The Pathology of Pelvic-Ovarian Epithelial (Epithelial-Stromal) Tumors

Molecular Pathogenesis

In recent years the molecular footprints of epithelial ovarian tumors have been increasingly refined, permitting investigators to uncover pathways and biomarkers that distinguish the individual groups and provide both pathogenetic information as well as potential diagnostic and therapeutic targets (see Fig. 25.1 ). They will be referred to under each tumor type within the chapter as appropriate.

Borderline (Atypical Proliferative) Serous Tumors

SBTs are bilateral in about 25% of cases. They are usually partly cystic with watery or mucinous cyst fluid and contain intracystic or surface soft white to tan cauliflower-like papillary projections ( Fig. 25.7 ). The papillae are usually more extensive than the firmer, smooth, bosselated papillae of benign tumors. However, one cannot reliably predict the histologic findings on the basis of the gross appearance, and frozen section examination of either type of projection is warranted.

The following characterize SBTs:

• About 70% of patients with SBTs have no evidence of spread beyond the ovary at the time of initial staging; the disease-free survival in these patients is 98.2%.

• The remaining 30% of patients have pelvic or abdominal peritoneal implants upon initial staging. These account for the great majority who suffer serious morbidity or death.

• However, in marked contrast to serous carcinomas, only 15% to 30% of these higher-stage tumors recur; even when they do, many of them still do not cause harm. Thus, since the late 1970s, the primary focus of research on SBTs has been to discover predictors of a poor outcome in these higher-stage cases.

• Studies have examined the prognostic significance of the histopathologic features of the primary tumor, the morphology of the implants, lymph node metastases, minor foci of invasion in the ovarian tumor, flow cytometry, and p53 expression in the tumor cells. These are discussed in greater detail later.

• The main finding that predicts outcome in higher-stage SBTs is the type of extraovarian peritoneal implants: whether they are “invasive” or “noninvasive.” The 7-year disease-free survival in patients with noninvasive implants is 95.3%; with invasive implants it is 66%.

SBTs are characterized by broad, branching papillae focally covered with highly stratified epithelial cells, some of which appear to be free floating in the extracellular space ( Figs. 25.8 and 25.9 ). This space is usually clear, but some tumors may elaborate mucin ( Fig. 25.10 ). The epithelial cells consist of a mixture of polygonal or columnar and exhibit variable potential for ciliated differentiation. The latter may manifest as abundant cilia or as eosinophilic cytoplasm that is sometimes ciliated or has surface blebs ( Fig. 25.11 ; see Fig. 25.9 ). In some cases, the cytoplasm is abundant, especially in tumors from pregnant women ( Fig. 25.12 ). Nuclear atypia varies among and within cases from mild to moderate, but it is generally less severe than that of LGSCs (see Figs. 25.9, 25.11, and 25.12B ). Mitotic figures are infrequent. The stroma is fibrous and variably cellular; it may be edematous, particularly in the papillary cores. There may be few or many psammoma bodies. Benign-appearing areas, either cystadenomatous or adenofibromatous, are often present and may account for the majority of the tumor.

Borderline tumors are distinguished from low-grade carcinomas by the absence of both micropapillary or cribriform (noninvasive low-grade carcinomas) growth and frank stromal invasion. In addition, the reader will note that (noninvasive) LGSCs often display a more irregular branching pattern, with large blunt papillae terminating abruptly into micropapillary growth patterns. Because of the oblique sectioning of complex papillae, there are areas where epithelial cells are included within the stroma. This appearance should not be mistaken for invasion, because there is no stromal reaction and one can easily conceptualize the included areas as secondary to the plane of section. It must be acknowledged that, aside from tumors that have the micropapillary/cribriform patterns described later, there are occasional low-grade serous tumors that lack obvious stromal invasion but have a particularly exuberant epithelial growth pattern ( Fig. 25.13 ). Experienced pathologists may differ on whether to interpret these as borderline tumors or as LGSCs with “expansile growth-type invasion.”

Having diagnosed an SBT, the pathologist may encounter five other features, the first four of which do not alter management but may slightly increase the overall risk of recurrence and in some cases, adverse outcome:

• Surface involvement ( Fig. 25.14A ): Exclusion of surface involvement depends on a careful gross examination and orientation such that the surface is evident on the slide, usually by its being inked prior to cutting into the tumor. These have a “stuck on” appearance and are identical to extraovarian noninvasive desmoplastic implants. These “auto implants” are associated with an increased frequency of high-stage tumors, but they do not alter the prognosis in stage I and are not currently considered to represent invasion. The presence or absence of surface involvement should be included in the final report.

  

• Stromal microinvasion: Most SBTs are homogeneous. However, generous and judicious sampling is required to exclude focal areas of invasion. In most cases, this requires at least one section per centimeter of tumor diameter. Invasive foci measuring less than 10 mm ² and less than 3 mm in greatest dimension are present in as many as 10% of cases. These should be reported as foci of microinvasion. Microinvasive areas consist either of dyshesive eosinophilic cells in the cores of the papillae (see Fig. 25.14B ) or small aggregates of cells with a cribriform or microcystic pattern, usually with a reactive-appearing stroma (see Fig. 25.14C ). Microinvasion does not by itself appear to alter the overall behavior of a borderline tumor but will co-segregate with more aggressive tumors (LGSCs), in which case certain patterns, such as micropapillae, may be more ominous. Vascular space invasion has also been observed in a few cases with microinvasion, but thus far it has not been demonstrated to worsen the prognosis. A borderline tumor with invasive areas that are larger than microinvasive areas should be reported as LGSCs associated with an SBT. Patients with these tumors often—but not invariably—have invasive extraovarian implants and a high mortality, albeit sometimes only after many years.

