Targeted Molecular Testing in Endometrial Carcinoma: Validation of a Clinically Driven Selective ProMisE Testing Protocol : International Journal of Gynecological Pathology

Secondary Logo

Journal Logo

Pathology of the Corpus: Original Articles

Targeted Molecular Testing in Endometrial Carcinoma: Validation of a Clinically Driven Selective ProMisE Testing Protocol

Talhouk, Aline Ph.D; Jamieson, Amy M.B.Ch.B., F.R.A.N.Z.C.O.G., F.R.C.S.C.; Crosbie, Emma J. B.Sc., M.B.Ch.B., Ph.D., F.R.C.O.G.; Taylor, Alexandra M.B.B.S., M.R.C.P., F.R.C.R., M.D.; Chiu, Derek M.Sc.; Leung, Samuel M.Sc.; Grube, Marcel M.D.; Kommoss, Stefan M.D.; Gilks, C. Blake M.D., F.R.C.P.C.; McAlpine, Jessica N. M.D.; Singh, Naveena M.D., F.R.C.Path., F.R.C.P.C.

Author Information
International Journal of Gynecological Pathology 42(4):p 353-363, July 2023. | DOI: 10.1097/PGP.0000000000000898
  • Free


Incorporation of molecular classification into clinicopathologic assessment of endometrial carcinoma (EC) improves risk stratification. Four EC molecular subtypes, as identified by The Cancer Genome Atlas, can be diagnosed through a validated algorithm Proactive Molecular Risk Classifier for Endometrial Cancer (ProMisE) using p53 and mismatch repair (MMR) protein immunohistochemistry (IHC), and DNA polymerase epsilon (POLE) mutational testing. Cost and access are major barriers to universal testing, particularly POLE analysis. We assessed a selective ProMisE algorithm (ProMisE-S): p53 and MMR IHC on all EC’s with POLE testing restricted to those with abnormal MMR or p53 IHC (to identify POLEmut EC with secondary abnormalities in MMR and/or p53) and those with high-grade or non-endometrioid morphology, stage >IA or presence of lymphovascular space invasion (so as to avoid testing on the lowest risk tumors). We retrospectively compared the known ProMisE molecular classification to ProMisE-S in 912 EC. We defined a group of “very low-risk” EC (G1/G2, endometrioid, MMR-proficient, p53 wild-type, stage IA, no lymphovascular space invasion) in whom POLE testing will not impact on patient care; using ProMisE-S, POLE testing would not be required in 55% of biopsies and 38% of all EC’s, after evaluation of the hysterectomy specimen, in a population-based cohort. “Very low-risk” endometrioid EC with unknown POLE status showed excellent clinical outcomes. Fifteen of 166 (9%) of all p53abn EC showed G1/G2 endometrioid morphology, supporting the potential value of universal p53 IHC. The addition of molecular testing changed the risk category in 89/896 (10%) EC’s. In routine practice, POLE testing could be further restricted to only those patients in whom this would alter adjuvant therapy recommendations.

The traditional binary algorithm for classifying endometrial carcinoma (EC) continues to be mirrored in international clinical guidelines, wherein low-grade endometrioid EC’s are separated from high-grade and nonendometrioid histotypes 1. While this simplistic classification has stood the test of time to a large extent, some tumors have been difficult to categorize owing to either ambiguous morphology or discrepant clinical behavior 2–4.

In 2013 the Cancer Genome Atlas (TCGA) demonstrated that ECs can be segregated into 4 different molecular clusters, as determined by integration of genomic, transcriptomic, and proteomic analysis: (i) ultramutated, characterized by hotspot mutations in DNA polymerase epsilon (POLE) and a markedly increased transversion mutation frequency; (ii) hypermutated, characterized by microsatellite instability and a 10-fold higher mutation rate than microsatellite stable tumors; (iii) copy number-high, composed of serous-like EC with extensive somatic copy number alterations and low tumor mutational burden; and (iv) copy number-low, composed largely of endometrioid, microsatellite stable tumors with relatively low somatic copy number alterations 5.

The validated Proactive Molecular Risk Classifier for Endometrial Cancer (ProMisE) demonstrated that by using immunohistochemistry (IHC) for p53, the protein product of the oncogene TP53, and mismatch repair proteins (MMR), and sequencing the exonuclease domain of the enzyme DNA polymerase epsilon (POLE), 4 molecular groups can be identified that replicate those described by TCGA 6–8. Although these 4 groups are mostly nonoverlapping, in <5% of ECs, more than 1 molecular feature may be present 9. The reason for this is that both POLE mutation and MMR deficiency result in high mutational states that can cause secondary or passenger mutations, in each other and in TP53, that do not influence prognosis in the same way as driver mutations. In other words, EC showing abnormal p53 IHC results or harboring TP53 mutations in the presence of a pathogenic POLE mutation or MMR defect, show similar clinical behaviour as POLEmut or mismatch repair deficient (MMRd) EC, respectively 9.

Initially the ProMisE algorithm first identified those that were MMRd EC, followed by those with mutations in the exonuclease domain of POLE (POLEmut EC), and finally classification of the remainder according to p53 IHC results. Having MMRd as the first step allowed all EC patients to be screened for Lynch syndrome. Since the validation of ProMisE, further research has led to refinement of the original algorithm. Firstly, it has been shown that only specific pathogenic POLE mutations are associated with the “ultramutator” phenotype that defines the outstandingly good prognosis of this class of EC 10,11. Furthermore, tumors harboring pathogenic POLE mutations with MMRd show the same clinical behavior as POLEmut ECs 10. For these reasons, it is now agreed that knowledge of the presence or absence of a pathogenic POLE mutation should be the first step in the algorithm.

