The diagnosis of endometrial precancers poses everyday problems in pathology practice.1,2 Endometrial biopsies may be limited or highly fragmented, and cyclical variations or exogenous hormones can further confound histologic features of endometrial precancers (atypical hyperplasia/endometrioid intraepithelial neoplasia, or AH/EIN). These challenges are widely acknowledged, as studies have consistently documented suboptimal interobserver agreement.3–6 One difficulty is that while some AH/EIN appear as tight clusters of crowded glands that are well-demarcated and distinctive cytologically or architecturally from adjacent “normal” glands, other AH/EIN can be more diffuse, with a continuum of architectural features that can make it difficult to clearly delineate a lesion and confidently classify it. In such cases, distinction from adjacent normal glands (to assess atypia) is not always possible. Gland crowding leading to a reduction in stromal volume (gland area>stromal area, or >50% gland area) is a useful histologic criterion,7–10 but its application can also be challenging in practice. For example, estimating gland area is problematic in highly fragmented specimens, and area quantitation by digital image analysis can be arbitrary because of the need to select specific areas, among other reasons.11,12 Also, benign endometria such as secretory endometrium, or with extensive areas of cystically dilated glands associated with atrophy, can exceed 50% gland area. Yet other challenges arise in the evaluation of endometrial polyps (EMPs), which sometimes harbor AH/EIN but can also exhibit increased gland crowding and architectural disarray.13–15

Building on earlier studies,16–20 we recently defined a panel of 3 immunohistochemical (IHC) markers—comprised of β-catenin, Pax2, and Pten—with practical utility in the diagnosis of AH/EIN.21 Reflecting their unique biology, each marker is scored per distinct criteria.22–24Arid1a and Mlh1 are sometimes lost in AH/EIN,25,26 and are reliably scored, but are rarely if ever the sole aberrant markers, arguing against their inclusion in the core panel, although they may be useful in select cases.1,21β-catenin, Pax2, and Pten are each aberrant in more than half of AH/EIN; with a panel of all 3, at least 1 is aberrant in 93% of cases.21

A prior report describing this 3-marker panel evaluated only completely normal endometria and definitive AH/EIN, which was helpful in establishing aberrancy criteria and defining aberrancy rates for each marker individually and in the optimal panel of 3 markers.27 However, this left open questions about the performance of the panel in ambiguous or subdiagnostic lesions. To advance our understanding of the 3-marker panel, we sought to evaluate its application in a spectrum of lesions including disordered proliferative endometrium (DPE) and nonatypical hyperplasia (NAH). These results were correlated to clinical disease progression and next-generation sequencing (NGS) analyses.

MATERIALS AND METHODS Case Selection

After approval from the UT Southwestern Institutional Review Board, we retrospectively identified cases via text searches with a final diagnosis of NAH or DPE accessioned between 2010 and 2021 at the UT Southwestern Clements University Hospital. Exclusion criteria were a prior diagnosis of AH/EIN or uterine malignancy. The following terms were used in text searches: endometrial/endometrium, “disordered proliferative endometrium,” “hyperplasia without atypia,” “non-atypical,” “gland crowding,” “crowded glands,” and “insufficient.” A diagnosis of cancer in an unrelated organ system was not an exclusion criterion. Following searches to preliminarily identify cases, the original hematoxylin and eosin (H&E) slides were reviewed by a panel of 3 specialty pathologists to reach a consensus diagnosis, blinded to the original diagnosis, and before the performance of the IHC panel. Standard histologic diagnostic criteria were used for the initial diagnosis of AH/EIN including gland:stroma ratio >1, overt nuclear atypia/cytologic demarcation from background endometrium, and exclusion of mimics.9,28 DPE was defined by the presence of well-spaced, cystically dilated/enlarged glands without cribriforming or epithelial stratification, resembling the pattern that is typical in the perimenopausal state. NAH was defined as the presence of more significant and/or diffuse gland crowding but falling short of the criteria for AH/EIN, with no cribriforming, epithelial stratification, or definitive cytologic distinctiveness/atypia.21 Most of the cases could be placed into a consensus category (proliferative endometrium [PE], DPE, NAH, or AH/EIN) but those that could not and raised suspicion for AH/EIN were labeled as ambiguous/difficult. The consensus diagnoses of these ambiguous cases contained phrases like “suspicious for,” “gland crowding,” and “complex.” Cases harboring definitive AH/EIN or carcinoma in the index biopsy or reclassified as normal in the expert pathology review were excluded (2 from the initial pool of 92 cases, leaving 90 cases). Patient charts for all 90 cases were then systematically reviewed for any follow-up endometrial tissue samples (biopsies and hysterectomies within our institution’s health care system) after the index case. For patients who developed subsequent AH/EIN or carcinoma (ie, progressors), the accession date for the first tissue specimen harboring AH/EIN or carcinoma was used for the progression-free survival (PFS) analysis. The most representative tissue block was selected for each index and progressor specimen for tissue-based analyses. The n=79 normal controls and n=111 AH/EIN were previously reported.21