• Lymph node involvement: Lymph nodes are not examined in many cases, and whether nodal involvement by SBTs carries significant weight in terms of overall outcome has not been well established. Pelvic and paraortic lymph node involvement has been found in as many as 25% of cases in which nodes are examined. The tumor is generally present in sinusoidal spaces rather than in the substance of the node, distinguishing them from müllerian rests (endosalpingiosis) that may also be present ( Fig. 25.15 ). The presence of lymph node involvement has not been shown to alter the prognosis of SBTs. However, there may be a relationship to lower disease-free survival.

  

• Noninvasive implants on the peritoneal surfaces: Sometimes the surface of the tumor or the lining of a cyst within the tumor may have plaques of reactive-appearing fibrous stroma containing psammoma bodies and epithelial cells lying singly or in small groups. They are illustrated later. Noninvasive implants on the peritoneal surfaces impart a slightly greater risk of an adverse outcome than when absent, but the differences are small. This will be discussed further later.

• Salpingoliths: The fallopian tubes in patients with borderline tumors may contain concentric calcifications trapped within the lamina propria, termed salpingoliths ( Fig. 25.16 ). When detected in tubal specimens, these concentric calcifications should raise the possibility of a coexisting borderline serous tumor in the ovary or pelvic surfaces.

  

In all, the prognostic impact of the preceding findings is low (see Table 25.1 ). Studies of stromal microinvasion and lymph node spread in borderline tumors have shown little impact on 5- to 6-year survival, although some feel that disease-free survival is reduced with lymph node involvement. Recurrence rates for noninvasive implants are less than 10%. Surface involvement is significant insofar as it enables peritoneal spread, the outcome ultimately being a function of the biology of tumor.

Borderline Serous Tumors With Micropapillary/Cribriform Architecture (Noninvasive Low-Grade Serous Carcinomas)

Women with SBTs that contain extensive areas of micropapillary or cribriform epithelial overgrowth have an increased frequency of bilateral tumors and invasive peritoneal implants, with the attendant risk of the latter for subsequent recurrence and death (discussed later). In the micropapillary pattern, epithelial cells surround a prominent, nonbranching fibrous stalk and protrude radially as long, thin micropapillae without stromal cores such that the low-power appearance resembles a Medusa head ( Fig. 25.17A and B ). When the micropapillae merge, they take on a ramifying appearance; alternatively the pattern can take on a cribriform pattern (see Fig. 25.17C ). The epithelial cells in these foci generally differ from those in other areas or in ordinary borderline tumors in that the cells are monomorphous and “clonal appearing” with scant cytoplasm (see Fig. 25.17D ). They are mostly cuboidal but may also have a hobnail, columnar, or flattened appearance. Cilia may be retained, particularly in the areas of cribriform growth (see Fig. 25.17C ). Nuclei are round and slightly more hyperchromatic than in the usual borderline tumor.

Ample data exist to consider the micropapillary pattern synonymous with noninvasive LGSCs. In a study examining 400 borderline tumors and well-differentiated carcinomas, 17 borderline tumors had micropapillary/cribriform foci at least 5 mm in extent. Of the 13 patients with follow-up data in this group, three had no implants and none recurred; three had noninvasive implants and one had a recurrence; six had invasive implants (defined so as to include micropapillary in the implant itself as invasive), four of these patients had a recurrence, and two died of their tumors.

In a companion study of 65 advanced-stage borderline tumors, 11 ovarian tumors had micropapillary/cribriform areas, 10 of which had invasive implants. Seven of the 10 patients had recurrences, and four died of their tumors. The frequency of invasive implants and recurrences in this group was much greater than in the remaining cases. Because of these findings and subsequent correlative molecular analyses, these authors concluded that SBTs with micropapillary/cribriform areas greater than 5 mm belong in a separate category that they designated “micropapillary serous carcinoma.”

Four subsequent studies of tumors with micropapillary/cribriform areas totaling 59 cases recorded noninvasive implants in 28 cases, three of which recurred, and invasive implants in five cases, all of which recurred. There were no recurrences in the 26 cases without implants. Another study of 99 advanced-stage SBTs contained 18 cases with micropapillary/cribriform areas. Only three of these 18 patients had invasive implants, but 14 of them had disease progression or recurrence, and five of them died. This compared with 25 recurrences and 12 deaths in 81 patients without micropapillary/cribriform areas in their ovarian tumors.

Taken together, these observations indicate that patients with micropapillary/cribriform foci in an SBT have a higher risk than those with ordinary SBTs for both noninvasive and invasive implants and thus for subsequent recurrence. The question that has persisted is whether micropapillary tumors without implants imposed a greater risk of recurrence. The earlier studies did not provide consistent outcome information; however, it is important to note a study by Hannibal et al. surveying registry data from Denmark between 1978 and 2002. Their study encompassed 946 women with APST and 80 with noninvasive LGSC. Given the extent of the population studied, several findings in their study merit notation:

• They found that the 5-year and 20-year risk of metastatic serous cancer (mostly LGSC) outcome following a micropapillary borderline (or noninvasive LGSC) was several fold higher than that for SBT/APST (5.0% and 13.9% vs. 0.9% and 3.7%). Thus, although spread from the ovary or invasive implants is still the most powerful predictor of adverse outcome, the risk of adverse outcome for stage I tumors is still elevated relative to ordinary borderline tumors (see Fig. 25.17E and F ).