The revised algorithm incorporating this new information has been included in the 2020 World Health Organization Classification of Female Genital Tumours 12. This refined algorithm requires POLEmut EC to be identified first as those harboring pathogenic POLE mutations, followed by MMRd EC, with the remainder classified as No Specific Molecular Profile (NSMP) EC or p53abn EC on the basis of normal versus abnormal p53 IHC. With superior reproducibility to conventional histologic classification, as well as better prognostication and prediction of treatment response, the new molecular classification has been incorporated into clinical guidelines developed by the National Comprehensive Cancer Network 13 and those produced jointly by the European Societies of Gynecological Oncology, Pathology and Radiotherapy and Oncology (ESGO/ESTRO/ESP) 14. Many factors will determine how readily we can incorporate these changes worldwide and, as we enter this transition phase, it is important to have agreement on possible testing strategies and reporting terminology. While the assessment of MMR and p53 IHC are widely available, the evaluation of POLE mutational status currently presents a major barrier to widespread adoption.

In this study, we proposed and retrospectively evaluated the prognostic value of a clinically driven, selective ProMisE classifier (ProMisE-S), whereby MMR and p53 IHC are evaluated on all EC’s, and POLE mutational analysis is only assessed in patients in whom adjuvant treatment may be considered (Figure 1). For the purposes of this study, EC’s fulfilling all of the following criteria: grade 1 or 2, endometrioid histology, p53 normal, MMR proficient, stage IA and no lymphovascular space invasion (LVSI), were labeled “very low-risk” with no further testing, and POLE status was only evaluated in the remaining tumors. In addition, we compared this more pragmatic molecular classification of EC (ProMisE-S) with universal testing using the ProMisE algorithm, the present gold standard. Finally, we compared the allocation of EC patients into risk groups as defined without and with molecular classification by the recent ESGO/ESTRO/ESP guidelines.

FIG. 1:
Selected testing protocol. On diagnostic evaluation, the restricted testing protocol would require the evaluation of histotype, grade, MMR, and p53 IHC status to determine who should receive full ProMisE molecular characterization. Postsurgically, patients with >stage IA and with any LVSI would also receive full molecular characterization. IHC indicates immunohistochemistry; MMR, mismatch repair; LVSI, lymphovascular space invasion.


Cohorts and Data

We retrospectively analyzed data on cohorts from Vancouver (Canada) and Tuebingen (Germany) previously described in the development 6, confirmation 7, and final validation 8 of the ProMisE molecular classifier. All clinicopathologic data analyzed were based on findings posthysterectomy, and no biopsy-hysterectomy correlation or histopathologic review was performed. The cohort used in the development and confirmation represented patients referred to a tertiary cancer center and treated at Vancouver General Hospital and BC Cancer, unlike the Tuebingen validation cohort which was population-based (all comers). We characterized molecular subtype as previously published and in 2 additional ways: the first used the new WHO-endorsed ProMisE, starting with the identification of POLE pathogenic mutations, followed by assessment of MMR status by IHC, and finally considering the p53 IHC result. The second approach considered, termed ProMisE-S, aimed to fully subtype a targeted subset and identify a “very low-risk” group that could safely forego POLE testing (low-grade endometrioid EC, normal MMR and p53 IHC, stage IA, negative for LVSI). This approach could potentially be applied preoperatively at first, by identifying patients as “likely low-risk” based on evaluation of histotype, grade, and p53 and MMR IHC on diagnostic biopsy specimens (Fig. 1). Consideration of postsurgical pathology findings would result in additional patients requiring full molecular classification; after surgery, EC’s with normal p53 and MMR IHC who were stage IA with no LVSI, and remained endometrioid grade 1 or 2 would be considered “very low-risk” and would not require POLE assessment. We estimated the “likely low-risk” group based on histotype, grade, and MMR and p53 IHC, and the “very low-risk group” based on all histopathologic findings, ie including stage and LVSI.

New ESGO/ESTRO/ESP guidelines have proposed 2 definitions of prognostic risk-groups for postsurgical decision-making 14. One version is based on clinicopathologic parameters only and the other integrates molecular classification. For all patients in our cohort, we assigned ESGO/ESTRO/ESP risk groups with and without molecular classification. Where guidelines do not provide a classification with molecular subtype (as with NSMP clear cell and other nonendometrioid histotypes, and high-stage POLEmut EC) or where subtype information was not available (as in the case of the ProMisE-S “very low risk” group), clinicopathologic risk groups as assigned in the absence of molecular classification were retained. We compared these classifications for the scenarios with (i) no molecular classification available, (ii) molecular data available for a subset of patients, that is all those not classified as “very low-risk” (ProMise-S), and (iii) full molecular classification available for all patients (ProMisE).

Statistical Methods

We used cross tabulation to compare the new WHO-endorsed refined ProMisE to the previously validated molecular classification algorithm, as well as to a targeted ProMisE-S. We characterized clinicopathologic characteristics of the preoperative ProMisE-S group using descriptive statistics. Univariable associations of individual clinicopathologic parameters with ProMisE-S preoperatively were compared using 1 way analysis of variance for continuous data (age at surgery, body mass index) and the χ2 test for categorical data [stage, grade, histologic subtype at diagnosis, LVSI, myometrial invasion, and lymph node status], and the ESMO (2020) risk group assignment, with and without molecular classification.

