Preinvasive Endometrial Neoplasia

Definition

EIN is a clonal proliferation of architecturally and cytologically altered premalignant endometrial glands, which are prone to malignant transformation to endometrioid (type I) endometrial adenocarcinoma. EIN lesions are noninvasive, genetically altered neoplasms that arise focally and may convert to a malignant phenotype on acquisition of additional genetic damage. Diagnostic criteria for EIN have been developed by histopathologic correlation with clinical outcomes, molecular changes, and objective computerized histomorphometry. EIN is conceptually similar to complex atypical hyperplasia (CAH) of the endometrium, and most EINs overlap with CAH. However, as defined, EIN is not exclusive to CAH, and not all CAHs are EIN lesions. This is due to differences in classification criteria and an inconsistent application of hyperplasia diagnosis by pathologists, as will be discussed later.

EIN should not be confused with unrelated serous EIC, which is an early phase of type II papillary serous adenocarcinomas of the endometrium. This is particularly important from the clinical perspective inasmuch as clinicians must be made fully aware of the risk attendant to a diagnosis of a noninvasive serous carcinoma (serous EIC).

Historical Background

In 1949, Arthur T. Hertig, a pathologist at Harvard Medical School and Boston Hospital for Women, described adenomatous hyperplasia and anaplasia of the endometrium as precancerous lesions that may antedate the appearance of endometrial adenocarcinoma by 1 to 5 years. At that time, estrogens were recognized as participants in the promotion of endometrial carcinoma, but there were inadequate tools to resolve field effects resulting from systemic hormonal exposures from neoplastic proliferation caused by somatic mutation. This remained a major source of confusion and contention for the next 50 years, during which the pace of nomenclature proposals outstripped actual advances in the understanding of this disease.

A 1985 landmark paper showed that subjective cytologic atypia within a lesion increased cancer risk 14-fold, the single variable conferring the largest incremental endometrial cancer risk discovered up to that time. In 1994, the World Health Organization (WHO) sanctioned a classification system whereby four classes of hyperplasia were divided by architecture (complex or simple) and cytology (nonatypical or atypical). Atypical endometrial hyperplasias, whether of simple or complex architecture, were widely construed as premalignant lesions requiring therapeutic ablation. The dominance of cytology in risk assessment has since been tempered by the realization that pathologists are unable to classify individual lesions reproducibly as atypical versus nonatypical and by the recognition that architectural features have a predictive value at least as important as cytology.

A parallel effort to use computers to classify low versus high cancer risk endometrial lesions played out against this backdrop of pathologist-derived classification systems. The availability of software and hardware for computerized tissue morphometry in the late 1970s enabled a greater degree of objectivity in the histopathologic description of premalignant endometrial lesions than previously possible. When statistically modeled against clinical outcomes of interest, individual cytologic and architectural predictive variables measured in a hematoxylin and eosin (H&E)–stained slide could be discovered and validated. This reached its pinnacle with the development of the deviation (D) score, a morphometric scoring system with superb cancer-predictive performance exceeding anything seen previously. The architectural variable of volume percentage stroma, a measure of gland crowding central to the d score, has since been incorporated as a key subjective EIN criterion (more glands than stroma) that can be understood and interpreted visually by pathologists without recourse to a computer.

Microdissection of targeted small lesions in paraffin-embedded tissues has enabled a precise correlation between molecular genetic and histopathologic features of endometrial precancers. Molecular studies showing a clonal growth pattern for endometrial precancers have established their focality of origin and subsequent centripetal expansion as geometric properties of evolving lesions. By the time such clones can be seen under the microscope, there are evident changes in cytology and architecture that offset them from the background. Lesion size and contrasting localizing lesion cytology and architecture to that of the background endometrium are characteristics that must be assessed during routine diagnosis and have been incorporated into EIN diagnostic criteria.

A formal proposal for the EIN diagnostic schema emerged from the realization that the histologic changes seen in genetically altered monoclonal endometrial precancers were identical to those previously defined by computerized morphometry of H&E-stained sections as increasing the likelihood of carcinoma. A high degree of clinical relevance was immediately justified by a series of preexisting clinical outcome studies, all of which validated the cancer-predictive value of the original D-score–based diagnostic algorithm. Just as diverse and independent methods of discovery were integrated to define the group of lesions that we now know as EIN, this is an entity that may be diagnosed in several ways. For those with access to the commercially available QPRODIT software and morphometry workstation (Leica, Cambridge, England), EIN lesions are those with a D-score less than a threshold of 1, which identifies lesions with a 27% chance of concurrent or 46-fold increased risk for future endometrial carcinoma ( Fig. 17.1 ). Practical implementation of subjective criteria for EIN diagnosis by pathologists working at a standard microscope with routine H&E slides is the subject of the remainder of this chapter.

