Twenty-one patients with high International Federation of Gynecology and Obstetrics (FIGO) stage (stage III or IV) endometrial carcinoma who had undergone total hysterectomy in Korea University Guro Hospital (KUGH) between 2008 and 2017 were included in this study. A number of general pathologists performed the original diagnosis. Three pathologists (Y.K.C., S.Y.L., and H.J.J.) reviewed the cases; Y.K.C. is a gynecology specialty pathologist, S.Y.L. is a board-certified general pathologist, and H.J.J. is a fourth-year anatomic pathology resident.

The diagnostic process proceeded as follows. First, three reviewers reached a primary diagnosis based on the H&E-stained slides. Each case was diagnosed based on the 2014 WHO Classification of Tumors of Female Reproductive Organs [3], and was categorized into one of the following entities: endometrioid carcinoma, serous carcinoma, mucinous carcinoma, clear cell carcinoma, undifferentiated/dedifferentiated carcinoma, carcinosarcoma, and mixed carcinoma. Cases diagnosed as endometrioid carcinoma were classified as low-grade (FIGO grade 1 and 2) or high-grade (FIGO grade 3).

In cases with discrepant original and primary diagnoses, reviewers reached a secondary diagnosis based on the H&E-stained and IHC-stained slides. Finally, all three pathologists conducted a discussion and reached a consensus diagnosis. If the discussion did not lead to a consensus diagnosis, NGS analysis was performed.

Concordance between the original and consensus diagnoses was calculated using the kappa statistics. A kappa value of 0.4 indicates poor agreement, 0.4–0.6 indicates moderate agreement, 0.6–0.8 indicates substantial/good agreement, and 0.8–1.0 indicates near perfect/ excellent agreement. The Windows version of IBM SPSS ver. 24.0 (IBM Corp., Armonk, NY, USA) was used for statistical analysis.

IHC staining was performed on whole slide sections using two automated staining systems, including a Bond-III autostainer (Leica Biosystems, Wetzlar, Germany) and a Bench-Mark ULTRA system (Ventana Medical Systems, Tucson, AZ, USA). In brief, 4-μm-thick paraffin-embedded tissue sections were deparaffinized and rehydrated across a graded series of ethyl alcohol concentrations. Heat-induced antigen retrieval was carried out in citrate buffer. Sections were incubated with the primary antibody in an automated immunostainer. Counterstaining with hematoxylin was performed. The antibodies used were as follows: anti-p53 (1:200, DO-7, Novocastra, Newcastle, UK), anti-ER (Ready-To-Use [RTU], SP1, Roche, Basel, Switzerland), anti-PR (RTU, 1E2, Roche), anti-p16 (RTU, E6H4, Roche), anti-MLH1 (1:100, ES05, Novocastra), anti-MSH2 (1:400, G219-1129, Novocastra), anti-PMS2 (1:100, MRQ-28, Cell Marque, Rocklin, CA, USA), anti-MSH6 (1:200, 44, Cell Marque), anti–hepatocyte nuclear factor-1β (HNF-1β; 1:200, polyclonal, Sigma-Aldrich, St. Louis, MO, USA), anti–napsin A (1:100, polyclonal, Cell Marque), anti-CD56 (1:200, 123 C3, DAKO, Glostrup, Denmark), anti-synaptophysin (1:200, polyclonal, Cell Marque,), anti-chromogranin (1:500, Dak-A3, DAKO), anti-GATA3 (1:50, L50-823, Cell Marque), anti-CD10 (RTU, CC1, Roche), anti–thyroid transcription factor-1 (TTF-1; 1:200, 8G7G3/1, DAKO), anti-WT1 (1:200, 6F-H2, Cell Marque), anti-cytokeratin (CK; 1:200, AE1/AE3, DAKO), anti-vimentin (1:1000, Vim 384, DAKO), anti-Ki67 (1:100, MIB-1, DAKO), anti-PTEN (1:400, 6H2-1, DAKO), and anti–c-KIT (1:200, polyclonal, Novocastra).