We considered univariable and multivariable survival of the different ProMisE variations to address the association with overall survival (OS), disease‐specific survival (DSS), and progression‐free survival (PFS). We assessed the significance of Kaplan-Meir survival curves using log rank tests in univariable survivals. For multivariable models we used cox proportional hazard, adjusting for age, body mass index, grade, histotype, and adjuvant treatment (any or none). We reported hazard ratios with corresponding 95% confidence intervals and likelihood ratio test P values. We used the Firth penalized maximum‐likelihood bias‐reduction method to estimate the hazard ratios (indicated by a superscript F) when the number of censored cases exceeded 80%, and the corresponding confidence intervals were obtained using the profile likelihood.

We used the R software for statistical computing 15. All statistical hypothesis tests performed were 2‐sided. Statistical significance was set at α=0.05, and we did not correct for multiple comparisons.


Data from 912 patients were used to evaluate the refined WHO-endorsed ProMisE and a simplified, targeted ProMisE-S algorithm. These data were obtained from previously published cohorts of which 460 were from Vancouver, Canada and 452 were from Tuebingen, Germany. Reclassification from the previously validated algorithm to the refined WHO endorsed ProMisE classifier was possible in 899 patients which we used for the remaining analyses. The change in the ordering of the classification tree left 13 patients unclassifiable; these were patients previously identified as MMRd but for whom we did not have confirmation of a pathogenic POLE mutation. Reclassification of the remaining 899 resulted in very few (7/899) tumors changing to a different molecular group: 2 changed from MMRd to POLEmut, 4 from POLEmut to NSMP, and 1 from POLEmut to p53abn (Table 1). In the first 2, the deficiency in MMR is considered secondary to the pathogenic POLE mutation, and in the latter 5 tumors, the POLE mutations were nonpathogenic.

TABLE 1 - Comparison of molecular classification results obtained from the previously validated ProMisE with the refined WHO endorsed molecular classification
Refined ProMisE, n (%)
POLEmut MMRd NSMP/p53wt p53abn Total, n (%)
Overall Pop based Overall Pop based Overall Pop Based Overall Pop based Overall Pop based
 POLEmut 77 (97.5) 39 (97.5) 4 (0.9) 2 (0.9) 1 (0.6) 1 (1.8) 82 (9.1) 42 (9.5)
 MMRd 2 (2.5) 1 (2.5) 226 (100) 120 (100) 228 (25.4) 121 (27.4)
 NSMP/p53wt 424 (99.1) 225 (99.1) 424 (47.2) 225 (50.9)
 p53abn 165 (99.4) 54 (98.2) 165 (18.4) 54 (12.2)
Total 79 (8.8%) 40 ( 9% ) 226 (25.1) 120 ( 27.1 ) 428 (47.6) 227 ( 51.4 ) 166 (18.5) 55 ( 12.4 ) 899 (100) 442 (100)
Bold values indicates the population-based cohort.
MMRd indicates mismatch repair deficient; ProMisE, Proactive Molecular Risk Classifier for Endometrial Cancer.

Of the 899 patients, 45% were deemed “likely low-risk” based on parameters available at the time of biopsy diagnosis (MMR and p53 IHC, histotype and grade). As expected, the proportion of “likely low-risk” patients was higher in the population-based German cohort (55%; Table 2), since the Vancouver cohort was biased toward high-risk patients reflective of tertiary referral center practice.

TABLE 2 - Comparison of molecular classification results obtained from a restricted molecular testing protocol versus testing every patient
Refined ProMisE (all patients), n (%)
POLEmut MMRd NSMP/p53wt p53abn Total
Overall Pop based Overall Pop based Overall Pop based Overall Pop based Overall Pop based
Restricted (preoperative) ProMisE-S
 Likely low-risk 47 (59.5) 33 (82.5) 358 (83.6) 210 (92.5) 405 (45.1) 243 (55)
 POLEmut 32 (40.5) 7 (17.5) 32 (3.6) 7 (1.6)
 MMRd 226 (100) 120 (100) 226 (25.1) 120 (27.1)
 NSMP/p53wt 70 (16.4) 17 (7.5) 70 (7.8) 17 (3.8)
 p53abn 166 (100) 55 (100) 166 (18.5) 55 (12.4)
Restricted (postsurgical) ProMisE-S
 Very low-risk 34 (43) 26 (65) 237 (55.4) 140 (61.7) 271 (30.1) 166 (37.6)
 POLEmut 45 (57) 14 (35) 45 (5) 14 (3.2)
 MMRd 226 (100) 120 (100) 226 (25.1) 120 (27.1)
 NSMP/p53wt 191 (44.6) 87 (38.3) 191 (21.2) 87 (19.7)
 p53abn 166 (100) 55 (100) 166 (18.5) 55 (12.4)
Total 79 (8.8) 40 ( 9 ) 226 (25.1) 120 ( 27.1 ) 428 (47.6) 227 ( 51.4 ) 166 (18.5) 55 ( 12.4 ) 899 (100) 442 (100)
Bold values indicates the population-based cohort.
MMRd indicates mismatch repair deficient; ProMisE, Proactive Molecular Risk Classifier for Endometrial Cancer.

As shown in Table 3, at the time of biopsy we would have identified 15 p53abn EC’s which were grade 1/2 with endometrioid histology; following surgery 10 of these patients were stage IA, 2 were stage IB and 3 were stage III. Within the “likely low-risk” group, 130/405 (32%) EC’s would need to undergo POLE testing after surgery because they were stage IB or higher (27%) and/or showing presence of LVSI (8%). Of these, 12/130 (9%) would be identified as POLEmut EC, and the remainder classified as NSMP EC. Patients classified postsurgically as “very low-risk’” made up 30% of the full cohort and 38% of population-based cohort (Table 2).