Implementation of the Endometrioid Intraepithelial Neoplasia Schema by the World Health Organization in 2014

EIN diagnostic criteria were sufficiently different from those used in the 1994 version of the WHO four-class hyperplasia schema that a formal adjustment of terminology and criteria was required. This was implemented as a new standard in 2014–2015 ( Table 17.1 ) on recommendation of WHO and the American College of Obstetricians and Gynecologists. In the current two-class 2014 WHO hyperplasia schema, hormonal field effects are designated as “Endometrial hyperplasia without atypia.” Correspondingly, precancerous high-risk lesions, documented extensively as EIN, were so designated. One minor refinement of the WHO was changing the endometrial to endometrioid to clarify its relevance to the general class of endometrioid carcinomas (type I endometrial adenocarcinomas). One other nomenclature accommodation in WHO 2014, made to facilitate transition from their legacy system, is that the high-risk group of EIN may be synonymously designated as EIN or as “atypical endometrial hyperplasia.” As might be expected, diagnostic reproducibility improves when using the 2014 compared to the 1994 version of the WHO criteria because fewer groups require distinction, and criteria are better defined.

Despite the nonatypical versus atypical terminology option in WHO 2014, pathologists must keep in mind that cytologic appearance alone is insufficient to discriminate among these entities. Rather, new diagnostic criteria, unavailable at the inception of the now-replaced 1994 four-class WHO schema, must be applied, according to the published EIN evidence base. These include minimum size of localized lesions, objectively defined gland crowding thresholds, and comparison of lesional with background glands to identify significant cytologic changes. For these reasons, entities named in the 2014 WHO schema do not correlate absolutely with earlier 1994 WHO classes. Fig. 17.2 shows general WHO 1994 to WHO 2014 correlations based on the combined experience of multiple European and American groups.

Endocrine Factors

Endocrine risk factors for endometrial precancers are essentially identical to those for endometrioid endometrial adenocarcinoma, with estrogens acting as promoters and progestins as protectors. In the Postmenopausal Estrogen–Progestin Interventions (PEPI) trial, 12% of women receiving unopposed estrogens developed atypical hyperplasia over the 3-year surveillance period compared with 0% of placebo controls. Estrogen risks are obviated by the addition of progestins such as medroxyprogesterone acetate, which protects against the development of endometrial hyperplasia. When administered in a combined low-dose oral contraceptive formulation, this may reduce the endometrial cancer risk below that of the population background.

Genetic Factors

Sporadic EIN lesions are precursors to the subset of endometrial carcinomas characterized by endometrioid differentiation, PTEN gene and PAX2 inactivation, and microsatellite instability. Precancer histology seen in women with a hereditary endometrial cancer risk go through a comparable premalignant phase of EIN, including PTEN inactivation. Hereditary nonpolyposis colon cancer (HNPCC), caused by abnormalities of factors involved in DNA mismatch repair, is an autosomal dominant condition conferring a 60% lifetime risk for endometrial cancer, generally of the endometrioid type, with more than 75% showing microsatellite instability and 68% having a loss of PTEN tumor suppressor function. These tumors occur on average about 30 years earlier than their sporadic counterparts, leading to a current management recommendation to begin annual or semiannual endometrial biopsies by the age of 30 to 35 years. A second heritable cancer syndrome characterized by an elevated risk for that type of endometrial cancer—with a premalignant EIN phase—is Cowden syndrome, caused by germline transmission of a mutant PTEN allele.

Early detection and treatment of premalignant endometrial disease is a mainstay of endometrial cancer therapy. The lifetime risk for endometrial cancer is 2.4% in the United States, primarily a sporadic disease driven by complex interactions between somatically acquired genetic lesions and ambient hormonal selection factors. Most endometrial cancers are discovered when the patient develops symptomatic bleeding, followed by a diagnostic endometrial biopsy. Under these circumstances, 21% of endometrial adenocarcinomas at the time of initial diagnosis have already extended beyond the subjacent myometrium, with extension to the cervix (stage 2; 5.8%), regional nodes or extrauterine tissues (stage 3; 7.7%), or distant sites (stage 4; 8.3%). If detected earlier, many of these patients could achieve surgical cure by hysterectomy alone.

Epidemiologic Factors

EIN has risk factors similar to those of endometrioid endometrial cancer. Specifically, women who are obese are at an increased risk to develop EIN. This relationship between obesity and endometrial proliferation is most likely secondary to excess peripheral conversion of estrogens in adipose tissue. Other epidemiologic risk factors include insulin resistance and polycystic ovary syndrome (PCOS).

Medication Factors

More women are taking tamoxifen now than ever before based on data suggesting that tamoxifen may decrease the incidence of breast cancer in women who are at high risk, but do not necessarily have the disease themselves. Although tamoxifen acts as an estrogen antagonist in the breast, it has been associated with various benign and malignant alterations of the endometrium, including EIN.

