Serous Carcinoma Mimicking Primary Urothelial Carcinoma on Clinical Evaluation and Pathology: A Potential Diagnostic Pitfall

This study was approved by the institutional review board at each institution. We performed a surgical pathology database search for all bladder specimens obtained during the years 2001 to 2015 and substratified that report to identify cases containing the term serous carcinoma. Reports were manually screened to identify cases in which tissues were obtained concurrent with resection for known ovarian, fallopian tube, or uterine primary carcinoma. Twenty-two cases of high-grade serous carcinoma involving the bladder wall were identified that included both local peritoneal resection of the exterior bladder wall and transurethral resection specimens to sample the luminal bladder surface. Material from 2 patients was unavailable for review, and those patients were excluded from further study.

Hematoxylin-eosin–stained slides from the 20 cases with available material were rereviewed by 3 genitourinary pathologists and a gynecologic pathologist. Bladder specimens were characterized for extent of carcinoma involvement of the bladder wall, morphologic features, and presence of urothelial colonization.

Clinical records for all patients were reviewed. Information collected included age at procedure, prior medical history, presurgical imaging, operative notes, pathology reports, postoperative therapies, and survival.

Prior immunohistochemical stains, when available, were reviewed. All cases were additionally restained for cytokeratin 7 (CK7; 1:1500; Abcam, Cambridge, Massachusetts); GATA3 (1:300; Cell Marque Rocklin, California), PAX8 (1:300; Cell Marque), uroplakin II (UPII; prediluted; Biocare Medical, Concord, California), p63 (1:600; Biocare Medical), and WT1 (1:10; Leica Microsystems, Buffalo Grove, Illinois). In brief, 4-μm sections were obtained from formalin-fixed, paraffin-embedded material. Antigen retrieval was performed using Cell Conditioning I (Ventana Medical Systems, Tucson, Arizona) for 40 to 72 minutes at 95°C, with range dependent upon epitope. Slides were incubated with primary antibody for 1 hour at 37°C, followed by application of secondary UltraMap (Vexcel Imaging, Graz, Austria) antibody-horseradish peroxidase conjugate and incubation of slides with 3,3′-diaminobenzidine. Slides were counterstained with hematoxylin, rinsed, dehydrated with alcohol and xylene, and coverslipped. The UPII antibody was developed using Fast Red chromogen. Stains were scored by 2 pathologists (D.E.H. and O.F.). For each antibody, cancer cells were categorized as negative, weakly positive (1⁺), or strongly positive (2⁺–3⁺) and as present in a focal or diffuse pattern. Appropriate positive and negative controls were used.

We identified 20 patients with a final diagnosis of serous carcinoma involving the bladder (Table 1). Mean patient age was 63 years, although some patients presented as young as 44 years. Fifteen patients (75%) had a known history of tubo-ovarian or uterine serous carcinoma. Three of the 20 patients (15%) had a history of serous carcinoma at least 1 year prior but which presented as a primary bladder carcinoma on clinical evaluation, and 2 patients (10%) had no precedent gynecologic history and presented with a large, regional mass that involved the full thickness of the bladder wall. Five patients (25%) had precedent breast cancer, and 1 patient (5%) subsequently developed breast cancer.

Imaging results were available for 17 of the 20 patients (85%) performed at the time of bladder involvement and showed a bladder-predominant mass in 2 patients (10%), a uterine or adnexal primary mass in 3 patients (15%), and extensive pelvic involvement in 12 patients (60%). Urine cytology was requested in 5 patients (25%) and showed atypical cells in 1 patient and malignant cells in a second. Papanicolaou tests showed the presence of adenocarcinoma in 2 of 13 patients (15%) in which it was performed.

In the 15 of 20 patients (75%) with a suspected high-grade serous carcinoma, surgery included removal of bladder implants in addition to resection of the primary lesion. The remaining 5 patients (25%) were presumed to have a likely primary bladder carcinoma and underwent transurethral resection (1 patient), partial cystectomy (1 patient), or pelvic exenteration (3 patients).

We rereviewed all hematoxylin-eosin slides from the 20 patients in the study to determine the spectrum of morphologic findings of serous carcinoma involving the bladder wall (Table 2). In 13 cases (65%), classic serous carcinoma morphology, including micropapillary, papillary, solid with slitlike spaces, cribriform, or glandular architecture; psammoma bodies; marked desmoplasia; nuclear pleomorphism; and prominent nucleoli were present in a subset of cases and did not pose a diagnostic dilemma (Figure 1, A through D). In addition, those cases typically showed minimal peritoneal involvement.

