OUP user menu

Utility of PAX2 as a Marker for Diagnosis of Endometrial Intraepithelial Neoplasia

Charles M. Quick MD, Anna R. Laury MD, Nicolas M. Monte, George L. Mutter MD
DOI: http://dx.doi.org/10.1309/AJCP8OMLT7KDWLMF 678-684 First published online: 1 November 2012


Diagnosis of endometrial intraepithelial neoplasia (EIN) requires learning new criteria. Two trainees rendered diagnoses based on biopsy findings, and then measured the effect of reviewing PAX2 on their interpretation. Fifty-two endometrial biopsy specimens diagnosed as having EIN were evaluated using EIN criteria. Background endometrial pattern, altered differentiation, and any features complicating diagnosis were noted. PAX2 stains were scored as confusing, helpful, or noncontributory. Fifty-two cases generated 104 passes; 82% were rediagnosed as EIN. The diagnosis was complicated because of altered differentiation (14%), EIN and background separation (13%), large lesions lacking background (11%), and secretory background (8%). PAX2 was most helpful in cases with secretory backgrounds and when EIN lacked adjacent normal tissue, and most confusing when scoring was ambiguous (14%). The diagnosis of EIN can be difficult when: (1) the lesion cannot be easily compared with background; (2) there is a confounding process; and (3) gland differentiation is altered. PAX2 can be of assistance in delimiting EIN lesions.

Key Words
  • PAX2
  • Utility
  • Endometrial intraepithelial neoplasia

Accurate diagnosis of premalignant endometrial disease, or endometrial intraepithelial neoplasia (EIN), requires the application of new criteria including gland area greater than stroma, cytologic alteration of crowded glands, size greater than 1 mm, and exclusion of mimics and carcinoma.1 Although these criteria were discovered using specialized techniques such as clonal analysis, computerized image analysis,2 and marker studies,3 they are intended to be applied to standard H&E slides viewed through a routine microscope. This works well in practice,4 but markers able to recognize EIN lesions provide a visual impression of clonal growth and precise delineation of lesion extent that lead to insights with educational and, perhaps in certain circumstances, diagnostic value.

Biomarkers that can be queried with immunohistochemistry have been described for EIN. Perhaps the best characterized are clonal loss of PTEN and PAX2 protein in 44% and 71%, respectively, of EIN lesions.5 Clinical use of both PAX2 and PTEN as specific markers for EIN has been constrained by the demonstration that both can be inactivated sporadically in normal endometrial tissues in the absence of a bona fide EIN lesion.5 In essence, changes in these genes can be early events occurring in small numbers of normal-appearing endometrial glands (“latent precancers”), which may or may not later evolve into high-risk EIN lesions. Nonetheless, when informative, PTEN and PAX2 can precisely delimit the extent of the EIN lesion in a majority of cases and facilitate appreciation of lesion characteristics relative to the background. In this regard, these markers are potential training tools for pathologists to recognize the patterns and distribution of clonal EIN lesions relative to the diffusely irregular field effects of unopposed estrogens. PTEN immunostaining is technically challenging, and its pancellular distribution sometimes confusing to interpret, creating a significant barrier to routine use. In contrast, PAX2 has proven to be technically robust and its normally strong nuclear staining pattern is very easy to score.

In several potentially confusing diagnostic contexts biomarkers capable of offsetting EIN from its background might be useful for diagnosis. Recognition of cytologic alteration of the premalignant glands requires comparison with normal background glands.6 Severely fragmented samples can lead to a physical separation of the premalignant and normal glands, making comparison between the 2 populations difficult. Alternatively, the focus of EIN may overrun the entire sample, thus obscuring residual normal glands. Normal secretory endometrium may exhibit patchy gland crowding at a density overlapping that of EIN. When EIN is present in a secretory endometrium, the recognition of subtle cytologic changes points one to the correct diagnosis. Endometrial polyps can also harbor foci of irregularly packed benign glands as well as EIN.7 Only the careful observation of coordinated changes in both cytology and architecture relative to polyp background permits accurate diagnosis. Altered differentiation of the premalignant glands, that is, mucinous, eosinophilic, or squamous metaplasia (morules), by definition changes the cytologic appearance of the glands themselves. Discrimination of the subset of metaplasias that are premalignant EIN lesions requires consideration of the background context, and larger-scale features of affected glands such as crowding and geographic distribution.

