Endometrial carcinoma is the most common form of genital malignancy diagnosed in women. Simple hyperplasia without atypia lies closest to normal proliferative endometrium. On the other end of the spectrum of preinvasive neoplasms is complex atypical hyperplasia (CAH), which from biologic, morphologic, and clinical perspectives resembles well differentiated endometrioid adenocarcinoma (WDA). Because CAH and WDA share so many important characteristics, tests for any one or a combination of these cannot definitively separate the two entities. Morphologically, CAH resembles WDA and, in many cases, substantial diagnostic sophistication is required to distinguish between them using current criteria. Two noteworthy articles based on the findings of the Gynecologic Oncology Group Protocol 167 have been published in this edition of Cancer and provide a glimpse of one group's approach to the diagnosis of CAH and its relevance.
See also pages 804–11 and 812–9.
Endometrial carcinoma is the most common form of genital malignancy diagnosed in women. Tens of thousands of hysterectomies are performed annually in the U.S. for this disease and its precursor, endometrial hyperplasia. The most prevalent classification of endometrial hyperplasia, used by the World Health Organization and the International Society of Gynecological Pathologists, is based on the seminal work of Kurman et al.1 and includes simple and complex forms, with and without atypia. Simple hyperplasia without atypia lies closest to normal proliferative endometrium. On the other end of the spectrum of preinvasive neoplasms is complex atypical hyperplasia (CAH), which from biologic, morphologic, and clinical perspectives resembles well differentiated endometrioid adenocarcinoma (WDA). Because CAH and WDA share so many important characteristics (e.g., microsatellite instability and mutations of PTEN, CTNNB1, and k-ras), tests for any one or a combination of these cannot definitively separate the two entities. Morphologically, CAH resembles WDA and, in many cases, substantial diagnostic sophistication is required to distinguish between them using current criteria.
The presence of CAH indicates a significant risk for coincident or subsequent adenocarcinoma. Both diagnoses result in hysterectomy in postmenopausal patients and, when appropriate, both diseases can be treated with progestational agents. A recent Gynecologic Oncology Group (GOG) protocol (LAP2) involving pelvic and paraaortic lymph node dissection for every patient diagnosed with WDA (but not CAH) reflects the most significant difference between the two from the standpoint of clinical management. WDA patients are offered hysterectomy and lymph node dissection, whereas CAH patients commonly undergo hysterectomy without lymph node dissection. However, the necessity for lymph node dissection in every patient with WDA has been debated.
Two noteworthy articles based on the findings of GOG Protocol 167 have been published in this edition of Cancer—one by Zaino et al.2 and the other by Trimble et al.3 These articles provide a glimpse of one group's approach to the diagnosis of CAH and its relevance. GOG Protocol 167 had two objectives: determining the interobserver variability in the diagnosis of CAH2 and the frequency of carcinoma in hysterectomies performed shortly after a diagnosis of CAH based on biopsy or curettage (close follow-up hysterectomies).3 Biopsy and curettage slides from patients diagnosed with CAH were forwarded to GOG review pathologists who either agreed with the original diagnosis or reclassified it as “less than CAH” or “carcinoma.” The review pathologists were instructed to use the original diagnostic criteria of Kurman et al.1 Agreement between two of three review pathologists was defined as a consensus. The reviewers then examined hysterectomy slides while remaining blinded to their own previously rendered diagnoses.
Zaino et al. reported substantial differences of opinion between referring pathologists and GOG review pathologists, and among the reviewers themselves.2 The consensus review diagnosis supported the referral diagnosis of CAH in only 39% of cases. Unanimous agreement among review panelists for any diagnosis (CAH, less than CAH, and carcinoma) was attained in only 40% of cases. These results are not surprising. In a recent article, Kendall (from Kurman's group) et al. described an experiment in which gynecologic pathologists from the same institution blindly reviewed endometrial biopsies and curettages that had been diagnosed as proliferative endometrium, hyperplastic endometrium, or endometrial adenocarcinoma.4 The interobserver and intraobserver agreement for hyperplasias less than CAH was excellent, as was agreement regarding what constituted carcinoma. However, there was suboptimal agreement with respect to the diagnosis of CAH. Aside from the prominence of nucleoli, there was no consensus regarding which features were sufficient to consider a hyperplasia “atypical.”
Trimble et al. reported that after a consensus review diagnosis of CAH on biopsy or curettage, 39% of hysterectomies contained adenocarcinoma and 33% of these were myometrial invasive.3 The review panel considered 29% of all cases to be diagnostic of carcinoma on biopsy or curettage,2 cases that were followed by a finding of residual carcinoma in 64% of hysterectomies. The panel downgraded a number of cases (from the total) to “less than CAH,”2 but even these were followed by carcinoma discovered at hysterectomy in 19% of the cases. The reviewers were unable to reach a consensus in 6% of the total cases.2 These were likely to yield carcinoma on hysterectomy (63%), with the majority being tumors with high-risk features such as myometrial invasion and/or moderate or poor differentiation. A total of 43% of hysterectomies performed for an original diagnosis of CAH harbored adenocarcinomas, and 31% of the carcinomas were myometrial invasive.3
This prevalence of myoinvasive carcinoma noted in a group believed to have CAH is astonishing. If anything, there has been a historical bias toward diagnosing what appears to be CAH as adenocarcinoma, thereby leaving CAH a mostly clinically inert entity, a neoplasm without the capacity for myoinvasion or metastasis. In the article by Trimble et al., Table 5 allows comparison with previous retrospective studies of CAH. The rates of myometrial invasive carcinoma after a diagnosis of CAH ranged from 2–39%, which is similarly surprising.