• The recurrence rate for all SBTs/APSTs was only 5.3%, increasing steadily up to 10 years following surgery and then leveling off.

• Although an LGSC outcome was infrequent for APSTs, it was nonetheless significantly greater than the expected rate of serous cancer in the population (34 for APSTs vs. 2.5 in the general population). Moreover, the risk of metastatic LGSC outcome for noninvasive LGSC was significantly higher than for APSTs with a hazard ratio (HR) of 4.3. Factors contributing to LGSC outcome risk for women with APSTs were surface involvement (HR = 4.8), residual disease following surgery (HR = 3.8), bilaterality (HR = 7.7), noninvasive implants (HR = 9.3), and, most telling, invasive implants (HR = 115.9). Less information is available on risk factors for adverse outcome following noninvasive LGSC given the small numbers, but nonsignificant risks were linked to advanced stage (HR = 2.6), surface involvement (HR = 2.2), and invasive implants (HR = 4.5).

For the purposes of management, our strategy is to adhere to the spirit of the WHO classification and classify these tumors as SBTs with complex or micropapillary architecture, noting their alternate designation of noninvasive LGSC, the latter important to convey their biologic potential. Both terms are included in Table 25.2 . Naturally, cases with micropapillary/cribriform areas must be sampled and sectioned thoroughly to exclude invasion within the ovarian tumor, and we extensively sample the omentum. Moreover, we append a note to the diagnostic report indicating the presence of micropapillary/cribriform foci and the fact that these foci slightly increase the risk of an LGSC outcome, as recently described.

It is important to stress that micropapillary-like mimics can be seen in not only noninvasive LGSCs but also ordinary borderline tumors ( Fig. 25.18A ), frank low serous carcinomas (see Fig. 25.18B ), and HGSCs (see Fig. 25.18C ). Thus, attention must be paid to the histologic criteria, the surrounding tumor, and the cytologic features of the tumor epithelium.

Implants Versus Benign Mimics

Implants must be distinguished from florid mesothe­lial hyperplastic reactions, which may form glandlike structures, contain psammoma bodies, and be associated with a fibrous reaction, similar to true implants. Mesothelial cell nuclei are paler than epithelial cell nuclei, and, when lining a surface, hyperplastic mesothelial cells often have a hobnail arrangement with spaces between them ( Fig. 25.19 ). Although usually not necessary, a positive calretinin stain will confirm their mesothelial nature.

Glandular, cystic, or small papillary inclusions, sometimes with psammoma bodies, are not uncommonly present on peritoneal surfaces, in lymph nodes, and in the omentum ( Fig. 25.20 ) in women with or without borderline tumors. Although some can question the benign nature of these foci in lymph nodes, as well as their link to the borderline tumor when present in the omentum, their presence has not been shown to alter the prognosis of an SBT, and they should not be diagnosed as implants or metastases. True implants have a more exuberant and complex epithelial cell proliferation than do these foci of endosalpingiosis.

Peritoneal Cytology

It is important to reiterate here that the status of the peritoneal cytology will not determine the nature of the peritoneal involvement. Low-grade serous tumors, be they SBT or LGSC, may not be distinguishable on cytology alone. In general, we refer to a positive peritoneal cytology in such cases as “positive for low grade serous neoplasia” ( Fig. 25.21 ) and leave the distinction of the nature of the tumor up to histology.

Invasive Versus Noninvasive Implants

Implants may or may not invade pelvic or abdominal structures. On the basis of this distinction, implants are divided into invasive and noninvasive types. These may coexist. Criteria to distinguish the implant types are given here and summarized in Table 25.3 .

Table 25.3

Invasive Versus Noninvasive Implants of Serous Borderline Tumors

Noninvasive a		Invasive
Epithelial	Desmoplastic	Accepted Criteria b	Proposed Additional Qualifying Criteria c
Papillary epithelium similar to that of the usual borderline tumor in submesothelial invaginations or between lobules of adipose tissue	Papillae, glands, cell clusters, or single cells within inflamed, reactive-appearing fibrous tissue that appears to be plastered on to serosal surfaces or in fibrous septae of adipose tissue	Proliferative epithelium resembling low-grade serous carcinoma (LGSC) invading abdominal/pelvic organs or omental fat, usually with a desmoplastic reaction and an infiltrative margin	Micropapillary architecture whether or not there is obvious invasion, or solid epithelial nests with surrounding clefts in a fibrous stroma

a If underlying tissue is not present in the biopsy, the implant is considered noninvasive.

b Konstantinopoulos PA, Matulonis UA: Current status and evolution of preclinical drug development models of epithelial ovarian cancer. Front Oncol 3:296, 2013.

c Bell KA, Smith Sehdev AE, Kurman RJ: Refined diagnostic criteria for implants associated with ovarian atypical proliferative serous tumors (borderline) and micropapillary serous carcinomas. Am J Surg Pathol 25(4):419-432, 2001.

Noninvasive implants consist of two subtypes—epithelial and desmoplastic. Epithelial implants are composed of papillae resembling those within a borderline tumor. They are found either in mesothelial-lined invaginations of a peritoneal surface or between lobules of fat ( Fig. 25.22A and B ). Desmoplastic noninvasive implants consist of single cells, small epithelial islands, papillae, or cribriform glands encased in stroma, usually with inflammatory cells and psammoma bodies. The entire implant appears well circumscribed and has a “plastered on” rather than invasive infiltrative appearance (see Fig. 25.22C ). Early in their evolution, desmoplastic noninvasive implants may appear necrotic and inflammatory.