Immunohistochemistry

For Pax2, Pten, β-catenin, and Mlh1 staining protocols previously validated for clinical testing were performed on 4 μm sections in the clinical IHC laboratory on a DAKO Autostainer Link 48 instrument. The following primary antibodies were used: Mlh1 (prediluted, clone ES05, #IR07961-2; Agilent), β-catenin (prediluted, clone β-catenin-1, #IR70261-2; Agilent), Pax2 (prediluted, clone EP235, #BSB2567; Cancer Diagnostics), and Pten (prediluted, clone 6H2.1, #PM278AA; BioCare) with antigen retrieval performed in low pH (6.0) for β-catenin and high pH (9.0) Tris/EDTA solution (Agilent) for the other markers at 97°C for 20 minutes. FLEX peroxidase block was performed for 10 minutes for β-catenin and 5 minutes for other markers. Primary antibody incubation time was 20 minutes for β-catenin, and 40 minutes for Pax2, Pten, and Mlh1. Incubation with Mouse Linker (Agilent) for β-catenin and Rabbit Linker (Agilent) for Pax2 was performed for 10 minutes. Secondary antibody (Envision/HRP) incubation time was 20 minutes for Pten, β-catenin, 30 minutes for Pax2, and 40 minutes for Mlh1. Arid1a IHC was performed on a DAKO Autostainer Link 48 instrument in a research core facility (1:200 dilution, clone D2A8U, #12354; Cell Signaling Technology) with low pH (6.0) Tris/EDTA solution (Agilent) for 20 minutes at 97°C. Primary and secondary incubation times were 20 minutes each. For all antibodies, the enzymatic conversion of the 3,3′-diaminobenzidine tetrahydrochloride chromogen was performed for 10 minutes at room temperature.

Scoring of the 3 Markers by IHC β-catenin

Aberrancy is manifested as nuclear localization versus its normal membranous/cytoplasmic localization, often associated with overexpression. Strong nuclear β-catenin is scored as aberrant, even if focal.22,23 Nuclear staining is assessed only in glandular epithelium (not squamous morules) since true morules always exhibit nuclear β-catenin. Low levels of nuclear β-catenin are normal; the criterion for strong nuclear expression is nuclear staining clearly greater than that of the lateral cell membranes.21

Pax2

Pax2 is a nuclear transcription factor expressed within the uterus only in endometrial epithelium, and scoring aberrancy requires complete loss of expression within all of the nuclei in an entire gland in cross-section. Decreased expression is not scored as aberrant. In most specimens, residual normal glands that retain Pax2 serve as internal controls for Pax2 IHC.

Pten

Pten is ubiquitously expressed in endometrial glands, stroma, and leukocytes, and true loss within the endometrial glandular epithelium gives rise to a “punched-out” appearance of glands relative to the surrounding stroma. Leukocytes interspersed within endometrial glands, which can be abundant, retain Pten expression even when true epithelial loss is present.

Focal loss of Pax2 and Pten in individual glands or small clusters of glands can occur in normal endometria, with >10% loss across the glands in the entire specimen used as the cutoff for aberrancy. However, most cases of AH/EIN exhibit Pax2 or Pten loss in >25% of glands, making scoring more straightforward.