Each reviewer independently assessed the results of IHC staining. When interpretation-associated discrepancies arose, all three reviewers conducted a discussion until a consensus was achieved. p53 immunostaining was interpreted as mutation type if the tumor exhibited: (1) diffuse strong nuclear positivity involving at least 80% of the tumor cells (aberrant pattern), (2) complete absence of staining with the presence of positive internal control staining of non-neoplastic cells such as lymphocytes (null pattern), or (3) unequivocal cytoplasmic staining accompanied by variable nuclear staining (cytoplasmic staining pattern). Cases were considered wild-type if any degree of non-diffuse nuclear staining (< 80%) of the tumor cells was present [12]. Immunostaining for ER and PR was scored based on the percentage of tumor cells exhibiting moderate to intense nuclear staining. p16 staining was interpreted as positive if ≥ 90% of tumor cells were stained. PTEN immunostaining was considered negative if there was a complete loss of expression in tumor cells. Immunostaining for MLH1, MSH2, MSH6, and PMS2 was considered negative (loss of expression pattern) if there was a complete absence of nuclear staining. Immunostaining for napsinA and WT1 was interpreted as positive if ≥ 1% of the tumor cells were stained, regardless of the staining intensity. HNF-1β immunostaining was interpreted as positive if there was positive staining in ≥ 50% of tumor cells. Immunostaining for neuroendocrine markers (CD56, synaptophysin, and chromogranin) was interpreted as positive if there was positive staining in ≥ 10% of tumor cells. Immunostaining for CK and vimentin was considered positive if there was diffuse staining in ≥ 50% of tumor cells. Immunostaining for CD10, GATA3, and TTF-1 was interpreted as positive if there was more than one area of focal staining. c-KIT immunostaining was considered positive if there was cytoplasmic staining in ≥ 1% of tumor cells. The Ki67 labeling index was measured by recording the tumor cell percentages that exhibited moderate to intense nuclear staining.

Up to five 10-μm-thick sections of formalin-fixed, paraffin-embedded tissue blocks were used for DNA and RNA sequencing analysis. Representative tumor areas were manually micro-dissected. DNA and RNA were extracted using the Invitrogen Recover All Total Nucleic Acid Isolation Kit (Invitrogen, Carlsbad, CA, USA). Massively parallel sequencing of cancer-related gene panels (Oncomine Comprehensive Assay v3, ThermoFisher Scientific Waltham, MA, USA) was performed using the Ion Chef and Ion Torrent sequencing platforms. Sequence data were aligned against the human reference genome (hg19 build). Bioinformatics data analysis was performed using Ion Reporter software (v.5.10.1.0) with the Oncomine Variants filter chain (v.5.6).

The original diagnoses in 21 cases were endometrioid carcinoma (n = 15), serous carcinoma (n = 5), and clear cell carcinoma (n = 1). Among the 15 cases of endometrioid carcinoma, five were low grade (G1, 2 cases; G2, three cases), and 10 cases were high grade (G3). The primary diagnoses made by the three reviewers were identical in 13/21 cases (62%). These cases included endometrioid carcinoma (n = 9), serous carcinoma (n = 3), and clear cell carcinoma (n = 1). Table 1 showed a summary of the primary diagnosis.

Secondary diagnosis due to disagreement in primary diagnosis or diagnostic difficulty was performed in 8/21 cases (38%). Table 2 summarized the IHC staining result in eight discordant cases. The original diagnosis of eight cases with discrepancies was endometrioid carcinoma in six cases (one case of G1 and five cases of G3) and serous carcinoma in two cases. Secondary diagnosis based on H&E and IHC staining was concordant in four of eight discrepant cases. The secondary diagnosis of four concordant cases was mesonephric-like adenocarcinoma (case 1), carcinosarcoma (case 2), clear cell carcinoma (case 3), and large cell neuroendocrine carcinoma (LCNEC) (case 4). The four discordant cases were as follows: mixed carcinoma vs. LCNEC (case 5), G3 endometrioid vs. dedifferentiated carcinoma (case 6), and G3 endometrioid vs. serous carcinoma (cases 7 and 8).

Among four discordant cases, two cases reached a consensus diagnosis through review of all slides and discussion between three pathologists. The consensus diagnosis was LCNEC (case 5) and dedifferentiated carcinoma (case 6). In two cases of G3 endometrioid vs. serous carcinoma (cases 7 and 8), NGS was performed because a consensus was not achieved even after discussion. Table 3 was a sequential diagnosis summary of eight discordant cases.

Through the above diagnostic process, three reviewers made a consensus diagnoses for 21 advanced-stage endometrial carcinoma cases. The final diagnoses included four cases of low-grade endometrioid carcinoma, five cases of high-grade (G3) endometrioid carcinoma, five cases of serous carcinoma, two cases of clear cell carcinoma, two cases of neuroendocrine carcinoma, one case of mesonephric-like adenocarcinoma, one case of carcinosarcoma, and one case of dedifferentiated carcinoma (Table 4). The overall kappa for concordance between the original diagnosis and consensus diagnosis was 0.566 (moderate agreement).

The original diagnosis of case 1 was G1 endometrioid carcinoma. Three pathologists’ primary diagnosis were discordant, as one pathologist diagnosed this case as mesonephric-like adenocarcinoma and two others diagnosed it as low-grade endometrioid carcinoma. The tumor exhibited diverse histological morphology such as papillary and ductal/glandular patterns with a low nuclear grade. Nuclear features of the cells included nuclear overlapping, nuclear grooves, and open, vesicular chromatin, similar to those of papillary thyroid carcinoma. Although the histopathological characteristics were consistent with those of mesonephric adenocarcinoma, mesonephric remnants and uterine cervical involvement were not identified. IHC staining showed positivity in GATA3 and TTF-1, negativity in CD10 and ER, and focal positivity in PR. Based on pathologic findings and IHC results, the consensus diagnosis was mesonephric-like adenocarcinoma (Fig. 1A, B).