Based on the GOG Protocol 167 articles,2, 3 we can conclude that CAH is not diagnosed reproducibly and that a preoperative diagnosis of CAH is frequently followed by carcinoma (even myometrial invasive carcinoma) in a follow-up hysterectomy. However, we still do not know why this is the case. Are the current diagnostic criteria for CAH flawed? Were there significant sampling errors? Might there be other unexplored issues?
There were several methodological problems with these studies: a lack of criteria regarding what constituted a representative sample; an assumption that the review panel had examined all the material on which the original diagnosis was based; the probability that the referral pathologists applied CAH criteria that differed from those of the review panel; the possibility that the diagnostic criteria for WDA were not applied as rigorously in hysterectomy specimens compared with in curettage specimens; and a lack of retrospective validation that would have allowed the review pathologists to determine whether missed WDAs were the result of sampling error or interpretive error. Further complicating the issue of sampling error was the inclusion of both curettage and biopsy specimens for review. Authors of future studies on the topic should address these points. We also need to be mindful of the possible clinical repercussions of the results of these studies. It would be best to avoid diagnosing myoinvasive carcinomas as CAH so as to lessen the impulse to perform upfront hysterectomy and lymph node dissection after a preoperative diagnosis of CAH.
The results of the studies by Zaino et al.2 and Trimble et al.3 also emphasize the need to more precisely define CAH and the criteria for its diagnosis. We should strive for consensus regarding what constitutes this entity and what type of information a diagnosis of “CAH” should convey. The criteria or diagnostic test should be informative, relevant, easy to apply, reproducible, and affordable. The biologist might demand that CAH possess some but not all of the molecular characteristics of WDA or have additional unique features. The clinicians will expect a disease that is for the most part hormonally responsive. It would persist in untreated cases and progress to carcinoma in some cases. I believe they would be surprised and disappointed if a hysterectomy performed after a recent biopsy with CAH yielded myometrial invasive carcinoma. The pathologist would expect a morphologic continuum between CAH and WDA but most likely would be tempted to draw between the two rigid diagnostic lines, perhaps derived empirically. Even after this scrutiny, we might still be left with fundamental questions: is CAH a disease entity (akin to cervical squamous dysplasia) or is it a descriptor that denotes abnormality and a risk for an undetected, clinically relevant neoplasm (such as atypical squamous cells of undetermined significance [ASCUS])?
There have been some laudatory attempts in the pathology community to refine the concept of CAH and, in the case of Mutter, to define a parallel but not entirely equivalent entity, endometrial intraepithelial neoplasia (EIN).5, 6 Development of the D-score through work first developed by Baak et al.7 and then modified to EIN by the combined groups of Mutter et al. and Baak et al.5, 6 is attractive because it is based on morphology and appears increasingly easier to put into practice. Biologic correlates, including the concept of the PTEN-null gland, have been described to date.5 These criteria are an attempt to define the lower limits of CAH/EIN, but not those that separate CAH/EIN from WDA. In addition, the EIN criteria define a lesion that is at some risk for progressing to carcinoma (precursor lesion),6 but to our knowledge relatively less attention has been focused on the rates of coincident carcinoma discovered during close follow-up hysterectomies (whether all EINs are WDA marker lesions). Therefore, we cannot determine whether immediate hysterectomy should be the standard of care for all patients with EIN. This has all engendered considerable controversy. It is noteworthy that large studies correlating endometrial biopsies and/or curettages with findings on close follow-up hysterectomy are to our knowledge lacking, but these data are now beginning to be reported.8
Longacre et al. eschewed the empiric approach and based their distinction of CAH from WDA on morphologic features that correlated retrospectively with myometrial invasion specifically, a clinically relevant “gold standard.”9 Using their criteria, myoinvasive carcinomas should be extraordinarily rare in hysterectomies after a diagnosis of CAH. In contrast to the studies of Mutter and Baak et al.,5–7 these criteria assist in distinguishing CAH from clinically relevant WDA, but they do not address the distinction of CAH from nonatypical hyperplasias or the risk of disease progression to carcinoma. At first glance, the criteria of Longacre et al.9 appear complicated, but the use of a pattern recognition chart that recalls the Gleason pattern chart (used in grading prostate carcinoma) and subsequent validation with practitioners of varying skill levels indicate that these criteria should be rather straightforward to apply. To our knowledge, biologic correlates have not been studied herein nor have the criteria been applied prospectively in a multiinstitution trial to determine their power.
The approaches of Mutter,5 Baak et al.,6, 7 and Longacre et al.9 should be studied in depth before any changes are made to the CAH category or the criteria that commonly define it, and I reiterate my recommendation concerning the need to clarify what type of information the designation of “CAH” should convey. It would be informative to redo the expert panel review segment of GOG Protocol 167 keeping in mind the importance of defining an adequate or representative sample using pathologists schooled in the approaches of Mutter, Baak, and Longacre.5–9 If consensus was reached, I would propose devising training sessions to be followed by tests of interobserver agreement and feasibility. It would be a worthy challenge.
The author thanks George Monemvasitis and Milicent Cranor for their editorial assistance.