UTSW Clinical NGS Panel

Unstained 4 μm sections were cut from one tissue block for each case. Areas of nonprecancer/cancer (ie, subjacent endomyometrium) were macrodissected away with a blade. DNA was also prepared from control somatic tissues removed during the hysterectomy, such as the ovary or fallopian tube (permitting distinction of acquired somatic mutations vs. inherited germline mutations). DNA and RNA were isolated using Qiagen Allprep kits. The custom panel of DNA probes was used to produce an enriched library, using a Kapa NextSeq. 550 instrument. DNA and RNA sequence analyses were done using custom germline, somatic, and mRNA bioinformatics pipelines run on the UTSW Bio-High Performance Computer cluster and optimized for the detection of single nucleotide variants, indels, and gene fusions.29,30 The median target exon coverage for the assay is ×900 with 94% of exons at ×100. FASTQ files were aligned into BAM files with BWA, then called with a combination of variant callers for somatic variants (platypus, GATK, SAMtools), copy number variants (OncoCNV), and fusions (PINDEL, StarFusion) to create a VCF file, which was annotated through ANSWER software.30 The RNA panel covers 1505 genes and can capture unbaited partners. The limit of detection is 5 RNA reads. Gene Set Cancer Analysis to compare the spectrum of mutations to TCGA data sets was performed on the GSCA portal (http://bioinfo.life.hust.edu.cn/GSCA/#/mutation)31 and cBioPortal and plotted with the Cancer Type Summary Tools.32

Quantitative Analysis of Gland:Stroma Ratios

H&E-stained tissue sections were scanned on an Aperio ScanScope CS. To quantify gland density (ie, gland:stroma ratio) representative areas of the scanned images were selected in Aperio ImageScope (v12.3.3). Representative examples of the areas selected as total (gland+stroma) and gland only are shown in Figure S1 (Supplemental Digital Content 1, https://links.lww.com/PAS/B523) and quantified by ImageScope. Areas judged to be surface and not glandular endometrium were excluded from total area. The % gland density was calculated as gland area divided by total area.

RESULTS Biomarker Aberrancy Across Endometrial Diagnostic Categories

We sought to analyze a range of samples with architectural abnormalities falling short of AH/EIN. A total of n=92 cases were identified by text searches. The only exclusion criteria were a prior diagnosis of any uterine/ovarian neoplasia including AH/EIN. All original H&E slides for each case were reviewed by 3 gynecologic pathologists to reach a consensus diagnosis and final placement into 3 diagnostic categories: DPE, NAH, and ambiguous/difficult. Cases that after review were diagnosed as PE (ie, no meaningful architectural abnormalities) or AH/EIN (definitive premalignancy) and thus did not fall into the diagnostic spectrum between normalcy and AH/EIN were redacted. This included 1 case rediagnosed as PE, and 1 as AH/EIN, leaving a set of n=90 new cases for further investigation (Table 1). An additional n=180 previously reported PE and definitive AH/EIN subjected to the 3-marker panel21 were included for comparative purposes.

TABLE 1 - Clinical Summary Total patients=90 (age range: 49-70 y [average: 57.3 y]) DPE=44 (age range: 50-75 y [average: 54.5 y]) NAH=40 (age range: 49-77 y [average: 59.5 y]) Ambiguous/difficult=6 (age range: 54-71 y [average: 63.7 y])

One block from these 90 index cases representing the most prominent architectural abnormalities was selected for immunostaining with 3-marker panel. The 270 IHC slides were scored per previously described criteria.21 The fraction of cases aberrant for ≥1 marker was 18.2% for DPE, 47.5% for NAH, and 50.0% for ambiguous/difficult cases (Fig. 1A). This difference in marker aberrancy between DPE and NAH was statistically significant (P=0.0052, Fisher exact test). Aberrancy for each of the markers across the diagnostic categories is presented in Figure 1B. In NAH, the 3 markers were aberrant in roughly equal proportions, whereas in DPE, Pax2 aberrancy was much more common. The distribution of cases aberrant for 1, 2, or 3 biomarkers is also shown for each diagnostic category (Fig. 1C). In summary, in contrast to PE, the 3-marker panel discriminated among endometria across the diagnostic categories, with some cases exhibiting marker aberrancy for 1 or more marker, a result that may be unexpected given current histopathologic diagnostic schema that regard DPE and NAH as unequivocally benign entities.