TABLE 3 - Clinicopathologic parameters according to a restricted preoperative ProMisE-molecular testing algorithm in all cohorts
Restricted preoperative ProMisE-S protocol
Likely low-risk POLEmut MMRd NSMP/p53wt p53abn Total P
Age, yr
 Median (IQR) 62 (54–72) 59 (56–67) 67 (59–74) 64 (56–73) 72 (66–78) 66 (57–74) <0.001
 Missing 0 0 1 2 0 3
 Median (IQR) 29 (24–37) 27 (23–29) 28 (23–33) 25 (22–31) 28 (23–32) 28 (24–34) <0.001
 Missing 32 5 12 9 5 63
FIGO 2009 stage, n (%)
 IA 294 (73) 14 (44) 107 (48) 26 (38) 61 (37) 502 (56) <0.001
 IB 68 (17) 14 (44) 57 (25) 14 (21) 20 (12) 172 (19)
 II-IV 42 (10) 4 (12) 61 (27) 28 (41) 85 (51) 220 (25)
 Missing 2 0 1 2 0 5
Tumor grade, n (%)
 Grade 1/2 405 (100) 3 (9) 137 (61) 4 (6) 15 (9) 564 (63) <0.001
 Grade 3 0 29 (91) 89 (39) 66 (94) 151 (91) 335 (37)
 Missing 0 0 0 0 0 0
Histologic subtype (reviewed)
 Endometrioid 405 (100) 21 (66) 196 (87) 51 (74) 39 (23) 712 (79) <0.001
 Serous 0 3 (9) 10 (4) 11 (16) 108 (65) 132 (15)
 Clear cell 0 1 (3) 4 (2) 2 (3) 5 (3) 12 (1)
 High grade mixed or other 0 6 (19) 12 (5) 4 (6) 14 (8) 36 (4)
 Undifferentiated 0 1 (3) 3 (1) 1 (1) 0 (0) 5 (1)
 Missing 0 0 1 1 0 2
LVSI, n (%)
 Positive 32 (8) 17 (55) 73 (33) 28 (43) 76 (48) 226 (26) <0.001
 Negative 365 (92) 14 (45) 146 (67) 37 (57) 81 (52) 643 (74)
 Missing 8 1 7 5 9 30
Nodal status, n (%)
 Positive 9 (2) 2 (6) 21 (10) 9 (13) 36 (23) 77 (9) <0.001
 Negative 266 (73) 25 (81) 166 (79) 40 (60) 102 (66) 599 (73)
 Not tested 88 (24) 4 (13) 23 (11) 18 (27) 17 (11) 150 (18)
 Missing 42 1 16 3 11 73
Postsurgical treatment, n (%)
 None 245 (66) 7 (22) 89 (43) 16 (24) 48 (32) 405 (49) <0.001
 Brachy only 86 (23) 5 (16) 46 (22) 7 (11) 7 (5) 151 (18)
 RT+/−brachy 26 (7) 6 (19) 32 (16) 17 (26) 21 (14) 102 (12)
 Chemo+/−RT 15 (4) 14 (44) 39 (19) 26 (39) 76 (50) 170 (21)
 Missing 33 0 20 4 14 71
Total 405 (45%) 32 (4%) 226 (25%) 70 (8%) 166 (18%) 899
BMI indicates body mass index; IQR, interquartile range; LVSI, lymphovascular space invasion; MMRd, mismatch repair deficient; ProMisE, Proactive Molecular Risk Classifier for Endometrial Cancer.

Of note, in the ProMisE-S algorithm, the groups labeled as NSMP and POLEmut correspond to subsets of these original ProMisE categories, that exclude those with the least aggressive clinicopathologic features (Table 2).

Survival results when comparing the original and refined ProMisE classifiers were very similar, as expected, given that only 7/899 of all EC’s changed classification (Table 1). Using the targeted testing with ProMisE-S, a slightly different prognostic trend emerges owing to the exclusion of patients with least aggressive forms of the disease. As would be expected, the “very low-risk” group had excellent OS, DSS and PFS outcomes rivaling outcomes in the POLEmut group, which in turn followed a similar OS and DSS as in the refined ProMisE algorithm, but with slightly inferior PFS (Fig. 2). The worst survival remained attributable to p53abn EC. However, we note that NSMP EC had worse survival outcomes relative to MMRd EC, this is likely because in this group, as defined by the ProMisE-S algorithm, the subset with the most favorable features were part of the “very low risk” group. Multivariable analyses which adjust for age, grade, histotype, LVSI, and stage show that all versions of the ProMisE classifier, original, refined as well as selective, remain prognostic, with the order of survival similar to that observed in univariable models (Table 4).