In case 2, the original diagnosis was serous carcinoma. This primary diagnosis was discordant as two pathologists diagnosed this as serous carcinoma, while the remaining pathologist could not decide between serous carcinoma and carcinosarcoma. Microscopic findings showed serous and mesonephric carcinoma components. Two small foci of atypical spindle and pleomorphic cell proliferation suggested the presence of a sarcomatous component. The spindle cell component exhibited strong vimentin positivity and aberrant p53 expression, while CK expression was focal and weak. After discussion, the three pathologists agreed upon the presence of a distinct sarcomatous component, and a consensus diagnosis of carcinosarcoma was made (Fig. 1C, D).

Case 3 showed solid or glandular proliferation patterns, and the tumor cells displayed severe nuclear atypia and abundant clear or eosinophilic cytoplasm. The primary diagnosis was discordant, with one diagnosis of clear cell carcinoma and two diagnoses of endometrioid carcinoma. On IHC stains, the tumor cells were p53 wild-type and negative for ER, PR, and HNF-1β with granular positivity for napsin A. The final consensus diagnosis was clear cell carcinoma (Fig. 1E, F).

In case 4, large polygonal cells with prominent nucleoli and abundant cytoplasm formed well-demarcated nests, which displayed peripheral palisading. Frequent mitotic figures were observed, and geographic tumor necrosis was identified. The primary diagnosis was discordant, with two diagnoses of neuroendocrine carcinoma and one diagnosis of G3 endometrioid carcinoma. The tumor cells revealed diffuse positivity to CD56 and chromogranin. A consensus diagnosis of LCNEC was rendered (Fig. 1G, H).

For case 5, the primary diagnoses were mixed carcinoma (endometrioid carcinoma, clear cell carcinoma, and neuroendocrine carcinoma) vs. G3 endometrioid carcinoma vs. serous carcinoma. Histopathologic analysis showed that the tumor was composed of endometrioid carcinoma (50%), clear cell carcinoma (30%), and LCNEC (20%). The IHC staining revealed an LCNEC component showing diffuse positivity for CD56 and synaptophysin and focal positivity for chromogranin. Although the tumor consisted of various components, according to the criteria of the WHO Classification of Tumors of Female Reproductive Organs [3], case 5 was diagnosed as LCNEC (Fig. 2A, B).

In case 6, the primary diagnosis was divided into two diagnoses of dedifferentiated carcinoma and one diagnosis of serous carcinoma. The secondary diagnosis was also discordant, as two pathologists diagnosed the case as dedifferentiated carcinoma and the remaining pathologist’s diagnosis was G3 endometrioid carcinoma. Microscopically, the tumor was composed of G1 endometrioid carcinoma and a poorly differentiated solid portion showing non-cohesive cells without readily visible gland formation. The three reviewers agreed that the solid portion was an undifferentiated carcinoma, not a high-grade endometrioid carcinoma component. A final consensus diagnosis of dedifferentiated carcinoma was reached (Fig. 2C, D).

The microscopic features of case 7 displayed columnar tumor cells with nuclear pseudostratification and pleomorphism. While the tumor’s dominant pattern was solid proliferation, several foci showed complex papillary or glandular patterns. IHC staining revealed a complete loss of p53 staining, negativity for ER and PR, and positivity for p16, WT1, MLH1, PMS2, MSH2, and MSH6. NGS analysis was performed as histopathological features (suggestive of endometrioid carcinoma) and IHC features (suggestive of serous carcinoma) exhibited discrepancies (Fig. 2E, F). NGS revealed mutations in TP53, PIK3CA, and NF1, and copy number variations (CNVs) in MYCL, FGFR3, CDK2, CDK4, ERBB2, and CCNE1. As the molecular profile corresponded to a copy-number high group, the final diagnosis was rendered as serous carcinoma.

Case 8 presented with high-grade tumor cells with severe nuclear pleomorphism. The tumor showed complex glandular and solid patterns. In the solid portion, nuclear pleomorphism was less severe than that of typical serous carcinoma. The primary diagnoses were divided into G3 endometrioid carcinoma versus serous carcinoma. On immunostaining, the tumor cells exhibited aberrant p53 expression, negativity for PR, diffuse positivity for p16, MLH1, PMS2, MSH2, and MSH6, and focal positivity for WT1 and ER. While the immunoprofile suggested serous carcinoma, the diagnosis of G3 endometrioid carcinoma could not be entirely excluded for the histomorphological findings (Fig. 2G, H). NGS analysis revealed mutations in TP53 and polymerase E (POLE) and CNVs involving ALK, ERBB2, CCNE1, AKT2, and AXL. Although POLE mutation was present, the total tumor mutation burden was not as high as that seen in the typical POLE-ultramutated type. Since a significant number of CNVs were detected, the case was assigned to the copy-number high group, and a final diagnosis of serous carcinoma was made.