FIGURE 1: Marker aberrance across diagnostic categories. A, Percent of cases per diagnostic category aberrant for at least 1 of the 3 principal markers (β-catenin, Pax2, Pten). B, Distribution of aberrancy for the 3 principal markers per diagnostic category. Data for PE and AH/EIN was derived from a recent study from our group with IHC performed in the same Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory.21 C, Distribution of aberrance for 1, 2, or 3 biomarkers per diagnostic category.Association of 3-marker Aberrancy With Risk of Disease Progression

The medical records for the n=90 women were then all methodically searched for subsequent tissue specimens that would permit formal analysis of endometrial disease progression (PFS). Overall, 52/90 patients (58%) had a subsequent endometrial tissue sample in our hospital system. The tissue follow-up intervals ranged from 1 to 9 years. Women with subsequent tissue-based diagnoses of AH/EIN or cancer (in all cases FIGO 1 endometrioid adenocarcinoma) were categorized as progressors. The original H&Es for these subsequent specimens were reviewed to confirm progression. In 1 case, the original diagnosis of AH/EIN (in a hysterectomy) was revised to grade 1 adenocarcinoma based on small foci of malignant epithelium without intervening stroma (patient 5). Formal PFS analysis of patients with marker nonaberrant versus marker-aberrant index cases showed that the latter were more likely to progress. Only 1/60 marker nonaberrant cases progressed (1.6%), whereas 8/30 marker-aberrant cases progressed (27%) (P=0.0005, Fisher exact test). The difference in the PFS curves (Fig. 2) was also statistically significant (P=0.0002, log-rank test; hazard ratio=18.0, 95% CI: 4.4-74).

FIGURE 2: PFS of all cases in this study (marker nonaberrant vs. aberrant). A diagnosis of AH/EIN or cancer () on a subsequent tissue sample was scored as progression. The difference in the curves is statistically significant (P=0.0002 per log-rank test).Patterns of Biomarker Aberrancy in Progressors

Next, histologic features and associated patterns of biomarker aberrancy were analyzed and are described below for several illustrative cases.

Patient 9 (DPE) (Fig. 3): The histologic features of the index case for this patient, the sole nonaberrant progressor, included cystically dilated glands without significant crowding or cribriforming; image analysis documented 24% gland area (see Fig. S1, Supplemental Digital Content 1, https://links.lww.com/PAS/B523 for examples of area selection for % gland quantitation including this case). AH/EIN was diagnosed 4 years and 5 months after the initial diagnosis of DPE (progressor specimen). Since the DPE and subsequent AH/EIN were nonaberrant for the 3 markers, the additional markers Arid1a and Mlh1 were performed. These were also nonaberrant in both the index and progressor specimens. Lack of marker aberrancy in the AH/EIN and the preceding DPE in this patient is concordant with prior observations that 7% of AH/EIN are marker nonaberrant,1,21 resulting in some “false-negative” cases with the 3-marker panel. Persistent patterns of biomarker loss among the marker-aberrant cases presented below suggest that DPE or NAH preceding AH/EIN or cancer are bona fide precursor lesions in most patients.

FIGURE 3:

Nonaberrant progressor, patient 9 (sole example in study). The index specimen showed DPE with nonaberrancy for the 3-marker panel or additional markers Mlh1 and Arid1a (performed for this case because of nonaberrancy for the 3 principal markers); % gland density per image analysis is in the upper left corner of the H&E images. The subsequent AH/EIN similarly did not show aberrancy for any of the markers.