Invasive implants are composed of cells and cell patterns similar to those of well-differentiated serous carcinomas. They are almost always associated with stromal desmoplasia but, in addition, have an irregular, infiltrative-appearing border. They appear to invade underlying structures or replace fatty tissue rather than sitting indolently on the peritoneal surface or between fat lobules ( Fig. 25.23 ).

If underlying tissue is absent in a biopsy, the lesion is classified as noninvasive by default. However, because this is a common scenario, one reported study of implant morphology classified either a micropapillary pattern alone or solid epithelial nests surrounded by clear spaces or clefts in a fibrous stroma ( Fig. 25.24 ) as invasive in the absence of an obvious infiltrative pattern. Using their expanded criteria, these authors found that of 60 patients with implants, 31 had the invasive type. Nineteen of these 31 patients had progressive disease, and six of them died. Of the 29 remaining patients, three had progression and two died. A caveat is that 25 of the 60 patients (45%) had micropapillary/cribriform areas in their primary tumor. This is a higher frequency than was found in two other studies of high-stage borderline tumors (where micropapillary/cribriform foci were present in only 21% of 146 ovarian tumors), suggesting that case selection may have skewed the data toward the finding of micropapillary foci in the implants. Nevertheless, these results suggest that an expanded definition of invasive implants is warranted.

Invasive Low-Grade Serous Carcinomas

On the basis of their nuclear features, almost all serous carcinomas can be divided into low-grade and high-grade types (grade 1 or grade 3), with the latter constituting the great majority of cases. LGSCs are often associated with areas that merge morphologically with SBTs ( Fig. 25.27A ); they typically combine the features of both an extensive papillary architecture (often with many psammoma bodies) and evidence of stromal invasion (see Fig. 25.27B ). The latter may consist of infiltrating gland-like structures or more commonly, multiple papillae projecting into a space or embedded in a fibrous stroma that may also contain glands, cysts, irregular islands of tumor cells, or cribriform glands (see Fig. 25.27B ). Often the invasive papillae are surrounded by a clear space, possibly caused by fluid secreted by tumor cells (see Fig. 25.27C ) and occasionally demonstrate a “macropapillary” invasive architecture ( Fig. 25.27D ). Nuclei are uniform, round, or oval with evenly distributed chromatin, with or without a prominent nucleolus ( Fig. 25.28 ). At this point in the spectrum of low-grade serous tumors, cilia are inconspicuous. The mitotic index is usually less than 10 per 10 high-power fields. In some cases, a conspicuous micropapillary or cribriform component may be present ( Fig. 25.29 ; see Fig. 25.27A ) with underlying invasion.

Peritoneal invasion by LGSC is an extension of the spectrum of invasive implants in borderline tumors. In contrast to the latter, the abundance of metastatic deposits and its nuclear morphology are more in keeping with a frankly malignant process. This can take the form of numerous nests of invasive tumor, more loosely arranged clusters of papillae ( Fig. 25.30A ), the latter of which might be delicate and might mimic a noninvasive implant (see Fig. 25.30B ). A rare variant of low-grade carcinoma and one that appears to predominate, if not originate in, the peritoneum is the “psammocarcinoma.” This tumor contains a marked preponderance of psammoma bodies to the point where the epithelial cells may be overlooked (see Fig. 25.30C ). Peritoneal metastases from either of these low-grade malignancies can focally mimic adhesions or noninvasive implants from a borderline tumor, but they will invariably demonstrate invasion into underlying tissues that equals or exceeds that seen in invasive implants described previously.

Nodal metastases can occur in LGSCs. They are typically more confluent and actively replace the nodal parenchyma (see Fig. 25.30D ).

Low- Versus High-Grade Serous Carcinoma

The pathologist who is contemplating the diagnosis of an HGSC of the ovary typically must make two distinctions: (1) whether the tumor is high- or low-grade serous and (2) whether it is an endometrioid carcinoma. As mentioned earlier, the separation of LGSC from HGSC might be difficult for two reasons. First, some tumors fall into a category of intermediate grade. Second, some LGSCs have areas virtually indistinguishable from HGSC.

LGSCs should not display accumulation of nuclear p53 protein, because they arise independent of perturbations in this pathway. Thus, they typically display focal or patchy staining ( Fig. 25.31A ). In fact, the presence of some staining, as will be discussed later, is helpful in that it implies the presence of an intact TP53 gene. Tumors displaying greater degrees of nuclear atypia might still be considered low-grade carcinomas (see Fig. 25.31B ). However, if they contain strong (or completely absent) staining for p53, they are best placed in the HGSC category (see Fig. 25.31C ). In essence, designation of grade 2, although justifiable for tumors that fall into a morphologic gray zone, has little practical value. Using p53 immunohistochemistry, the pathologists must attempt to determine the tumor grade for the purposes of prognosis and management.

High-Grade Serous Carcinomas

HGSCs display a range of morphologic patterns including (1) classic HGSC and (2) SET patterns. The two are distinguished primarily based on growth pattern and the fundamental, consistent feature is enlarged nuclei, usually with prominent nucleoli and many mitoses, which are often abnormal.