Carcinomas that penetrated deeper into the bladder wall, however, showed more-variable features that could potentially be misinterpreted as urothelial carcinoma variants in the context of limited clinical information or a clinical impression of a primary bladder cancer, including features that overlapped with urothelial carcinoma variants. Those features included solid growth with basaloid features (3 of 20 patients; 15%), squamouslike differentiation (2 patients; 10%), bland nests (2 patients; 10%), spindled/sarcomatoid change (2 patients; 10%), and syncytial growth with inflammation (1 patient; 5%). Both basaloid and syncytial growth patterns occur in serous carcinoma and can thus show morphologic overlap with poorly differentiated urothelial carcinoma and a lymphoepithelioma-like carcinoma appearance, respectively. Four of the 20 cases (20%) showed these variant morphologies of serous carcinoma. The basaloid appearance was notable for irregular sheets of basophilic cells with peripheral palisading that were associated with a desmoplastic stromal background (Figure 2, A and B). Brisk mitotic activity and prominent nucleoli were appreciated at higher magnification, although psammoma bodies were absent in these cases. One case showed a more-epithelioid appearance (Figure 2, C and D). A syncytial growth pattern was evident in one case, which extended to immediately underlie the surface urothelium (Figure 2, E and F). The carcinoma cells contained a syncytial growth pattern, with carcinoma cells containing vesicular nuclei and prominent nucleoli that could readily mimic a lymphoepithelioma-like carcinoma; however, the prominent inflammatory infiltrate in that patient consisted of neutrophils, rather than lymphocytes and plasma cells.

Several unusual patterns were also evident that could readily mimic an urothelial carcinoma in that location. A subset of cases showed small- to medium-sized epithelial nests with only mild to moderate nuclear pleomorphism infiltrating the full thickness of the bladder wall (Figure 3, A and B). An early suggestion of desmosomes in these cases may give the appearance of squamous differentiation. In addition, one case was also notable for exceptionally bland morphology, which is unusual in the setting of metastatic serous carcinoma (Figure 3, C and D). Two cases also showed a spindled/sarcomatoid appearance that could mimic a sarcomatoid carcinoma of the bladder (Figure 3, E and F).

We also identified 2 of the 20 cases (10%) with surface colonization of the urothelium by serous carcinoma. In both cases, those carcinomas showed exophytic lesions that resembled high-grade papillary urothelial carcinoma with the presence of thin, fibrovascular cores and overlying neoplastic cells (Figure 4, A and B). Two additional cases that were clinically considered a primary bladder carcinoma contained diagnostic comments related to the surface urothelium, including the presence of urothelial atypia (Figure 4, C) and a possible papillary lesion (Figure 4, D). Each of these 2 latter cases was ultimately benign and were either marked reactive atypia or a tangential section.

We next applied a panel of immunohistochemical stains to 17 cases in which blocks were still available (Table 3). We selected CK7, WT1, UPII, p63, GATA3, and PAX8, which would allow for a likely distinction between urothelial carcinoma and serous carcinoma when combined. Individually, CK7, p63, GATA3, WT1, and PAX8 have been reported in each of these entities. UPII has recently been described and has not yet been applied in this setting, to our knowledge. In conventional serous carcinoma with micropapillary or cribriform features, the anticipated immunostaining pattern was present, including diffuse CK7, WT1, and PAX8 immunoreactivity and negative immunostaining for UPII, p63, and GATA3 (Figure 5, A through D). However, a subset of cases showed variations in this anticipated antibody profile that could confound the diagnosis. Basaloid morphology was more likely to show patchy CK7 staining (Figure 5, E), and nested and spindled morphology could show focal GATA3 immunoreactivity (Figure 5, F). Two cases, one of classic cribriform architecture and one with nested architecture, showed expression of p63 (Figure 5, G and H). One case of serous carcinoma even showed focal, strong immunoreactivity for UPII, although findings from the other urothelial markers, which included p63 and GATA3, were negative, and the morphology was consistent with classic serous carcinoma (Figure 6, A through D). UPII has not previously been described in serous carcinoma, and this represents one of the first reports of that finding; however, most prior studies on UPII have primarily focused on its sensitivity in identifying urothelial carcinoma.^8–10  Use of this panel of immunohistochemical stains confirmed serous carcinoma as the diagnosis in all cases, despite occasional aberrant staining patterns.