Patient 1 (NAH) (Fig. 4): The histologic features of this case included diffuse (nonfocal) gland crowding (33% gland area in the representative area shown) without definitive cytologic distinctiveness. Nonetheless, the lesion showed widespread Pax2 and Pten loss (>50% each). One notable feature was the presence of areas of degeneration/necrosis, but such areas were not surrounded by more concerning features or increased gland crowding. A hysterectomy performed 9 months later for persistent abnormal uterine bleeding and thickened endometrial stripe showed FIGO 1 endometrioid adenocarcinoma that was similarly Pax2 and Pten aberrant. With the relatively short time interval, the findings in this patient could reflect sampling error in the index biopsy. This is consistent with the presence of NAH-like areas at the periphery of the adenocarcinoma that also showed Pax2 and Pten loss (lower panels). Identical patterns of marker aberrancy in the index NAH and the adenocarcinoma and its adjacent NAH regions in the hysterectomy specimen strongly support a common biological origin in a Pax2/Pten-deficient precursor detected in the index specimen. This and the other cases below suggest that the 3-marker panel can be diagnostically informative even with some sampling error, revealing endometria more likely to harbor bona fide precursors at increased risk for neoplastic progression.

FIGURE 4:

Aberrant progressor, patient 1. The index specimen showed NAH; % gland density per image analysis is in the upper left corner of the H&E. *=area of necrosis/stromal degeneration without more concerning features; †=highlights scattered glands retaining Pten expression. The subsequent FIGO 1 endometrioid adenocarcinoma was similarly deficient for both Pax2 and Pten. An adjacent area was morphologically consistent with NAH and closely resembled the NAH in the index specimen; its pattern of Pax2 and Pten loss matched that of the index NAH. The Pax2 and Pten IHC images in the lower panels correspond to the boxed area on the H&E on the left.

Patient 2 (NAH) (Fig. 5): The index case also showed NAH with diffuse but mild gland crowding (21% gland density) and lack of cytologic distinctiveness. Pax2 was retained but Pten showed >50% loss. The subsequent FIGO 1 adenocarcinoma (diagnosed 9 years later) showed Pax2 and Pten loss. This case illustrates that additional markers can be lost during disease progression, as would be expected, although sampling issues cannot be entirely excluded. Also, marker aberrancy does not occur in a determinate order: in this case, Pten loss was the initial event, followed by Pax2. The extent of Pten loss in the NAH strongly suggests a neoplastic process, as diffuse Pten loss is not observed in normal endometria. Consistent with this order of marker loss (Pten→Pax2), large areas of the adenocarcinoma showed retention of Pax2, whereas Pten was lost throughout.

FIGURE 5:

Aberrant progressor, patient 2. The index case was an NAH with widespread Pten loss despite minimal gland crowding (21%). The ensuing FIGO 1 adenocarcinoma was Pax2 and Pten-deficient. While Pten was diffusely lost, Pax2 was lost only in some areas (*).

Patient 5 (NAH) (Fig. 6): The index case was diagnosed as NAH and harbored diffuse admixtures of mildly dilated glands and small areas of focally increased gland density. However, no areas approached 50% gland density considered the usual threshold for AH/EIN. The representative area shown has a gland area of 26% (Fig. S1, Supplemental Digital Content 1, https://links.lww.com/PAS/B523). Pax2 and Pten were aberrant in >50% of the total glands, and β-catenin was aberrant with nuclear localization, albeit only focally. The subsequent FIGO 1 adenocarcinoma, diagnosed 5 weeks later, was aberrant for Pax2, Pten, and β-catenin, and concordantly exhibited squamous metaplasia, commonly seen in CTNNB1-mutant endometrioid adenocarcinomas.22,23,33 Thus, this is another example of marker concordance in a precursor lesion that was diagnosed as NAH by histologic features, but in fact, is revealed by the 3-marker panel to be a neoplastic precursor mimicking a non-neoplastic process. It is also another example of the 3-marker panel’s ability to detect aberrancy even with suboptimal sampling (which seems likely in this case given the extremely short time interval between index and progressor).

FIGURE 6:

Aberrant progressor, patient 5. The index specimen showed NAH, which was extensively deficient for Pax2 and Pten (>50%, as evident from low and high magnification) and also aberrant for β-catenin, albeit focally; †=adjacent normal gland. Percent gland density per image analysis is in the upper left corner of the H&E image. The subsequent endometrioid adenocarcinoma exhibited squamous metaplasia and was also aberrant for β-catenin, Pax2, and Pten.

