with discordant results between the MMR and MSI status ( Supplementary Figure 5 and Table 3 ). Eight cases depicted the loss of one or more MMR proteins but showed an MSS phenotype, and two cases displayed an MSI-H status despite the expression of all MMR proteins ( Table 3 ). Concerning TP53 status, among the 114, we observed 31 (27.2%) mutated cases ( Supplementary Figure 6 ) and found six discordances methodologies ( Table 4 ). Overall, the traditional and molecular PROMISE approaches showed a concordance rate of 86.8% (99/114 cases) with an overall accuracy of 0.87 and a Kappa coefficient of 0.82 ( Table 5 ).

We further evaluated the prognostic impact of ProMisE classification, both traditional and molecular approaches. After a median follow-up of 23.2 months (± 10.6 months), a statistically significant difference was observed in terms of OS (p=0.027) and PFS (p=0.0034) related to the four molecular subtypes using the ProMisE classifier ( Figure 1 and Table 6 ). Assessed by molecular classification, both outcomes maintain statistically significant differences with curves very similar to those of the traditional methodology (OS: p = 0.0036; PFS: p = 0.0023) ( Figure 1 and Table 6 ). In this prospective cohort, attention is drawn to the excellent overall survival not only of the POLEmut subgroup, but also of MMRd and NSMP.

In the present study, we implemented the ProMisE classifier in a Brazilian prospective cohort of patients with EC from a tertiary cancer hospital. Furthermore, we proposed comparing this traditional methodology with a molecular methodology to evaluate its accuracy and reproducibility.

We observed that the proportion of patients in each of the four subgroups considering the ProMisE classifier was 15.8% POLEmut, 28.1% MMRd, 27.2% p53abn, and 28.9% NSMP. Overall, our findings align with the range reported in populations, namely from North America, Europe, Asia, and Oceania, which showed prevalence rates of these subgroups of 4-15% POLEmut, 17-38% MMRd, 9-25% p53abn, and 40-64% NSMP (6, 8, 14, 22–25). Proportionally, we had fewer cases in the NSMP subgroup and slightly higher cases in the p53abn subgroup. Further studies are needed to elucidate these differences, mainly the higher frequency of p53abn cases. Nevertheless, we can hypothesize that it could be due to the reference nature of our hospital, which receives cases with more advanced diagnoses (26). Additionally, it could be related to distinct exposure or the admixture ethnicity of the Brazilian population, as we recently reported the higher frequency of TP53 mutations in lung cancer patients associated with African ancestry (21). Notably, the p53abn subgroup classification was a significant factor for both PFS and OS, exhibiting the poorest outcomes.

Considering the histological subtypes and their molecular profile, we had 84 patients (73.7%) with endometrioid histology, 17.9% of these in the POLEmut subgroup, 36.9% in the MMRd subgroup, 10.7% in the p53abn subgroup, and 34.5% in the NSMP subgroup. The other 30 patients (26.3%) had non-endometrioid histologies (serous, clear cell, carcinosarcoma), with a prevalence of molecular profile represented as follows: 10.0% POLEmut, 3.3% MMRd, 73.4% p53abn, and 13.3% NSMP subgroup. Compared with TCGA data, the enormous prevalence of patients with p53 mutation is represented by non-endometrioid histologies, and less than 5% of patients with POLE mutation are in this histological subgroup. The present cohort has less than a third of patients when compared to the TCGA study, which makes it difficult to perform more detailed subgroup analyses such as that carried out in the pivotal study (low-grade endometrioid versus high-grade endometrioid; serous versus clear cell versus carcinosarcoma) (5, 27). Further studies in a more extensive Brazilian series of endometrioid and non-endometrioid histologies are warranted to compare the differences in molecular subgroups among them.

In our study, the concordance and accuracy comparing ProMisE and molecular classifiers were 86.8% and 87%, respectively. Two studies, one Canadian led by Jessica McAlpine (8), and the other Chinese led by Jinaliu Wang (28), evaluated the comparison between two methodologies, the traditional ProMisE classifier and another using next-generation sequencing (NGS) molecular classifier (ProMisE NGS) and found slightly higher values, probably due to higher number of cases assessed. The Canadian study identified a concordance rate of 97% (159/164), whereas in the Chinese study, this rate was 94.1% (451/479) (8, 28). Concerning the MSI, we evaluated using a PCR-based approach previously validated at BCH for distinct tumor types (18, 20, 29, 30). The concordance rate was 86.8% (99/114) and is in accordance with reported discrepancies in methodological agreement in endometrial cancer. Dedeurwaerdere and colleagues evaluated three different molecular techniques, including one using PCR, to compare with immunohistochemistry results to define colorectal and EC patients with MSI. The concordance rate for patients with colorectal cancer was 100%, while for patients with EC, it ranged from 58% to 75% (31). A Spanish study also evaluated different methodologies for defining MMR status (IHC, PCR, NGS) in EC. The results showed discordance rates among the three techniques (32), which in part could be due to the known discordant behavior of the MSH6 pattern loss (33). McConechy et al. demonstrated a 93.3% concordance rate comparing pentaplex mono and di-nucleotide PCR tests for MSI with MMR status by IHC, suggesting that further studies are needed to define the best diagnostic technique for MMRd in EC (34).

The concordance rate for p53 by IHC and TP53 sequencing by NGS was 90.3%, similar to other reports. Singh et al. conducted a study with just over 200 patients, comparing p53 status by IHC and TP53 sequencing by NGS, demonstrating an overall agreement rate of 92.3% (35). Another study evaluating patients with high-risk EC in the PORTEC-3 trial reported a 90.7% concordance rate between p53 IHC and TP53 NGS analysis (36).

Although the ProMisE methodology has been well documented internationally (6), there has never been a report of its use in Brazilian endometrial cancer patients. Furthermore, we conducted a second classification, based only on molecular methodologies, and carried it out in a public hospital in Brazil. Additionally, this is a prospective cohort where we included all cases referred for treatment at the BCH, thus minimizing selection bias that could influence the results demonstrated. Of note, the entire therapeutic definition of the patients was discussed in a weekly tumor board of the gynecologic oncology department of the aforementioned hospital, thus minimizing possible influences on the oncological outcomes of these patients.

Despite our study’s significant findings and relevance, it also exhibits some limitations. The number of patients included is a limiting factor for a large and significant statistical association, as is the short follow-up time to define oncological outcomes in patients with EC. Furthermore, the molecular approach using distinct methodologies is not ideal, and we are currently working on a single NGS assay to assess all the biomarkers of the ProMisE classifier.

In this prospective cohort of endometrial cancer treated at a single public institution in Brazil, we successfully implemented the ProMisE classification for all patients and demonstrated its impact on stratifying patients’ outcomes. The molecular subgroups were in line with the international literature, with a slight increase in the proportion of TP53 mutated patients. Furthermore, with the current development of more comprehensive NGS panels, with the inclusion of POLE, TP53 genes, MSI, and other putative actionable endometrial genes, we foresee that a single NGS methodology would be the most effective, quick, objective, and cost-effective approach for endometrial cancer molecular classification.