FIG. 2:
OS (DSS and PFS are provided for the 3 ProMisE algorithms: (i) refined ProMisE, as endorsed by WHO (2020), whereby the assessment of POLE status precedes that of mismatch repair (MMR), (ii) ProMisE-S based on tumor features that would be available at biopsy (histotype, grade and MMR and p53 IHC), and (iii) ProMisE based on hysterectomy findings, incorporating all of the following: histotype, grade, MMR and p53 IHC, stage, and LVSI. DSS indicates disease-specific survivival; OS, overall survival; PFS, progression-free survival; ProMisE, Proactive Molecular Risk Classifier for Endometrial Cancer.
TABLE 4 - Univariable and multivariable hazard ratios of molecular subtypes with p53abn set as reference
Univariable Multivariable*
# Of events/n Hazard ratio (95% CI) LRT, P # Of events/n Hazard ratio (95% CI) LRT, P
Overall survival
  NSMP/p53wt 232/912 0.23 (0.16–0.31)F <0.01 216/877 0.57 (0.36–0.9)F 0.05
   POLE mut 0.18 (0.09–0.31)F 0.51 (0.25–0.98)F
  MMRd 0.36 (0.26–0.49)F 0.61 (0.4–0.95)F
 Refined ProMisE
  NSMP/p53wt 229/899 0.23 (0.17–0.31)F <0.01 213/865 0.58 (0.36–0.91)F 0.06
   POLE mut 0.17 (0.08–0.3)F 0.49 (0.23–0.95)F
  MMRd 0.37 (0.26–0.51)F 0.62 (0.4–0.96)F
 ProMisE-S (postsurgical)
  NSMP/p53wt 229/899 0.39 (0.27–0.54)F <0.01 213/865 0.59 (0.37–0.95)F 0.07
   POLE mut 0.19 (0.08–0.38)F 0.4 (0.15–0.87)F
  MMRd 0.37 (0.26–0.51)F 0.63 (0.4–0.97)F
  Very low-risk 0.12 (0.08–0.18)F 0.63 (0.32–1.22)F
Disease-specific survival
  NSMP/p53wt 150/900 0.15 (0.1–0.23)F <0.01 136/866 0.63 (0.36–1.1)F <0.01
   POLE mut 0.07 (0.02–0.18)F 0.21 (0.06–0.57)F
  MMRd 0.31 (0.21–0.46)F 0.71 (0.42–1.17)F
 Refined ProMisE
  NSMP/p53wt 147/887 0.15 (0.1–0.23)F <0.01 133/854 0.64 (0.36–1.12)F 0.01
   POLE mut 0.07 (0.02–0.18)F 0.23 (0.06–0.62)F
  MMRd 0.31 (0.21–0.46)F 0.7 (0.41–1.16)F
 ProMisE-S (postsurgical)
  NSMP/p53wt 147/887 0.32 (0.21–0.48)F <0.01 133/854 0.67 (0.38–1.17)F 0.03
   POLEmut 0.13 (0.04–0.33)F 0.25 (0.07–0.69)F
  MMRd 0.31 (0.21–0.46)F 0.68 (0.4–1.13)F
  Very low-risk 0.03 (0.01–0.07)F 0.34 (0.11–0.99)F
Progression-free survival
  NSMP/p53wt 152/865 0.16 (0.1–0.23)F <0.01 141/836 0.49 (0.28–0.86)F <0.01
   POLE mut 0.09 (0.03–0.2)F 0.2 (0.06–0.5)F
  MMRd 0.32 (0.22–0.47)F 0.63 (0.38–1.03)F
 Refined ProMisE
  NSMP/p53wt 150/852 0.16 (0.1–0.23)F <0.01 139/824 0.48 (0.27–0.84)F <0.01
   POLE mut 0.09 (0.03–0.21)F 0.21 (0.07–0.54)F
  MMRd 0.33 (0.22–0.49)F 0.64 (0.38–1.05)F
 ProMisE-S (postsurgical)
  NSMP/p53wt 150/852 0.3 (0.2–0.46)F <0.01 139/824 0.51 (0.28–0.89)F <0.01
   POLEmut 0.16 (0.05–0.38)F 0.25 (0.08–0.63)F
  MMRd 0.33 (0.22–0.49)F 0.62 (0.37–1.02)F
  Very low-risk 0.05 (0.02–0.1)F 0.29 (0.11–0.73)F
*Adjusted for age, grade, histotype, LVSI and stage, F.
F indicates Firth bias correction; LRT, likelihood ratio test; MMRd, mismatch repair deficient; ProMisE, Proactive Molecular Risk Classifier for Endometrial Cancer.

We used ESGO/ESTRO/ESP guidelines 14 to stratify all patients in our cohorts into risk groups. Molecular classification would have resulted in 89/896 (10%) patients being assigned to a different risk category (Table 5). Without molecular classification 35/89 would be assigned to a lower risk group representing patients who would potentially be undertreated. Conversely, 54/89 would be assigned to a higher risk group representing patients who would potentially be overtreated. Using the ProMisE-S classification all patients would have been placed within the appropriate ESGO/ESTRO/ESP risk category.

TABLE 5 - Risk categorization of endometrial carcinoma with and without molecular classification
With refined ProMisE molecular classification, n (%)
Low Intermediate High-intermediate High Advanced metastatic Total
Without molecular classification
 Low 349 (89.9) 6 (4.6) 3 (1.3) 358 (40.0)
 Intermediate 10 (2.6) 117 (90.0) 5 (2.2) 132 (14.7)
 High-intermediate 21 (5.4) 0 (0.0) 123 (93.9) 21 (9.3) 165 (18.4)
 High 8 (2.1) 7 (5.4) 8 (6.1) 198 (87.2) 221 (24.7)
 Advanced metastatic 20 (100.0) 20 (2.2)
Total 388 (43.3) 130 (14.5) 131 (14.6) 227 (25.3) 20 (2.2) 896 (100)


The comprehensive genomic analysis of EC published by the TCGA discovered 4 nonoverlapping and prognostically relevant molecular groups 5, which have been amply validated in several retrospective series 7,8,16–18. This classification goes a long way toward explaining the deficiencies of a simplistic binary classification of EC, wherein many tumors are difficult to classify and there is imperfect correlation with clinical outcomes 17,19,20. For this reason the WHO and clinical guidelines recommend the use of molecular classification alongside traditional risk assessment, and the combined ESGO/ESTRO/ESP guidelines have provided algorithms for risk categorization both without and with knowledge of the molecular class 12,14,21.

