Prognostic value of the TCGA molecular classification in uterine carcinosarcoma

Abstract Background The TCGA molecular groups of endometrial carcinoma are “POLE‐mutated” (POLEmut), “microsatellite‐instable/mismatch repair‐deficient” (MSI/MMRd), “TP53‐mutated/p53‐abnormal” (TP53mut/p53abn), and “no specific molecular profile” (NSMP). Objective Prognostic assessment of the TCGA groups in uterine carcinosarcoma (UCS). Search strategy Systematic review from January 2000 to January 2021. Selection criteria Studies assessing the TCGA groups in UCS. Data collection and analysis Progression‐free survival (PFS) and overall survival (OS) were assessed by Kaplan–Meier and Cox analyses (reference: TP53mut/p53abn group) and compared with endometrioid and serous carcinomas (original TCGA cohort), with a significant P < 0.050. Main results Five studies with 263 UCS were included. Compared with TP53mut/p53abn UCS, MSI/MMRd UCS showed significantly better PFS (P < 0.001) but similar OS (P = 0.788), whereas NSMP UCS showed similar PFS (P = 0.936) and OS (P = 0.240). Compared with their endometrioid/serous counterparts, NSMP and TP53mut/p53abn UCS showed significantly worse PFS (P < 0.001 and P = 0.004) and OS (P < 0.001 and P < 0.001), while MSI/MMRd UCS showed similar PFS (P = 0.595) but significantly worse OS (P < 0.001). The POLEmut group showed neither recurrences nor deaths in both the UCS and the endometrioid/serous carcinoma cohorts. Conclusion POLEmut UCS show excellent prognosis, whereas TP53mut/p53abn and NSMP UCS show a prognosis even worse than that of TP53mut/p53abn endometrioid/serous carcinomas. The prognosis of MSI/MMRd UCS remains to be defined.


| INTRODUC TI ON
Uterine carcinosarcoma (UCS) is a biphasic neoplasm constituted by an epithelial component and a stromal component, both of which are malignant and typically high-grade. The epithelial component reflects the endometrial carcinoma histotypes, such as serous (most common), endometrioid, clear cells, mixed, and undifferentiated. The stromal component may be "non-otherwise specified" or reflect the typical uterine sarcomas; in these cases, the stromal component is defined as "homologous". If a differentiation towards extrauterine mesenchymal tissues is present, the stromal component is defined as "heterologous". 1,2 The classification of UCS has changed over time. Indeed, UCS had previously been listed among the "mixed Müllerian tumors" and was lumped together with uterine sarcomas in terms of staging and treatment. 3,4 In recent decades, it has emerged that UCS represents a carcinoma with secondary sarcomatous transformation. 2 Therefore, UCS is now classified as a subtype of endometrial carcinoma and is staged and managed accordingly. 1,5,6 The rise of The Cancer Genome Atlas (TCGA) molecular classification has brought about a revolution in the prognostic stratification of endometrial carcinoma. The TCGA study and subsequent studies have identified four molecular prognostic groups: the "POLEmutated" (POLEmut) group, characterized by excellent prognosis; the "microsatellite-instable/mismatch repair-deficient" (MSI/MMRd) group and the "no specific molecular profile" (NSMP) group, characterized by intermediate prognosis; and the "TP53-mutant/p53-abnormal" (TP53mut/ p53abn) group, characterized by poor prognosis. [7][8][9][10][11][12] However, the original 2013 TCGA study only included endometrioid and serous carcinomas, 7 and the prevalence and prognostic value of the TCGA groups have been shown to vary across the different histotypes. [13][14][15][16][17] Among UCS, the vast majority (>70%) fall into the TP53mut/ p53abn group, which is consistent with its aggressive behavior. 16 However, the prognostic value of the other TCGA groups in UCS is not well defined. Moreover, it is unclear if TP53mut/p53abn UCS have a prognosis similar to their endometrioid and serous counterparts. The aim of this quantitative systematic review was to assess the prognostic value of the TCGA groups in UCS and to compare the prognosis of the TCGA groups between UCS and the original 2013 TCGA cohort.

| MATERIAL S AND ME THODS
Methods of this study were defined a priori, based on previous studies. 16,18 Each review stage was carried out by two independent authors; disagreements, if any, were resolved by consensus. The PRISMA statement 19 was followed to report this study.

| Search strategy and study selection
Four electronic databases (Scopus, ISI Web of Science, PubMed, Google Scholar) were searched from January 2000 to January 2021 for all studies assessing the TCGA molecular signatures in UCS. We adopted the following word combination: (uterine OR endometrial) AND (carcinosarcoma OR mixed malignant Müllerian) AND (TCGA OR ProMisE OR ultramutated OR POLE OR hypermutated OR mismatch OR microsatellite OR copy number OR TP53 OR p53). We also assessed the reference lists of eligible studies. We included all studies that assessed the TCGA groups and survival outcomes in UCS. Given the rarity of POLEmut UCS, 16 we also included studies that did not assess POLE status. Exclusion criteria, defined a priori, were: sample size less than 10, data not extractable, reviews.

| Risk of bias assessment
The QUADAS-2 20 were used to assess the risk of bias within studies, as previously described. 16,18 Four domains were assessed: (1) "patient selection" domain (were selection criteria and period of enrollment reported and adequate?); (2) "index test" domain (were immunohistochemical and/or molecular methods reported and adequate); (3) "reference standard" domain (were follow-up data reported and adequate?); and (4) "flow" (were all patients assessed for the TCGA classification and survival outcomes?). The risk of bias was categorized as "low", "unclear", or "high" as previously described. 16
The POLEmut group was assessed in three studies by using POLE sequencing. The MSI/MMRd group was assessed by MMR immunohistochemistry in three studies and by MSI testing in two studies.
The TP53mut/p53abn group was assessed by p53 immunohistochemistry in two studies, by TP53 sequencing in two studies and by both methods in the remaining study. Characteristics of the included studies are reported in Table 1.

| Risk of bias assessment
For the patient selection domain, three studies were considered at low risk of bias and two studies at unclear risk (period of enrollment not reported). For the index test domain, three studies were considered at low risk of bias and two studies at unclear risk (POLE mutations not assessed). For the "reference standard" domain, all studies were considered at low risk of bias. For the "flow" domain, the risk of bias was considered low for four studies and unclear for one study (only part of the cases was assessed by MMR immunohistochemistry). Authors' judgments are reported in Figure S2

| DISCUSS ION
This study showed that POLEmut UCS had the same excellent prognosis as POLEmut endometrioid carcinomas, whereas TP53mut/ p53abn and NSMP UCS had a prognosis even worse than TP53mut/ In the last decades, the risk stratification of endometrial carcinoma has been based on poorly reproducible histologic features and on simplistic clinical classifications. [26][27][28][29] Since their publication in 2013, the TCGA findings have progressively changed the approach to the study of endometrial carcinoma. 6,7,30 The "Proactive Molecular Risk Classifier for Endometrial Cancer" (ProMisE) is a surrogate of the TCGA classification, which has allowed a wider applicability of the TCGA groups. 8

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
The study was conceived by AT, AR, DR, DA, AS, RS, and GFZ and  Abbreviations: CI, confidence interval; HR, hazard ratio; UCS, uterine carcinosarcoma. a NV, not evaluable, because no event occurred in POLEmut cases.