In this study, we evaluate the usefulness of PAX2 immunostaining as an educational and diagnostic tool for interpreting endometrial biopsy findings with EIN arising in various settings. Two postdoctoral trainees, relatively new to established EIN criteria, diagnosed routinely stained (“H&E”) EIN-bearing endometrial biopsy specimens and then assessed their impressions of corresponding PAX2 staining results.

Materials and Methods

Fifty-two endometrial biopsy specimens with a pathology report diagnosis of EIN made at Brigham and Women's Hospital (Boston, MA) were recut and stained with H&E and PAX2. Full details of case selection, PAX2 staining, and the finding of 71% (37/52) PAX2 null phenotype have been reported previously.5 In addition to the diagnosis of EIN, which was the basis for case retrieval, all available slides were rereviewed by an expert senior pathologist (G.L.M.) to confirm the EIN diagnosis. One recut H&E-stained section and 2 PAX2-immunostained sections from each of the original 52 cases were used for the current study. No follow-up data or specimen was sought for this study.

Cases were presented independently to each of 2 pathology trainees (C.M.Q. and A.R.L., in years 6 and 4, respectively, of their post–medical school pathology training). The trainees initially rendered a blinded diagnosis on the H&E slide alone using standard EIN criteria.1,6 Criteria included (1) area of glands greater than that of stroma, (2) altered cytologic features of crowded focus compared with background endometrium, (3) size of at least 1 mm in a single fragment, and (4) exclusion of mimics of EIN. Features that potentially confounded the diagnosis of EIN were noted, including small sample size, tissue fragmentation, physical separation of lesional and background endometrium, lack of normal background glands for comparison, and subdiagnostic cytologically altered glandular areas. In addition, altered differentiation (eg, mucinous, eosinophilic, or squamous metaplasia) in either the lesion itself or background endometrium was recorded. Diagnoses were separately recorded for the background endometrium, including presence of endometrial polyps, anovulatory, or cycling endometrium.

PAX2 staining results were subsequently reviewed independently by the trainees. For each case, the H&E and PAX2 stains were reviewed side by side to evaluate whether PAX2 was useful in identifying neoplastic EIN glands. The trainees were required to record whether the PAX2 was helpful, noncontributory, or confusing in reaching a diagnosis of EIN. PAX2 stain was defined as helpful in cases in which a questionable or difficult H&E diagnosis of EIN was reinforced by concurrent PAX2 review. PAX2 stain was noncontributory in cases in which H&E diagnosis was clear and PAX2 staining was superfluous. PAX2 staining was recorded as confusing when a clear H&E diagnosis was not reaffirmed with the PAX2 stain. In addition, reasons why PAX2 was helpful or confusing in the aforementioned situations were recorded in free-text format Table 1.

View this table:
Table 1


Impressions of the 52 cases were recorded by the 2 trainees, for a total of 104 diagnostic “passes.” Most of the H&E diagnoses were either EIN (82%) or crowded glands subdiagnostic of EIN (12%). In addition, secondary diagnoses of the background endometrium included anovulatory endometrium (23%), endometrial polyp (14%), and secretory endometrium (10%) Table 2. It is worth noting that both reviewers concluded that gland crowding, as opposed to EIN, was present in 6 of the 52 cases. In most of these cases, diagnostic areas seen on the original slides were depleted in deeper levels available for the current study.