The benefits of this combined approach are, firstly, superior diagnostic reproducibility between pathologists and laboratories 21 and between biopsy and hysterectomy specimens 22, as compared with morphologic classification alone 4. Secondly, it has been demonstrated that molecular classification alongside traditional clinicopathologic risk assessment results in more accurate prediction of risk of recurrence and death than either system alone, providing more meaningful information for treatment planning 6,14. p53abn EC’s account for 20% of EC’s 23 and 50% to 70% of EC mortality 24, and are heterogeneous in regard to histotype and grade 25. On the other hand, POLEmut EC ‘s, account for 6% of EC’s overall 23 and have outstandingly good clinical outcomes with exceedingly low rates of recurrence/death and high salvage rates for recurrences, regardless of morphologic features and stage at presentation 11. Thirdly, molecular classification can be performed on endometrial biopsies, providing prognostic information earlier in the patient pathway 21,22. Fourthly, molecular subtypes differ in their response to conventional treatments 18. Finally, and most importantly, molecular classification is the basis for personalizing the treatment of EC; the major gain of applying this system lies in correct identification of patients for the most appropriate therapy, whether this may be avoidance of any adjuvant treatment, the recommended approach for stage I and II POLEmut EC patients, or the use of chemotherapy for all p53abn EC with myoinvasion, or a range of targeted therapies 19,20,26. Many questions remain unanswered, such as the prognostic impact of grade, histotype, and additional genomic changes within individual molecular groups.

We have previously developed 6, confirmed 7, and validated 8 the ProMisE molecular classifier that serves as a surrogate system for identification of the TCGA molecular classes using simple and accessible laboratory tests. The algorithm recommended by the WHO, based on subsequent research, represents a minor modification of the ProMisE approach, wherein accurate interpretation requires knowledge of pathogenic POLE mutation status, followed by the MMR status and, lastly, the p53 result 12. This algorithm thereby implies that POLE testing is mandatory for the accurate classification of all patients.

There are major barriers to universal implementation of molecular classification. The prospective evaluation of this classification and its therapeutic implications remain to be seen and are the subject of numerous clinical trials worldwide. In the absence of prospective trial results, however, several clinical implications are already part of routine care, such as the need for referral to Clinical Genetics Services and consideration of immunomodulatory treatments in certain subsets of patients with MMRd EC. On the basis of the current state of knowledge, there are many potential implications for prognosis and clinical management. Financial constraints, a major barrier, require pathologists and laboratories to be looking to refine their diagnosis as much as possible through judicious use of available resources. While MMR and p53 IHC are relatively readily available, access to POLE testing, currently only achievable through gene sequencing and lacking an immunohistochemical surrogate, remains a challenge. Cheaper and accessible testing options for POLE testing are therefore needed, and in addition, there is a need to examine whether testing is necessary on all patients, or whether an amended protocol can diminish the burden of testing.

In this study we assessed the extent to which an amended ProMisE algorithm (termed ProMisE-S) can reduce the number of EC’s needing POLE mutational testing. We evaluated a selective testing algorithm, based on test accessibility, whereby MMR and p53 IHC would be carried out on all biopsies; with POLE status only assessed after excluding patients deemed “likely low-risk”: grade 1 or 2, endometrioid histology, p53 normal, MMR proficient. Following surgical staging, there would be no clinical benefit from knowing the POLE status of a “very low-risk” group, defined as fulfilling all of the above parameters as well as being stage IA, grade 1 or 2, with absence of LVSI. POLE testing would be recommended in all of the other substage and histotype designations for accurate risk assessment through complete molecular classification, where adjuvant treatment recommendations would differ based on POLE status.

Our findings were, firstly, that universal p53 IHC identified 15/166 (10%) p53abn ECs that were reported as grade 1/2 endometrioid tumours. These constituted <3% (15/564) of all grade 1/2 endometrioid EC in this cohort. We recommend universal p53 IHC in this algorithm to prevent missing any potential high-risk cases. It remains unknown whether p53abn grade 1/2 “endometrioid” EC have a prognosis and response to treatment equivalent to the p53abn EC group as a whole and thus are candidates for adjuvant chemotherapy. We are unable to address this issue satisfactorily based on this series as there are only 15 such cases and no slide review was performed to confirm the morphologic diagnosis; this is, however, and important area for future study. Only if the p53abn grade 1/2 “endometrioid” EC have the poor prognosis of p53abn EC would the expense of universal p53 IHC be justified.

It is also possible that the grade 1/2 EC with mutant pattern p53 staining can be recognized based on their nuclear features, as suggested by Kang et al. 27. In their study they were able to detect p53abn EC based on nuclear features, including smudged chromatin, pleomorphism, atypical mitoses, and tumor giant cells, and this allowed for identification of cases that have mutant pattern p53 expression with a high degree of sensitivity (98.5%). This work, if validated, suggests that a selective approach to performing p53 IHC could be used, but it is important to note that their approach did not detect subclonal mutant pattern p53 expression (only 33% sensitivity) and this could limit its applicability, should it be demonstrated that subclonal p53 expression is equivalent to uniform p53abn. Studies on the clinical significance of subclonal p53 expression are needed to address this issue, but it should be noted that subclonal p53 expression, with mutant pattern expression of p53 in >10% of tumor cells, was considered to be p53abn in previous studies demonstrating the poor prognosis associated with this molecular subtype 18. As already stated, no histologic review was carried out in our study and we are thus unable to validate the findings of Kang et al. 27.

Several studies have previously reported a low proportion of p53 abnormal IHC or TP53 mutation within morphologically low-grade EC with varying frequency 18,27–29, and this finding has been reported to have prognostic significance 28,29. As evaluation of nuclear features remains somewhat subjective, the value of carrying out universal p53 is to allow for the variation in thresholds for recognition of nuclear atypia and avoid missing p53abn EC’s that show deceptively bland morphology, as would occur if molecular classification were to be restricted to high-grade and high-risk patients.