• Classic HGSC is immediately recognized as “serous carcinoma.” The range of pleomorphism is broad and may include bizarre cells ( Figs. 25.32 through 25.34 ). The architectural features are more variable, including large papillae (see Fig. 25.31B ); smaller papillae may be frequent or sparse, small, or large ( Fig. 25.35A ) and micropapillary architecture devoid of supporting stroma (see Fig. 25.35B ). Foci showing neither distinct papillary nor glandular architecture with slit-forming fenestrations are common ( Fig. 25.36 ). The stroma is fibrous, desmoplastic, edematous, or myxoid. Psammoma bodies may be plentiful, infrequent, or absent. There may be extensive necrosis. Additional patterns that are distinctly “nonserous” in their appearance include foci of squamous differentiation, osseous metaplasia, intracytoplasmic lumina with mucinous inclusions, a microcystic appearance ( Fig. 25.37A ), intracellular or extracellular hyaline globules, multinucleated cells sometimes resembling syncytiotrophoblasts (often beta–human chorionic gonadotropin [β-hCG] positive), an adenoid cystic–like pattern, a yolk sac tumor–like reticular pattern, and mural nodules of sarcoma or anaplastic carcinoma.

  
  
  
  
  
  

• SET (solid, endometrioid-like, and transitional) patterns of HGSC are the group that are not always readily classified as HGSC, because they lack the striking features attributed to the classic group, mainly they show a predominance of more solid growth with less striking atypia. In addition to their morphologic features, SET variants have the following attributes: (1) they have been described somewhat more frequently in association with BRCA1 inactivation or germ-line mutations, (2) they have been observed to have a higher rate of chemoresponsiveness, and (3) they a significantly lower frequency of association with STIC than the classic type.

SET patterns include the following: (1) solid resembling grade 3 endometrioid, with the exception of somewhat higher nuclear grade; (2) pseudoendometrioid, with well-defined glands with partial preservation of pseudostratified epithelium; and (3) transitional, essentially a solid pattern with stromal cores lending a resemblance to a urothelial tumor yet retaining a müllerian immunophenotype (PAX8, WT1, CK7, p53 positive, and CK20 negative) (see Fig. 25.37B and C ). Another pattern that might be encountered is “undifferentiated.” Typical serous morphology may not be seen and the diagnosis may require immunostaining. In such cases, we simply apply the term “high-grade müllerian adenocarcinoma.”

Determining the Origin of High-Grade Pelvic Serous Carcinoma

Potential origins for HGSC are detailed earlier in this chapter, and there are a number of recent publications promoting algorithms to ascertain the probable origin of HGSCs. To summarize what is known:

• The only epithelium that contains a serous carcinogenic sequence that can be agreed on is the distal fallopian tube and endometrial lining, the former being the putative source in a very high percentage of incidentally discovered extrauterine HGSCs in asymptomatic women.

• An STIC cannot be found in approximately 60% of symptomatic women with HGSC.

• Some STICs are very likely intramucosal metastases rather than primary lesions.

• Other sources, including ovarian adenofibromas and endometriomas, as well as endosalpingiosis and pelvic ovarian surface epithelium, are under consideration but have yet to yield a definable carcinogenic sequence. We do recommend noting the presence of any potential precursor sequence, including STIC ( Fig. 25.38 ). For management purposes and staging, STIC is not a critical component unless found in isolation, and we do not resort to the more traditional approaches to sorting out origin at this time.

  

In our diagnostic practice, we adhere to the following guide:

• 1.

  Stage I tumors involving the ovary or fallopian tube are classified as such and assigned to the organ of origin.

• 2.

  Higher stage tumors are simply classified as HGSC, irrespective of whether they are on the ovary, mesosalpinx, pelvic peritoneum, or endosalpinx.

• 3.

  If STIC or a serous tubal epithelial proliferation/lesion is identified, it is noted in the report.

• 4.

  If the endosalpinx is involved by tumor, it is noted in the report.

• 5.

  Any associated ovarian or extraovarian features, including borderline or lower grade serous tumors, adenofibromas, endometriosis, or benign mullerian remnants (endosalpingiosis), are noted in the report.

• 6.

  The tumor is staged according to the FIGO guidelines.

This approach provides for the acquisition of data that might be relevant to site of origin, should this become an important management issue in the future. At this time, staging does not require assigning a specific site of origin; however, many advocate assigning the site of origin to the fallopian tube, ovary, or peritoneum.

Benign Serous Tumors

The principal issue with benign serous tumors is their distinction from non-neoplastic cysts or surface epithelial changes . Serous surface papillary adenofibromas should be at least 1 cm in greatest extent to be considered neoplastic. Small, knobby, fibrous protrusions on the surface of the ovary are common. Sometimes they are submitted for frozen section with the clinical suspicion of an implant or small carcinoma.

The rare rete cystadenoma resembles a serous cystadenoma. However, it is located in the ovarian hilus and lined by nonciliated cuboidal, columnar, or flattened epithelial cells, often forming an undulating or creviced appearance. There is smooth muscle in the wall, and hilus cells are often present.

The epithelium of an endometrioid adenofibroma can resemble serous epithelium except for the lack of cilia. The distinction in some cases is arbitrary. When uncertain, we simply classify these as benign cystadenofibromas.

Borderline Tumors

Many borderline epithelial tumors of the ovary consist of both ciliated and secretory cells, or both and endometrioid differentiation, and it can be argued that the classification of a given tumor will depend on the direction in which the differentiation pathway is tilting. Endocervical-like (müllerian) mucinous and mixed epithelial (mucinous, ciliated, and endometrioid) borderline tumors, as well as lower grade mixed endometrioid and serous carcinomas, typify this spectrum and are classified accordingly. This group may be associated with endometriosis or hemorrhagic cysts, suggesting at least a form of endometriosis.

