Portal de Periódicos da CAPES

The Fruits of our Labors: Distinguishing Endometrial From Endocervical Adenocarcinoma

2002; Lippincott Williams & Wilkins; Volume: 21; Issue: 1 Linguagem: Inglês

10.1097/00004347-200201000-00001

1538-7151

Richard J. Zaino,

Garlic and Onion Studies

Distinguishing endocervical adenocarcinoma from endometrial adenocarcinoma might be compared to distinguishing apples from oranges. Both of the latter are fruits of similar size, both are sweet, but their distinction is too easily made by casual inspection to resemble the clinical situation. For this problem, perhaps a closer analogy would be the distinction of one apple variety from another, such as Macintosh from Winesap. Texture, glossiness of the skin, and subtle shape differences provide points of distinction. But these two tumors are even more closely related. Perhaps this problem is more akin to distinguishing Macintosh apples from one orchard from Macintosh apples grown in another orchard. Potentially the same genetic composition but arising in a different environment. Still not quite right. To form an accurate comparison, we would need two apples in which expression of a host of features varies from one region to another in the same fruit. Well, such analogies break down upon too close an analysis, but the issue remains to confront us. Let us begin with a common situation for a surgical pathologist in a busy community practice. An endometrial biopsy and an endocervical curettage have been obtained from a perimenopausal woman with a history of abnormal bleeding per vagina. The diagnosis of endometrioid or mucinous adenocarcinoma is immediately evident upon microscopic examination of both specimens, but there is no associated in situ adenocarcinoma of the cervix, or atypical hyperplasia in the endometrium. Where has this tumor arisen, in the endometrium or the endocervix? Does it matter? If so, are there histologic, immunohistologic, or molecular means to distinguish between these two sites of origin? Should additional samples be requested? Two papers in this issue of the Journal clearly address this vexing and common problem of distinguishing endocervical from endometrial adenocarcinoma. First, it may be helpful to consider whether this distinction is significant. It is. The choice of definitive therapy differs depending on the site of origin. Radiation therapy or radical hysterectomy and pelvic lymphadenectomy are appropriate modalities for the treatment of localized endocervical adenocarcinoma, while surgical staging including simple hysterectomy and para-aortic lymph node sampling are generally indicated for endometrial adenocarcinoma. Consequently, it is important that the gynecologist be aware of the site of origin based on the initial biopsy or curettage sample. However, the problem persists even after the uterus is removed in a minority of cases, when the tumor is centered in the lower uterine segment or upper endocervical canal. But the need to distinguish the site of origin also persists because the chemotherapy of advanced cervical carcinoma differs from that used for advanced endometrial adenocarcinoma, and hormonal therapy is limited to treatment of endometrial, not endocervical, adenocarcinoma. Second, is the histology of endometrial adenocarcinoma distinctive from that of endocervical adenocarcinoma? Unfortunately, it is not. The histology displays significant overlap. Most endometrial adenocarcinomas are of endometrioid type, but so are a significant minority (or a majority in my institution) of endocervical adenocarcinomas. Mucin production may be found in tumors arising in either site. The presence of a precursor lesion such as endometrial hyperplasia or adenocarcinoma in situ may provide support for an origin in the endometrium or endocervix but often is absent. That leads us to consider whether immunohistochemistry can provide additional information, and this is the essence of the two articles in this issue of the Journal. Castrillon et al. (1) studied 29 cases of endocervical adenocarcinoma and 30 cases of endometrial adenocarcinoma. They intentionally selected cases with a high potential for confusion, with about equal numbers of endometrioid and mucinous types in each site. Immunohistochemistry was performed on blocks from the hysterectomy specimens, with antibodies directed against keratins CK7 and 20, epithelial membrane antigen (EMA), vimentin (VIM), and monoclonal carcinoembryonic antigen (CEA). Results were considered positive if staining was present in more than 10% of neoplastic cells, focally positive if staining was present in 1% to 10% of neoplastic cells, and rare or negative if fewer than 1% of cells localized chromogen. Because almost all tumors, irrespective of site of origin, stained with CK7 and EMA, and almost all were negative for CK20, these antibodies were not considered useful. In contrast, VIM was positive in a characteristic lateral cell border (or perinuclear) pattern in all endometrial adenocarcinomas, but in only 7% (2 of 29) of the endocervical adenocarcinomas. CEA was positive