Our second finding was that using the ProMisE-S protocol at the time of diagnostic biopsy, POLE testing would be indicated on all patients other than those deemed “likely low-risk,” constituting 55% of the total. The rationale for carrying out POLE testing on p53abn and MMRd EC is to exclude cases showing these abnormalities secondary to an underlying POLE mutation, ie to exclude multiple classifiers and validate the p53 and MMR IHC result. The rationale for testing high-grade endometrioid and nonendometrioid cases would be to unmask potentially low-risk EC masquerading morphologically as high-risk. Thirdly, 134/405 (33%) of EC’s considered to be “likely low-risk” based on findings potentially available at the time of biopsy (namely histotype, grade if endometrioid, MMR and p53 IHC results) were found to be stage IB or higher and/or positive for LVSI; in these patients POLE testing would be recommended post-surgery, with possible impact on adjuvant treatment decisions. Overall, the adoption of the targeted ProMisE- algorithm would reduce the proportion of EC’s requiring POLE testing by a minimum of ~40%. In real life this proportion could be higher still for at least 2 reasons: this additional testing could be restricted to those patients in whom this information would truly impact on treatment decision-making, excluding those in whom adjuvant treatment would either be given or withheld, regardless knowledge of POLE mutational status. The second reason relates to the definition of “substantial” LVSI; in the retrospective cohorts studied in this project, LVSI is recorded as present versus absent, while current definitions of extensive or substantial LVSI would exclude cases showing only focal LVSI, as the latter finding does not affect risk categorization 14.

We also compared risk categorization based on clinicopathologic parameters without and with molecular classification and found that knowledge of ProMisE results altered the risk category in 10% of all EC’s, with 4% moving to a higher risk group, as a result of universal p53 testing, and 6% moving to a lower risk group, as a result of POLE testing; these results would be unchanged using universal ProMisE or ProMisE-S approaches.

The ProMisE-S algorithm thus provides a pragmatic way forward to introduce molecular classification into routine diagnostic practice. It saves on unnecessary, POLE testing within the very low-risk group. It improves on the alternative of restricting molecular classification only on high-grade and high-risk EC by ensuring, through universal p53 IHC, that no EC’s belonging to the worst clinical category of p53abn EC are missed. It ensures that EC’s showing more than one of the key molecular changes, that is “multiple classifiers,” are accurately categorized. Overall the data directly supports ESGO/ESTRO/ESP recommendation of omitting POLE mutation analysis in low/intermediate risk EC, by providing objective histopathologic criteria. At present there are several possible approaches to molecular testing in EC, summarized in Table 6. We present the results of retrospective evaluation of strategy 4 to demonstrate one economical approach.

TABLE 6 - Possible approaches to move forward with molecular testing, depending on resources available; these in turn will be subject to changes in test costs and further refinement of molecular groups
Approach Description Comment
1. Continue with standard histologic classification (status quo) Based on pathology and histomorphology alone
2. Universal MMR and p53 IHC (or low threshold for p53 IHC in all morphologically ambiguous cases), and accurate histologic diagnosis; no POLE testing MMR/MSI testing is universally recommended by numerous national and international clinical management guidelines, and pathologists should have a low threshold for carrying out p53 IHC. This will increase diagnostic sensitivity for high-risk EC POLEmut EC accounts for 12% of high-risk EC and 3%–5% of EC show multiple classifiers; these cases are potentially over-treated in the absence of POLE testing, estimated to be 6% of all cases based on our results
3. Universal MMR and p53 IHC and complete molecular classification (including POLE testing) of intermediate-high-risk cases in whom treatment modulation on the basis of molecular category is considered (Fig. 2) Cases classified as intermediate or high risk are currently offered additional radiation and/or chemotherapy. In some cases the decision to offer or withhold additional treatment will not be affected by the molecular classifier, ie these will be offered or will not be offered treatment regardless of molecular class. These can be classified based on given information. This will further narrow down the need for NGS testing to only those in whom treatment modulation will be considered based on NGS results: this may be due to patient age, comorbidities, patient/clinician choice or other factors
4. Universal MMR and p53 IHC and complete molecular classification of all cases other than very low-risk EC (Fig. 2) A majority of EC cases fall into a low-risk category and do not require adjuvant treatment. As long as universal p53 IHC is carried out to identify any potential high-risk cases that could be misclassified by histology alone Based on our results this will diminish the POLE testing burden by ~40%
5. Universal ProMisE testing (Fig. 1) If resources permit testing could be carried out on all cases though a combination of POLE sequencing with MMR and p53 IHC
6. Universal NGS panel testing In some health care settings and going forward it may be preferable and more cost-effective to carry out universal panel testing on all solid tumors, such that POLE mutational analysis, microsatellite instability testing and TP53 mutational analysis are included and molecular subtype can be assigned based on the panel results This would need to be validated and compared with conventional ProMisE algorithm that had been developed according to Institute of Medicine guidelines through discovery, confirmation and validation phases that used POLE NGS and IHC for MMR and p53 proteins
EC indicates endometrial carcinoma; IHC, protein immunohistochemistry; MMR, mismatch repair; MSI, microsatellite instability; NGS, next-generation sequencing.


In conclusion, adoption of molecular testing has tremendous potential to provide reproducible diagnosis, inform prognosis, and impact treatment decisions. In the absence of unlimited resources for molecular testing in all EC, we present clinically relevant alternative strategies for selective testing that appear to be safe and informative. We also demonstrate the extent to which testing can be restricted while allowing accurate categorization into ESGO/ESTRO/ESP risk groups.