The retiform variant of Sertoli-Leydig cell tumor is papillary and may simulate an SBT or LGSC ( Fig. 25.39 ). However, it contains other features of a Sertoli-Leydig cell tumor, is usually found in teenagers and young women, and may be androgenic.

Serous Carcinomas

Because of overlapping features, serous carcinomas are most commonly confused with other epithelial-stromal tumors, either of the endometrioid, clear cell, or transitional cell type. Less frequently, other papillary neoplasms, such as the retiform Sertoli-Leydig cell tumor (discussed earlier), mesothelioma (discussed later), metastatic serous carcinomas from elsewhere in the female genital tract, or papillary carcinomas from distant sites, such as the breast, can mimic ovarian serous carcinoma.

Unlike serous carcinomas, low-grade endometrioid carcinomas may have squamous foci or aggregates of low-grade spindle cells without obvious keratinization blending with glandlike spaces. Psammoma bodies are rare in endometrioid carcinomas, and endometrioid papillary areas are usually villous rather than broad based and complex as in serous carcinomas. Endometriosis may accompany endometrioid tumors but is only rarely demonstrated with serous tumors. Immunostaining for the Wilms tumor gene, WT-1 , may be of assistance but in lower grade tumors is less likely to be needed. There is strong WT-1 nuclear staining in most serous carcinomas, but most endometrioid carcinomas stain only weakly or not at all. An equally promising biomarker is p16, which is diffusely positive in HGSCs and patchy in distribution in low-grade serous tumors and low-grade endometrioid carcinomas. However, it will not distinguish between high-grade serous and high-grade endometrioid carcinomas.

Clear cell carcinomas with papillary architecture can resemble serous carcinomas. However, most of the epithelial cells have either clear or oxyphilic cytoplasm and a characteristic nucleus with an angular periphery and a prominent nucleolus. The papillary cores are often hyalinized in clear cell carcinomas, and there are almost always other characteristic patterns. As with endometrioid carcinomas, residual endometriosis may be present. Immunohistochemistry may aid in this distinction, as clear cell carcinomas uncommonly demonstrate aberrant p53 staining, and are often positive for HNF1-B, Napsin A, and AMACR. Additionally, WT-1 is typically strongly positive in the nuclei of HGSC, whereas it is generally negative in clear cell carcinoma.

Transitional cell carcinomas associated with Brenner tumors or endometrioid adenocarcinomas conceivably might be confused with the transitional (SET) pattern of HGSC. The latter display greater nuclear atypia and will be p53 positive, and the associated benign Brenner tumor will not be seen.

Malignant mesotheliomas may sometimes be primarily or exclusively confined to the ovaries ( Fig. 25.40 ). They are often both papillary and solid and may simulate either an SBT or a carcinoma. The papillae of mesotheliomas have hyaline cores and are less complex than those of serous tumors. Although there are exceptions, mesothelioma cells are usually less stratified and pleomorphic than those of serous carcinomas. They have a round to oval nucleus with fine chromatin, a small nucleolus, and densely eosinophilic cytoplasm. The majority of serous carcinomas mark positively for a variety of epithelial markers, including epithelial membrane antigen (EMA), CAM-5.2, AE-1/AE-3, CK-7, B-72.3, LeuM-1, CA-125, Pax8, and WT-1 (nuclear staining), some of which overlap with mesothelioma. Serous carcinomas are negative for CK-20, calretinin, D2-40, and other mesothelial markers. Additionally, Bap1 loss by immunohistochemistry supports a diagnosis of malignant mesothelioma.

Serous carcinomas of the endometrium may metastasize to the ovaries, bringing up the question of origin when both organs are involved. These cases are best evaluated by assessing the volume of tumor, depth of penetration, and vascular invasion in the uterus compared with the location and extent of the tumor in the ovaries. WT-1 immunostaining may be helpful. It is generally strongly and diffusely positive in primary ovarian serous carcinomas and either negative or only focally positive in uterine serous carcinomas. One study showed that nearly all peritoneal serous carcinomas scored positive for WT-1, whereas only 20% of uterine tumors did. In another study, small carcinomas in a uterine polyp were typically WT-1 negative, even when associated with considerable peritoneal disease. Thus, the percentage of superficial uterine serous carcinomas that can be attributed to a peritoneal primary is considered small. Conversely, with increasing use of the Sectioning and Extensively Examining the FIMbriated End (SEE-FIM) protocol for evaluating tubes in uterine serous carcinomas, it is accepted that associated STICs are often mucosal metastases from the uterus.

Cervical “Papillary Serous Carcinoma,” p53 and p16 Staining

A small subset of cervical adenocarcinomas have been classified as papillary serous carcinomas owing to their close resemblance to their counterparts in the upper genital tract. Like the latter, they are intensely positive for p16, and this marker will not distinguish the two groups. However, TP53 mutations do not characterize the cervical carcinomas, and a p53 stain should be weakly positive and heterogeneous (“wild-type” pattern) ( Fig. 25.41A ). If the stain is either diffuse and intensely positive or completely negative (“mutation-type” pattern), a diagnosis of a metastasis from the upper genital tract should be considered (see Fig. 25.41B and C ). The latter is an important distinction. We have seen examples of HGSCs that have metastasized to the cervix but were devoid of p53 staining because the antigenic targets on the protein had been lost due to mutation, resulting in a null “mutation-type” staining pattern.