Hysteroscopy and Sonography

Endometrial visualization by hysteroscopy or vaginal ultrasonography can be a useful adjunct to biopsy, but practice in this regard is not standardized. Hysteroscopically guided endometrial biopsies have been increasingly performed in clinical settings where practitioners are comfortable with this technology. Hysteroscopically guided resection has also been used for the purpose of fertility preservation. A smooth regular lining by hysteroscopy can reassure the clinician that an occult carcinoma was not missed, and direct visual guidance can improve access to remote regions of an irregularly shaped cavity. Transvaginal ultrasound is an insensitive and nonspecific screen for endometrial cancer, and there are few studies on the ability to detect physically small EIN lesions. Cancer detection sensitivity for transvaginal ultrasound with a threshold endometrial thickness of 6 µm is only 17%, and it is 33% for a threshold value of 5 mm. The specificity is very low, making this an expensive (during follow-up of numerous false-positives) and insensitive test. Ultrasound will, however, identify conformational abnormalities of the uterine cavity that may complicate access for sampling.

Endometrial Sampling

Endometrial biopsy and curettage remain the primary diagnostic modalities to evaluate potential endometrial disease. These are invasive procedures that can cause cramping and bleeding and carry minimal risks of uterine perforation or contamination of the cavity by pathogens. For these reasons, endometrial sampling cannot be considered a screening test but rather a procedure undertaken in response to specific symptoms or cancer risk factors. The most common setting in which an EIN-yielding endometrial biopsy is performed is a workup of menstrual irregularity in a perimenopausal woman or symptomatic bleeding in a postmenopausal patient. Introduction of the Pipelle biopsy apparatus, which unlike curettage does not necessitate cervical dilation and anesthesia, has greatly reduced the cost and morbidity of endometrial sampling. Outpatient, office-based, Pipelle biopsies are now the most commonly performed endometrial sampling procedure.

Sampling Devices

Transcervical endometrial sampling is the mainstay of endometrial diagnosis. Sampling adequacy is of particular concern with localizing EIN lesions that are not uniformly represented in all fragments. Coverage is affected by the device used, guidance by hysteroscopy or ultrasound, operator performance, and uterine anatomy. Short of hysterectomy, no sampling strategy is foolproof. If there is any clinical or pathologic concern that the available sample is inadequate or nonrepresentative, the diagnostic process is incomplete, and resampling should be undertaken. It is incumbent on the pathologist to have a clear understanding of sources of sampling and interpretative errors and how these influence options for follow-up diagnostic procedures and to communicate clearly to the clinician.

Curettage by a rigid sharp device is the most established method but has the drawback of requiring cervical dilation and attendant anesthesia. For lesions whose distribution or site of origin is particularly relevant to management, separate acquisition of endometrial and endocervical curettage fractions (fractional curettage) has traditionally been used; however, currently it is not relied on, and the determination of site of origin (cervix vs. endometrium) is determined by histology, immunohistochemistry, and imaging. In addition, curettage is not without its limitations, sampling less than half of the uterine cavity in about 60% of cases.

The Pipelle biopsy instrument, a 3.2-mm diameter flexible cannula that aspirates a core of tissue of approximately 1.5 to 2 µm ( Fig. 17.3 ) as it sweeps across the endometrial surface is small enough that cervical dilation is unnecessary. Because no anesthesia is needed, it may be used readily in a private office setting and is more comfortable to patients than a sharp curettage. Vabra aspiration biopsy is not recommended, and there is widespread use of the Pipelle instrument in many outpatient contexts.

There are now many studies comparing the endometrial sampling adequacy of Pipelle biopsy to curettage. In aggregate, tissue adequacy and diagnostic accuracy weigh in favor of the Pipelle, with a few notable caveats. Physical mass lesions that impinge on the uterine cavity, such as polyps or uterine leiomyomata, may deflect the flexible Pipelle device and lead to blind spots. Insight regarding these underlying conditions may be gained by ultrasonographic studies or physical examination, and sampling errors may be reduced by the use of a rigid sampling device or hysteroscopic guidance. Overall sensitivity and specificity of an endometrial precancer (atypical hyperplasia) diagnosis is 82% to 100% for Pipelle and 67% to 99.8% for Vabra aspiration. Detection of endometrial carcinoma, generally bulky lesions, is even more sensitive—99.6% for Pipelle and 97.1% for Vabra aspiration.

Hysteroscopic biopsies are the least abundant tissue biopsy format, relying on accurate target selection rather than coverage extent to minimize sampling error. In the case of physically small, grossly inapparent EIN lesions, a larger random sample may be advantageous. Small, often crushed, tissue fragments yielded by tiny jaws of the hysteroscopic biopsy device present a diagnostic challenge to the pathologist. Not only is interpretation usually compromised by artifact, but the background context necessary to recognize a localizing EIN lesion may also be missing or poorly represented. Cauterizing sampling devices should be avoided whenever possible because the low resistance of fluid-filled lumens directs the current specifically to glands, causing severe distortion.

Rationale

Much as the German pathologist Koch developed a series of postulates that must be fulfilled to prove the pathogenesis of disease scientifically by a specific infectious organism, similar postulates may be formulated for premalignant disease. Here we list criteria to be fulfilled for a clinically relevant and biologically distinctive category of endometrial precancers, all of which have been met in the case of EIN.