1. Bokhman JV. Two pathogenetic types of endometrial carcinoma. Gynecol Oncol 1983;15:10–7.
2. Soslow RA. Endometrial carcinomas with ambiguous features. Semin Diagn Pathol 2010;27:261–273.
3. Soslow RA. High-grade endometrial carcinomas—strategies for typing. Histopathology 2013;62:89–110.
4. Gilks CB, Oliva E, Soslow RA. Poor interobserver reproducibility in the diagnosis of high-grade endometrial carcinoma. Am J Surg Pathol 2013;37:874–881.
5. Cancer Genome Atlas Research Network, Kandoth C, Schultz N, Cherniack AD, et al. Integrated genomic characterization of endometrial carcinoma. Nature 2013;497:67–73; Erratum in: Nature. 2013 8;500(7461):242.
6. Talhouk A, McConechy MK, Leung S, et al. A clinically applicable molecular-based classification for endometrial cancers. Br J Cancer 2015;113:299–310.
7. Talhouk A, McConechy MK, Leung S, et al. Confirmation of ProMisE: A simple, genomics-based clinical classifier for endometrial cancer. Cancer 2017;123:802–13.
8. Kommoss S, Bunz A-K, Taran F-A, et al. Validation of ProMisE molecular classifier in a large population-based series: a new era in endometrial carcinoma diagnosis and treatment. Oncol Res Treat 2018;41(suppl 1):1180–8.
9. León-Castillo A, Gilvazquez E, Nout R, et al. Clinicopathological and molecular characterisation of “multiple-classifier” endometrial carcinomas. J Pathol 2020;250:312–22.
10. León-Castillo A, Britton H, McConechy MK, et al. Interpretation of somatic POLE mutations in endometrial carcinoma. J Pathol 2020;250:323–35.
11. McAlpine JN, Chiu DS, Nout RA, et al. Evaluation of treatment effects in patients with endometrial cancer and POLE mutations: an individual patient data meta-analysis. Cancer 2021;127:2409–22.
12. WHO Classification of Tumours Editorial Board. WHO Classitifcation of Tumours: Female Genital Tumours (Vol 5). Lyon: International Agency for research on Cancer; 2020.
13. Koh WJ, Abu-Rustum NR, Bean S, et al. Uterine neoplasms, version 1.2018: clinical practice guidelines in oncology. J Natl Compr Canc Netw 2018;16:170–199.
14. Concin N, Matias-Guiu X, Vergote I, et al. ESGO/ESTRO/ESP guidelines for the management of patients with endometrial carcinoma. Radiother Oncol 2021;154:327–53.
15. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2021. Available at:
16. Raffone A, Travaglino A, Mascolo M, et al. TCGA molecular groups of endometrial cancer: pooled data about prognosis. Gynecol Oncol 2019;155:374–83.
17. Bosse T, Nout RA, McAlpine JN, et al. Molecular classification of grade 3 endometrioid endometrial cancers identifies distinct prognostic subgroups. Am J Surg Pathol 2018;42:561–8.
18. Leon-Castillo A, de Boer SM, Powell ME, et al. Molecular classification of the PORTEC-3 trial for high-risk endometrial cancer: impact on prognosis and benefit from adjuvant therapy. J Clin Oncol 2020;38:3388–97.
19. McAlpine J, Leon-Castillo A, Bosse T. The rise of a novel classification system for endometrial carcinoma; integration of molecular subclasses. J Pathol 2018;244:538–49.
20. Piulats JM, Guerra E, Gil-Martín M, et al. Molecular approaches for classifying endometrial carcinoma. Gynecol Oncol 2017;145:200–207.
21. Plotkin A, Kuzeljevic B, de Villa V, et al. Interlaboratory concordance of ProMisE molecular classification of endometrial carcinoma based on endometrial biopsy specimens. Int J Gynecol Pathol 2020;39:537–45.
22. Talhouk A, Hoang LN, McConechy MK, et al. Molecular classification of endometrial carcinoma on diagnostic specimens is highly concordant with final hysterectomy: Earlier prognostic information to guide treatment. Gynecol Oncol 2016;143:46–53.
23. Jamieson A, Huvila J, Thompson EF, et al. Variation in practice in endometrial cancer and potential for improved care and equity through molecular classification. Gynecol Oncol. 2022;165:201–214.
24. Jamieson A, Thompson EF, Huvila J, et al. P53abn Endometrial Cancer: Understanding the most aggressive endometrial cancers in the era of molecular classification. Int J Gynecol Cancer 2021;31:907–13.
25. McCluggage WG, Singh N, Gilks CB. Key changes to the World Health Organisation (WHO) classification of female genital tumours introduced in the 5th edition (2020). Histopathology 2022;80:762–78.
26. Vermij L, Smit V, Nout R, et al. Incorporation of molecular characteristics into endometrial cancer management. Histopathology 2020;76:52–63.
27. Kang EY, Wiebe NJ, Aubrey C, et al. Selection of endometrial carcinomas for p53 immunohistochemistry based on nuclear features. J Pathol Clin Res 2022;8:19–32.
28. Kurnit KC, Kim GN, Fellman BM, et al. CTNNB1 (beta-catenin) mutation identifies low grade, early stage endometrial cancer patients at increased risk of recurrence. Mod Pathol 2017;30:1032–41.
29. Yano M, Ito K, Yabuno A, et al. Impact of TP53 immunohistochemistry on the histological grading system for endometrial endometrioid carcinoma. Mod Pathol 2019;32:1023–31.

Endometrial carcinoma; TCGA; Molecular classification

Copyright © 2022 by the International Society of Gynecological Pathologists