Serous Versus Endometrioid Carcinoma

The distinction of serous from endometrioid carcinoma is a common conundrum in the endometrium due to both the spectrum of morphologies that might defy precise separation and some immunophenotypic overlap. p53 staining may help in some cases, and the presence of squamous differentiation will exclude a serous neoplasm. If a case proves impossible to classify, a diagnosis of “high-grade müllerian carcinoma” is permissible. Examples of such endometrioid tumors are illustrated under the section on high-grade endometrioid carcinomas (see Fig. 25.107 , later).

Benign Mucinous Tumors: Cystadenomas, Cystadenofibromas, and Adenofibromas

These three tumors form a spectrum with varying proportions of epithelial-lined cysts, glands, and stroma. The cystadenoma is by far the most common, being composed of one or more thin-walled cysts that are lined by a single layer of mucinous columnar cells with benign-appearing, basally located nuclei ( Fig. 25.49 ). Mitotic figures may be present. There may also be occasional goblet cells and neuroendocrine cells, the latter requiring special stains for visualization. Sometimes small papillae with stromal cores project into the cysts; however, there is no associated epithelial stratification or tufting. The stroma may be fibrous or resemble ovarian stroma. In some instances, it may contain smooth muscle, luteinized cells ( Fig. 25.50 ), or calcifications. Because of cyst rupture and mucin extravasation, large areas of necrosis accompanied by histiocytes and inflammatory cells may be present ( Fig. 25.51 ). Adenofibromas and cystadenofibromas contain a grossly observable solid stromal component. Microscopically, they contain mucinous glands and small cysts uniformly distributed in a fibromatous stroma ( Fig. 25.52 ).

Borderline Mucinous Tumors: Intestinal and Endocervical-Like Types

About 85% of mucinous borderline tumors are of the intestinal type. These in turn make up three subtypes: (1) with atypia only, referred to simply as “borderline,” (2) with intraepithelial carcinoma, and (3) with stromal microinvasion.

Intestinal-Type Mucinous Borderline Tumors With Atypia Only

These differ from cystadenomas and cystadenofibromas in that, in addition to benign-appearing areas with a single layer of mucinous columnar cells with banal, basally located nuclei, other areas are characterized by the presence of epithelial cells that are crowded and stratified and often form either stroma-free papillary tufts or papillae with thin fibrous cores ( Figs. 25.53 and 25.54 ). In these foci, the nuclei are slightly enlarged and hyperchromatic and have increased mitotic activity, most noticeably at the base of the papillae ( Fig. 25.55 ). The epithelium has the morphologic appearance of intestinal epithelium with goblet cells ( Fig. 25.56 ) and sometimes Paneth cells. Neuroendocrine cells are also present but may require special stains for visualization. The gland and cyst lumens contain mucin, sometimes admixed with histiocytes and inflammatory cells. There may be areas of necrosis secondary to mucin extravasation .

Intestinal-Type Mucinous Borderline Tumors With Intraepithelial Carcinoma

If the epithelial cell nuclei appear cytologically malignant (significant nuclear enlargement, nuclear hyperchromasia, often prominent nucleoli) and there is no obvious stromal invasion, the suffix “with intraepithelial carcinoma” should be added to the diagnosis of mucinous borderline tumor ( Fig. 25.57 ). In most cases, there is prominent stratification of the malignant-appearing cells with stroma-free papillae or a cribriform pattern of growth ( Figs. 25.58 through 25.60 ).

Intestinal-Type Mucinous Borderline Tumors With Microinvasion

One or several small foci of stromal invasion, often inciting a desmoplastic response, may be present in a borderline tumor ( Fig. 25.61 ). If the invasive areas are all less than 10 mm ² and less than 3 mm in any linear dimension, the tumor should be interpreted as “borderline with focal microinvasion,” further specifying the number of invasive foci. Areas of mucin extravasation with a fibrous and histiocytic response containing entrapped benign epithelial cells may be misinterpreted as invasion ( Fig. 25.62 ). Generally, true invasion will have more developed desmoplasia, fewer histiocytes, and more obviously malignant epithelial cells arranged in an infiltrative pattern. All reported stage I cases with microinvasion have had a benign clinical behavior.

Endocervical-Like Borderline Tumors

The low-power microscopic appearance, with broad bulbous papillae, associated with cell stratification and epithelial tufting, is similar to that of serous rather than the multicystic appearance of an intestinal-type mucinous borderline tumor ( Fig. 25.63 ). The papillae, however, are lined by both mucinous columnar cells resembling those of the endocervix and stratified polygonal eosinophilic cells ( Figs. 25.64 and 25.65 ). Characteristically, many acute inflammatory cells are present in the epithelial cell cytoplasm and in the extracellular mucin (see Fig. 25.65 ). Adjacent endometriosis may be present, and sometimes the tumor appears to arise from an endometriotic cyst. Nuclei are usually only slightly atypical, but rare cases with areas of intraepithelial carcinoma or microinvasion have been reported; the behavior of these has been benign. Even in the absence of identifiable stromal invasion, approximately 6% of endocervical-like borderline tumors have had either noninvasive or invasive-appearing abdominal and peritoneal implants ( Figs. 25.66 and 25.67 ) and rarely lymph node metastases. These have not, however, led to recurrences.

The comparative features of the various types of mucinous borderline tumors are given in Table 25.1 .

Mucinous Carcinomas

Only 5% to 10% of mucinous carcinomas are entirely malignant; the rest have a benign and borderline component. Within this background, there are two forms of invasion: (1) expansile and (2) infiltrative types, the former being more common.