Patient 6 (DPE) (Fig. 7): This aberrant progressor had an index case with features of DPE, with cystically dilated glands and no areas of significant gland crowding or cytologic distinctiveness. However, the endometrial fragments were aberrant for both Pax2 and β-catenin. The ensuing AH/EIN showed morphologically distinct areas with significant gland crowding and abundant squamous metaplasia, and other areas with crowding but no overt squamous metaplasia. The AH/EIN was similarly aberrant for Pax2 and β-catenin, with β-catenin being distinctly nuclear and overexpressed even in areas without squamous metaplasia, underscoring the utility of β-catenin as an informative marker even in cases or areas without diffuse squamous metaplasia.

FIGURE 7:

Aberrant progressor, patient 6. The index case was a DPE. Despite morphologically banal features, the lesion showed extensive Pax2 loss and focal β-catenin aberrancy (overexpression and nuclear localization); †=adjacent glands with membranous β-catenin expression. The subsequent AH/EIN showed squamous metaplasia in some areas (sq, left H&E panel) but not in others (right H&E panel). The AH/EIN was diffusely Pax2 and β-catenin aberrant with nuclear localization and overall overexpression in areas±squamous metaplasia. β-catenin IHC images show higher magnifications of areas in the H&Es directly above.

Patient 3 (EMP with gland crowding) (Fig. S2, Supplemental Digital Content 2, https://links.lww.com/PAS/B524): For a complete description of this complex case, which was categorized as ambiguous/difficult and gave rise to a FIGO 1 adenocarcinoma aberrant for all 3 markers (Fig. S2, Supplemental Digital Content 2, https://links.lww.com/PAS/B524).

Mutation Detection by Cancer Gene Panel Rationalizes Biomarker Aberrancy in Progressors

Although Pax2 loss has not been associated with gene mutations, β-catenin, Pten, and Arid1a aberrancy are often associated with gene mutations in the CTNNB1, PTEN, and ARID1A genes.34 AH/EIN and FIGO 1 endometrioid adenocarcinoma progressor cases (patients 1 to 9) were analyzed with a comprehensive clinical NGS gene panel, which was designed for mutation detection in samples with high cellularity such as the progressor cases. The progressor specimens exhibited mutational spectra expected for endometrioid adenocarcinomas, with mutations in KRAS, FBXW7, FGFR2, PIK3CA, and PIK3R1, among other endometrial cancer genes (Fig. 8).24,27,35–37 Of cases with somatically acquired CTNNB1 single nucleotide variants (all representing well-known exon 3 hotspots), 2/3 were scored as aberrant for β-catenin (see patients 4, 5, 6). For patients 1 and 2, biallelic truncating PTEN mutations fully rationalized the observed loss of protein. Of the other 4 patients with at least 1 PTEN mutation, 3 were aberrant for IHC, with the exception of patient 6 (PTEN p.R130Q, see the Discussion section). Patient 3 harbored 2 truncating alleles of ARID1A. IHC for the 2 ancillary AH/EIN markers Arid1a and Mlh1 was performed on the specimens for this patient, which confirmed the expected Arid1a loss and also incidentally revealed Mlh1 loss (Fig. S2, Supplemental Digital Content 2, https://links.lww.com/PAS/B524). For patient 9, the sole nonaberrant progressor (Fig. 3), mutations were not detected in either PTEN or CTNNB1. Thus, IHC aberrance for β-catenin, Pten, and Arid1a generally correlated with mutations in the respective genes, although concordance was incomplete as in prior studies.23,38,39

FIGURE 8:

Clinical NGS analysis of 9 progressor cases including 1 nonaberrant progressor (patient 9). All identified mutations are listed for the progressor cases for the 9 patients, with mutations relevant to the 3-marker panel and ancillary markers (CTNNB1, PTEN, ARID1A) highlighted in red. The 3-marker results plus Arid1a and Mlh1 (performed only for patient 3 to validate the NGS data, and for patient 9 because of lack of aberrancy of the 3 principal markers) are shown below for the progressor and index cases for each patient.