in only 27% of endometrial adenocarcinomas, but was positive in 62% of endocervical adenocarcinomas. They found the latter distinction to be particularly helpful for the tumors of endometrioid cell type for which only 14% of endometrial primaries stained, while 67% of those arising in the endocervix stained. The authors concluded that diffuse VIM positivity favors an endometrial origin, although negativity must be interpreted with caution. CEA may be diagnostically helpful in combination with VIM. McCluggage et al. (2) conducted a study of similar size and purpose, but with some notable differences in design. They also included 30 cases of primary endometrial adenocarcinoma, but all were of endometrioid cell type, and 26 cases of endocervical adenocarcinoma, all of which were of typical endocervical type with mucin production. Their antibody panel included CEA, VIM, estrogen receptor (ER), and the high molecular weight cytokeratin, 34βE12. They also found that the cytokeratin they employed, 34βE12, stained almost all of the tumors in either a focal or diffuse fashion irrespective of origin, and concluded that it was unhelpful. There was diffuse cytoplasmic staining for VIM in 97% of the endometrial adenocarcinomas, but in only 8% of the endocervical adenocarcinomas. CEA stained the cytoplasm and membranes of 96% of endocervical adenocarcinomas, and 70% of endometrial adenocarcinomas. Strong intranuclear localization for ER was identified in 93% of endometrial adenocarcinomas, although focal weak localization was noted in 38% of endocervical adenocarcinomas. The authors of this study suggested that a limited panel including VIM, CEA, and ER could help distinguish endocervical from endometrial adenocarcinomas. These papers do not stand alone, but do complement and extend the work of prior investigators. In the early 1980's, Cohen et al. (3), using a polyclonal CEA, found all endocervical adenocarcinomas and about one half of endometrial adenocarcinomas positive, the latter with a generally lower intensity of stain sometimes confined to the luminal aspect of the cells. A few years later, using an indirect immunoperoxidase technique, Maes et al. (4) reported that CEA did not discriminate between endocervical and endometrial adenocarcinomas, staining about one half of the tumors from each site. More recently, Dabbs et al. (5) found immunolocalization of CEA in 65% to 95% of endocervical adenocarcinomas, depending on the antibody used, but no staining in endometrial adenocarcinomas. These divergent results highlight some of the limitations of reliance upon CEA alone, because the methods and antibodies employed dramatically affect the frequency of a positive reaction. Most investigations have found a positive reaction for VIM to be effective in distinguishing between an origin in endometrium or endocervix, with localization described as either perinuclear or along the lateral cell border in 50% to 81% of endometrial adenocarcinomas but in <13% of endocervical adenocarcinomas (5–7). In a recent abstract, Staebler et al. (8) reported that all 23 (grade 1 and 2) endometrial adenocarcinomas were ER positive, and all 14 endocervical adenocarcinomas were ER negative. Other investigators have found that about 70% of endometrial adenocarcinomas are ER positive, with a higher percentage in well differentiated tumors, although only about 10% to 20% of endocervical adenocarcinomas contain ER (9–11). The authors also performed in situ hybridization for human papilloma virus (HPV) DNA in each of the cases. HPV was detected in 4 of 9 endocervical adenocarcinoma, but in none of the 13 endometrial carcinomas examined. Hording et al. (12) also found that HPV DNA was present in 70% of endocervical adenocarcinomas, but in no endometrial adenocarcinomas. In summary, based on these two clear articles and previous published data, there is good evidence that a limited immunohistochemical panel including VIM, ER, and CEA can resolve the site of origin for the majority of these problematic cases. The tumor that is ER positive, VIM positive, and CEA negative is almost certainly of endometrial origin, although an endocervical source is very likely for the tumor that is ER negative, VIM negative, and CEA positive. When the pattern of the panel of results is less definitive, then the detection of HPV by ISH might be considered, if available, because its presence provides strong evidence of an endocervical origin. Not only can this panel be used in the problematic biopsy or curettage specimen, it also may help to assign the correct origin to those tumors found at hysterectomy in the uterine isthmus. Finally, it is equally important to recognize that none of these studies are needed in the majority of cases in which the clinical evidence, history, or associated pathologic findings in the endometrial or endocervical specimen provide compelling data about the site of origin for the tumor. So, let us now celebrate the work of these investigations in gynecologic pathology and our ability to separate one apple from another, and enjoy the fruits of this day.