In expansile (confluent) invasion, glands and cysts lined by malignant-appearing cells have little or no intervening stroma such that they appear to have invaded by expanding into and replacing the stroma across a broad, sharply demarcated front. The invasive areas may be composed of back-to-back glands, ramifying channels lined by epithelial cells, or cystic spaces lined by complex papillae and containing eosinophilic material and nuclear debris ( Figs. 25.68 and 25.69 ). The epithelial cells are columnar or polygonal with eosinophilic or mucinous cytoplasm, sometimes with goblet cells ( Figs. 25.70 and 25.71 ). In some cases, the nuclei are highly stratified, resembling those of endometrioid carcinoma, but in contrast to endometrioid carcinoma, the nuclei are characteristically long and narrow and there is slight residual cytoplasmic mucin ( Fig. 25.72 ). Because small confluent areas are difficult to certify as invasive, expansile invasion should not be considered unless the area exceeds 10 mm ² and is at least 3 mm in each of two linear dimensions.

In infiltrative invasion, glands, sheets, or single epithelial cells haphazardly invade the stroma ( Fig. 25.73 ). The invasive cells may appear endocervical or intestinal but usually have a nondescript eosinophilic cytoplasm. Occasional mucinous carcinomas have colloid carcinoma–like foci ( Fig. 25.74 ). Signet ring cells may be present, but abundant signet ring cells should raise the question of a metastasis. The stroma may be desmoplastic with or without inflammatory cells, or it may be cellular and fibromatous. In the latter instance, an invasive carcinoma may be mistaken for a borderline tumor, especially if the infiltrative glands are large and evenly distributed ( Fig. 25.75 ). However, in carcinomas, the glands are entirely lined by malignant cells, rather than the mixture of benign-, atypical-, and malignant-appearing cells seen in borderline tumors ( Fig. 25.76 ). Tumors with infiltrative invasion are biologically more aggressive and spread beyond the ovary more readily than those with purely expansile invasion.

Another much less common subtype of mucinous carcinoma, the endocervical-like (müllerian) type , may also occur. It has a papillary, cystic, and glandular growth pattern with abundant mucin in the extracellular spaces ( Fig. 25.77 ). Most of the epithelial cells are well-differentiated columnar cells, similar to those of the endocervix ( Fig. 25.78 ). There may be minor components of endometrioid ( Fig. 25.79 ), serous, or squamous cell types. When the serous component is prominent, these tumors have been termed “seromucinous” carcinomas. Endocervical-like mucinous carcinomas may be bilateral and associated with endometriosis. We view these as a malignant counterpart to the broad spectrum of differentiation that permits mixed epithelial patterns encompassing mucinous, endometrioid, and serous tumors. The reader can well appreciate this phenomenon when struggling to precisely type some of these neoplasms.

Mucinous Tumors With Mural Nodules

Ordinary mucinous tumors may contain one or more solid areas in the wall of a cyst. The term mural nodule, however, is reserved for those very uncommon nodules that are widely divergent microscopically from the remainder of the tumor, which is either a borderline tumor or a carcinoma. Grossly, mural nodules may appear uniform or have a variegated yellow, pink, or red appearance; hemorrhage and necrosis may be present. The nodules may be single or multiple and range from less than 1 cm to 30 cm. Mural nodules may be benign or malignant, the latter in the form of anaplastic carcinoma, sarcoma, or carcinosarcoma.

Three varieties of benign mural nodules have been described. In the pleomorphic and epulis-like type, osteoclast-like multinucleated giant cells and extravasated red blood cells are present among pleomorphic spindle cells, similar to an aneurysmal bone cyst or giant cell reparative granuloma of bone ( Fig. 25.80 ). A second variant contains a purely spindle cell population with no or only a few giant cells. The third pattern consists almost entirely of groups of large multinucleated cells with abundant ground-glass cytoplasm separated by thin fibrous bands. Some benign mural nodules have been called “sarcoma like” because the cells may be quite pleomorphic, and there may be numerous mitotic figures, including atypical forms. However, benign nodules are distinguished from sarcomatous or carcinomatous nodules by their grossly and microscopically sharply circumscribed borders and an absence of vascular invasion. They also tend to be smaller than malignant nodules (1 to 6 cm) and occur in younger women. There have been no recurrences in any of the 26 cases with follow-up data.

Mural nodules with anaplastic carcinoma contain pleomorphic cells with markedly enlarged nuclei with prominent nucleoli and abundant eosinophilic cytoplasm, the latter sometimes imparting a rhabdoid appearance ( Figs. 25.81 and 25.82 ). There may be a background of spindle cells or spindle cells may predominate. Both the epithelioid and spindle malignant cells stain strongly with cytokeratin antibodies, a helpful confirmatory finding. However, focal or weak keratin staining in spindle cells is nonspecific, and one must be cautious not to interpret a pseudosarcomatous nodule with focal keratin positivity as an anaplastic carcinoma with spindle cells. Patients with mural nodules composed of anaplastic carcinoma have been reported to have a high risk of metastatic carcinoma and its attendant dire prognosis. However, in a study of 34 patients with mural nodules, all nine stage IA patients with anaplastic carcinoma were alive and clinically free of disease after a median follow-up time of 5 years.

Sarcomatous and carcinosarcomatous mural nodules are rare. Fibrosarcomas, rhabdomyosarcomas, and pleomorphic undifferentiated sarcomas have been reported. The prognosis is poor, with six of nine patients dead with metastases.