DISCUSSION

In this study, we found that previous criteria for scoring the IHC markers β-catenin, Pax2, and Pten based on analyses of AH/EIN and benign endometria1,21 are also applicable to endometria in between these diagnostic categories, including morphologically ambiguous or subdiagnostic specimens. One surprising finding in this study is the relatively high incidence of biomarker aberrancy in DPE (18.2%) and NAH (47.5%) compared with 0% in PE (92.8% in AH/EIN),21 given that DPE and NAH are considered entirely benign per some classification systems. For example, in the WHO Classification of Female Genital Tumors (2020), “benign endometrial hyperplasia” is considered acceptable terminology for “endometrial hyperplasia without atypia” (ie, NAH).40 On the other hand, AH/EIN are definitive precancers that harbor most of the mutations present in the carcinomas to which they give rise.41–44 Thus, there must be “pre-AH/EIN” lesions that harbor neoplastic drivers, putting them at increased risk of progression relative to non-neoplastic endometria (mimics) with architectural disorder due to external factors such as higher circulating estrogen levels during perimenopause.1 Also, classic studies argued that endometria can be histologically stratified into diagnostic categories with increasing risks of neoplasia, including cases that do not reach current diagnostic criteria for AH/EIN.45

Although distinction between DPE and NAH in clinical specimens is commonplace, there is a continuum between the 2 and the distinction is likely to be poorly reproducible.13 Also, there are no agreed-upon criteria to distinguish them, and it is also likely that different groups/investigators utilize different criteria. Nonetheless, we found a greater incidence of marker aberrancy in NAH versus DPE (47.5% vs. 18.2%, P=0.0052), suggesting that more severe architectural abnormalities along a histologic spectrum are biologically significant, even if not highly reproducible.

In NAH, the 3 markers were aberrant in roughly equal proportions, whereas in DPE, Pax2 aberrancy was much more common. This is consistent with prior observations that Pax2 silencing can be an early or initiating event in endometrial carcinogenesis.16,17 Although definitive conclusions cannot be drawn, the absence of β-catenin aberrant cases in the ambiguous/difficult category is in agreement with prior observations that well-defined clusters of crowded glands (ie, clonal-type patterns) are often β-catenin positive and thus classifiable as AH/EIN even if minute.1 However, some DPE and NAH were β-catenin aberrant (Fig. 1B), arguing that CTNNB1 mutations can also be relatively early events in endometrial cancer progression.

Aberrancy for 2 of the core markers (β-catenin and Pten), as well as Arid1a and Mlh1, is associated with underlying oncogenic driver mutations. In contrast, Pax2 is believed to be silenced by epigenetic mechanisms, not classic gene mutations.1 Nonetheless, even for the factors that undergo mutations, NGS does not fully correlate with IHC aberrancy in individual cases per numerous studies. For example, most endometrial cancers with Pten protein loss have at least 1 PTEN mutation, but biallelic mutations or 1 mutation plus loss-of-heterozygosity would be needed to explain complete loss of protein. For 2 of the progressor cases with Pten loss (patients 1 and 2), truncating biallelic mutations fully rationalized Pten loss, but for several others only 1 PTEN mutation was detected (patients 3, 5, 6) (Fig. 8). Patient 8 harbored biallelic PTEN mutations, including a splice site alteration that may abolish protein synthesis from that allele. However, the other allele was a C136Y amino acid substitution. In the absence of additional experimental evidence, it is not possible to accurately predict if a missense mutation will destabilize the protein leading to apparent protein loss. Patient 6 harbored a monoallelic R130Q missense mutation, but there was no Pten loss by IHC. A previous study found that the R130Q mutation was not associated with IHC loss in 1 endometrial cancer,46 while another study found that an engineered R130Q mutant protein was as stable as the wild-type version.39 Thus, for patient 6, the retention of Pten despite the presence of an underlying mutation could be explained by a negligible impact of the R130Q missense variant on overall protein levels.

A major limitation of NGS not sufficiently emphasized in the literature is that it cannot detect deletions and chromosome level events that result in loss-of-heterozygosity and are particularly common for PTEN. Indeed, the PTEN locus was originally identified by mapping homozygous chromosomal deletions in tumors.47 Some studies have concluded that for Pten, IHC is more sensitive than NGS, likely reflecting the inherent limitations of NGS in detecting large-scale genetic events inactivating the locus.48 Other recent studies have similarly concluded that NGS and IHC are complementary in evaluating Pten inactivation, but also found that appropriately stringent IHC criteria can reliably identify biologically significant and clinically significant Pten inactivation in the majority of cases.38

For Pax2, the underlying mechanisms and functional consequences of inactivation are less clear, but widespread loss is associated with endometrial neoplasia and appears to be an early or initiating event.16,17Pax2 was the most frequently aberrant marker in DPE (Fig. 1), consistent with a role for Pax2 as an early or initiating event. However, patient 2 (Fig. 5) underwent Pax2 loss after Pten, demonstrating that order of aberrancy can vary and further suggesting that there is selection for Pax2 loss even in later stages of neoplastic progression.