View this table:
Table 2

The most common confounding factor affecting the H&E diagnosis of EIN Table 3 was altered differentiation (“metaplasia,” 14%). Metaplasia consisted of eosinophilic, tubal, mucinous, or squamous changes of the neoplastic epithelium. In several cases, combinations of metaplasia were noted. Slightly less frequently (13%), fragments that were diagnostic of EIN were physically separated from normal background, thus making the evaluation of altered cytology, one of the diagnostic features of EIN, difficult. A similar situation was noted when extensive EIN lesions were present, resulting in minimal or absent comparison with background normal endometrium (11%). The least common, albeit most difficult, confounder was a background of secretory endometrium (8%) Image 1. Secretory EINs were composed of compactly arranged glands with an altered cytology characterized by a combination of abundant eosinophilic cytoplasm and nuclei that were usually smaller than those seen in the adjacent normal endometrial glands Image 2.

View this table:
Table 3
Image 1

Corresponding specimens stained with H&E (A) and PAX2 (B) demonstrating endometrial intraepithelial neoplasia in a secretory endometrial background with rare “normal” glands (arrow, PAX2 detail) for comparison.

Image 2

Secretory endometrial intraepithelial neoplasia displaying eosinophilic cytoplasm compared with the overrun gland (arrow).

PAX2 staining was helpful in 53%, confusing in 41%, and noncontributory in 6% of cases Table 4. PAX2 was most helpful when there was a distracting background of secretory (Figure 1) or anovulatory Image 3 endometrium. In these cases, PAX2 staining delineated the areas suspicious for EIN. PAX2 staining was also deemed helpful in cases in which normal background endometrium was minimal or unavailable for comparison. This included cases in which the focus of EIN was in fragments separate from the normal background, EIN was extensive with little comparison background, or biopsy specimens were scant with only rare background glands.

View this table:
Table 4
Image 3

Corresponding specimens stained with H&E (A) and PAX2 (B) demonstrating a dispersed geographic area of cytologically altered glands (arrows) with a low volume to percentage of stroma; however, the atypical glands are PAX2 null.

In several specific circumstances, PAX2 staining was confusing. This occurred in 4 passes (4%) when immunoreactivity for PAX2 was ambiguous. Typically these cases contained glands with attenuated, but non-null PAX2 immunoreactivity compared with the adjacent background glands Image 4. Discrepancies between the extent of EIN lesion distribution as perceived with H&E compared with PAX2 stains was another problem. In 11 passes (11%), the biopsy result was diagnostic of EIN on H&E, but the extent of PAX2 loss was much more widespread than expected.

Image 4

Corresponding specimens stained with H&E (A) and PAX2 (B) demonstrating crowded glands with decreased (arrow), but not null, expression of nuclear PAX2 staining relative to adjacent normal glands (arrowhead).

Diagnostic dilemmas occasionally arose when scattered PAX2 null glands or small, subdiagnostic clusters of PAX2 null glands were encountered in cases that were nondiagnostic of EIN on H&E staining Image 5; however, PAX2 staining was generally helpful in cases of cytologically altered crowded glands subdiagnostic of EIN. This was the case in 12% (6/52) of cases having small foci of crowded glands that were suspicious for EIN but lacked either gland density or size necessary to meet EIN diagnostic criteria Image 6. Negativity for PAX2 was seen as helpful in classifying these small foci as distinct from the background. A description of “crowded focus of cytologically altered glands nondiagnostic of EIN” with a suggestion for appropriate clinical follow-up could thus be made.8

Image 5

Corresponding specimens stained with H&E (A) and PAX2 (B) demonstrating endometrium that is nondiagnostic for endometrial intraepithelial neoplasia with “distracting” PAX 2 null glands (arrows) randomly distributed between normal glands.

Image 6

Corresponding specimens stained with H&E (A, C) and PAX2 (B, D) demonstrating a small focus of crowded glands that are subdiagnostic for endometrial intraepithelial neoplasia, but show loss of nuclear PAX2 staining relative to adjacent normal (arrow) epithelium. Small size of cytologically altered focus may be because of limited lesion extent (A, B) or fragmentation (C, D).