Previous studies in support of the EIN system (see WHO 202028) have cited 50% gland density as a threshold for discriminating benign from neoplastic endometria. There is a practical necessity for thresholds to promote diagnostic uniformity among pathologists for optimal clinical management. However, any single parameter or absolute numerical threshold will have limitations, which can stem from specimen fragmentation and other difficulties in estimating the exact percentage (especially at or near the threshold) without image analysis. Even formal image analysis requires some selection of specific areas for quantitation, which can itself be subjective,11,12 see also Figure S2 (Supplemental Digital Content 2, https://links.lww.com/PAS/B524). More fundamentally, a single parameter is not likely to capture the considerable biological complexity of endometrial carcinogenesis, which is characterized by an unusually wide range of molecular driving events leading to aberrations in multiple biochemical pathways.24,34,36,37,49,50 With quantitative image analysis, we documented several examples of cases with gland density falling well short of this threshold, but which nonetheless exhibited clear-cut marker aberrancy and progressed to AH/EIN or cancer. Along these lines, the presence of definitive nuclear atypia or cytologic distinctiveness relative to adjacent glands is a helpful criterion but is not always present, and the 3-marker panel can be helpful in such situations. Our findings underscore that current histologic criteria can miss neoplastic processes detected by the 3-marker panel and also suggests that DPE and NAH are biologically divergent combinations of benign non-neoplastic entities, and more significant precursor lesions at increased risk for malignant progression. These findings are especially significant given that DPE and NAH are underrecognized as malignancy precursors, and there is no clinical consensus or evidence-based guidelines on optimal long-term management of DPE and NAH nor on the need for (with DPE) and timing of (with NAH) resampling in patients with these lesions.51,52

In some patients, the short time interval to “progression” and other findings suggest that incomplete sampling contributed to underdiagnosis at the time of biopsy (eg, patient 1, Fig. 4). While pipelle biopsy has very good sensitivity and is favored in the outpatient setting, small endometrial lesions may be undetected.53 This underscores some strengths of the 3-marker panel. Interpretation is across an entire biopsy, not limited to the most worrisome foci (making it useful even with limited samples), and it is able to detect aberrancy across a wide range of morphologic alterations as documented in this and prior studies.27,54

At the same time, the 3-marker panel has some shortcomings. In addition to the time and cost associated with 3 immunostains, it too has nuances in interpretation. Perhaps most significantly, some AH/EIN (<10%) are nonaberrant for the 3 markers (and the ancillary markers Arid1a and Mlh1).21 This explains patient 9 (Fig. 3) where a nonaberrant AH/EIN was preceded by a DPE that was similarly nonaberrant. Such cases demonstrate that misapplication of—or overdependence on—the 3-marker panel could lead to underdiagnosis of a definitive AH/EIN. Other unresolved issues are the utility and proper interpretation of the panel in other settings, such as the assessment of AH/EIN within EMPs. One recent study suggested a higher aberrancy rate for Pax2 within benign EMP relative to normal endometria,15 a result that merits further investigation. Nonetheless, we have found the 3-marker panel to be helpful in many cases including EMPs and its use has become routine in our practice.

It is intriguing to speculate that the 3-marker panel could be applied to DPE or NAH for improved risk stratification; however, additional studies are required to further explore this idea. This study has several significant limitations: it was a relatively small retrospective study from a single institution, and therefore, our findings may not be generalizable to all populations. Also, the absolute risk of progression cannot be reliably determined from this study. Thus, the results should be validated and further extended, ideally through a large, multi-institutional prospective study to permit more accurate risk stratification based on integrated molecular, marker, and morphologic features. Incorporating both pathologic and clinical factors in future studies will also be esse