We confirmed several specimen characteristics in which EIN diagnosis is difficult for the inexperienced pathologist, and found that PAX2 stain is helpful in some, but not all, circumstances. Diagnostic criteria for EIN are relatively new in the arena of gynecologic pathology. Implementation requires education (or reeducation) of the diagnostic pathologist in the proper application of previously unknown criteria that have emerged from recognition of its clonal character (lesion size and direct comparison of lesional glands with flanking normal glands) and computerized image analysis (gland area exceeds stromal area). The current study was developed with this in mind, and it examined the process of learning the diagnostic criteria for EIN along with application of immunohistochemistry to visualize lesion distribution in difficult cases.

PAX2 is a consistent and robust immunostain for a nuclear transcription factor normally expressed in the endometrial epithelium. Its inactivation in the majority (71%) of EIN lesions, availability of companion background glands as a convenient internal control, and crisp nuclear distribution are all factors facilitating its use as a delimiter of EIN lesions. PAX2 staining was technically adequate and scorable in the majority (96%) of cases studied. Observed technical failures involved weak staining or staining that was patchy in individual glands.

Lesions in cycling endometrium and those with altered differentiation presented difficult H&E interpretations that were improved by the addition of PAX2. Loss of PAX2 staining was especially helpful in EINs presenting against a secretory background. In these lesions, the premalignant glands were very similar to the background seen on H&E staining. Because of this subtle morphologic alteration, PAX2 loss highlighting the premalignant glands was considered very helpful. Similarly, in cases of EIN with altered differentiation (“metaplasia”), PAX2 staining was sometimes useful for demonstrating the cohesive geographic distribution of the clonal affected glands, which must be distinguished from reactive or hormonal response effects.

Another common diagnostic confounder was small and/or fragmented tissue samples. For this problem, addition of PAX2 into the diagnostic algorithm was of marginal benefit, because it did not obviate the underlying dilemma of a poor specimen.

PAX2 was useful in biopsies containing microscopic foci of cytologically altered glands in which the lesion size was insufficient for a diagnosis of EIN. PAX2 staining sometimes reaffirmed the cytologic impression of a cluster of abnormal glands. However, the pathologist must be careful not to rely on PAX2 staining patterns as sufficient for EIN diagnosis, but always require diagnostic criteria to be met on the H&E slide. Rather, a recommendation for followup sampling is appropriate.

PAX2 staining is not a panacea for resolution of all, or even most, EIN lesions. A third of EIN lesions do not exhibit loss of this marker, so its inactivation is not essential for EIN diagnosis. In other cases, PAX2 findings were discordant with the H&E impression, such as when PAX2 loss appeared over a region either greater or less than the lesion extent apparent on the H&E. These examples highlight the need for caution in using PAX2 as a diagnostic tool, because relying too heavily on PAX2 loss may lead to over- or underdiagnosing of apparent lesions.

Both trainees agreed that PAX2 was useful as an educational tool for learning the criteria for EIN diagnosis in a retrospective setting. Exceptions were cases with discrepant or technically suboptimal staining. The greatest educational benefit was gained in cases with very subtle histologic differences between the benign and premalignant glands (ie, secretory morphology or metaplasia). By comparing the H&E and PAX2 profiles, the lesions were highlighted and the differences better appreciated. Similarly, confidence in recognizing subdiagnostic discrete foci requiring followup was improved with PAX2 correlation. In cases without readily visible background, either because it is overrun by extensive EIN or so altered by excess estrogen as to be unrecognizable as background, morphologically atypical crowded glands could better be defined as such even if the diagnostic criteria for EIN were not met.

There is, however, educational value in using small, prescreened training sets of dual-stained EIN lesions. In particular, selected cases are helpful in illustrating patterns of lesion distribution and the variable character of EIN cytology and architecture. PAX2 immunostaining may have clinical value in carefully selected cases in which contrast between the lesion and background are represented in different areas of the sample or obscured by a secretory background.


This study was supported by National Institutes of Health (Bethesda, MD) grant RO1-CA100833 (Dr Mutter).


View Abstract