Both bladder and serous carcinoma occur most frequently in patients in the sixth decade and older, although a broad age range may be seen.^11,12  Advanced-stage bladder cancer presents with transmural involvement of the bladder wall and may show regional extensions into adjacent organs, including the colon, prostate, uterus, vaginal wall, and occasionally, adnexal structures.^13–15  By contrast, serous carcinoma arising in the gynecologic tract typically shows a complex adnexal and/or uterine mass that may be associated with pelvic carcinomatosis and regional implants, including to the bladder peritoneal surface.^12,16,17  Given those unique presentations, the clinical impression derived from imaging, serum and urine analyses, and clinical examination readily distinguishes those 2 entities in most patients, which in turn, informs histopathologic diagnosis if tissue analysis is undertaken before surgical intervention. However, significant diagnostic conundrums associated with areas of clinicopathologic overlap between those entities may occasionally occur.^7,15,18,19  We identified 5 patients whose clinical presentation resulted in a differential diagnosis of bladder carcinoma. Review of those and additional cases identified a spectrum of clinical, morphologic, and immunohistochemical features that could lead to an erroneous diagnosis and improper patient treatment. We review the diagnostic pitfalls of serous carcinoma involvement of the bladder wall and discuss approaches to avoid misdiagnosis of a potential mimicker of bladder carcinoma.

Most patients with serous carcinoma involving the bladder wall will undergo excision of the peritoneal bladder wall implant(s) concurrent with surgical staging of the primary gynecologic carcinoma. The 2 situations identified as clinical confounders in this study include the presence of a large mass that involves pelvic structures and extends through the full thickness of the bladder wall, and a secondary development of a bladder mass at a time remote from the original serous carcinoma resection with clinical findings suspicious for a new bladder cancer primary. For example, low CA 125 levels, absence of a clear-cut adnexal primary carcinoma, clinical assumption of serous carcinoma remission, presence of gross hematuria, and a cystoscopically identified luminal bladder mass were all clinical findings in different patients that resulted in a heightened suspicion of a separate bladder cancer process.

Review of morphologic features associated with serous carcinoma involving the bladder wall identified additional histopathologic features that could also pose a diagnostic dilemma in the setting of bladder wall involvement by serous carcinoma. Of the 20 cases evaluated, conventional serous carcinoma features were present in most and included either micropapillary or cribriform architecture, psammomatous calcification, marked desmoplasia, and prominent nucleoli.^11,20  Micropapillary architecture, however, may also be seen in the context of urothelial carcinoma,²¹  which represents the most common morphologic overlap between these 2 entities and may present a diagnostic challenge in the setting of large, transmural masses or an absence of known serous carcinoma. However, the presence of psammomatous calcification is exceedingly rare in the micropapillary variant of urothelial carcinoma and, if seen, should, therefore, increase the suspicion for a primary gynecologic carcinoma.

A subset of cases deviated from conventional serous carcinoma morphology and resulted in diagnostic challenges, especially because urothelial carcinoma can itself demonstrate a broad spectrum of morphologic findings in the context of variant histology.²²  Specifically, morphology that demonstrated nested growth patterns with or without squamouslike foci, spindled/sarcomatoid growth, basaloid morphology, and syncytial growth patterns were all potential mimickers of conventional or variant urothelial carcinoma. Rereview of those cases identified morphologic features that should warrant caution in the diagnosis, including the presence of dense desmoplasia and absence of surface involvement in most cases.

One unusual aspect of 2 cases was the ability of serous carcinoma to colonize the urothelial lining of the bladder and result in exophytic lesions that resembled high-grade papillary urothelial carcinoma, a finding that has not previously, to our knowledge, been reported in the literature. Fortunately, neither of those 2 cases affected the patients who were considered to have de novo bladder cancer at clinical evaluation. The ability of carcinomas to colonize the urothelium is not unique to this study and is known to occur in colonic adenocarcinoma and other metastatic lesions to the bladder, which can mimic both papillary and flat, high-grade urothelial lesions. Cytologic evaluation can also lead to confusion regarding the presence of an in situ bladder lesion; for example, in their study of the cytology of metastatic lesions in the bladder, Edgerton et al⁶  found that 1 of 3 adenocarcinoma diagnoses (33%) in a series of 4677 urine specimens from female patients originated from the ovary, rather than the bladder. In instances in which a known prior history of serous carcinoma has been reported, however, the presence of putative, in situ lesion should still be viewed with caution and not be considered “diagnostic” of a new urothelial carcinoma.

Given the potential for serous carcinoma to morphologically mimic urothelial carcinoma, we tested a directed panel of antibodies to determine their utility in distinguishing serous and urothelial carcinoma. We selected CK7, WT1, UPII, p63, GATA3, and PAX8 because they are readily available in most laboratories and could be used to address the differential diagnosis. We opted not to use p53 immunohistochemical stain because it has been reported to be immunoreactive in both serous and urothelial carcinoma. Use of this panel could correctly identify the diagnosis in all cases we evaluated, despite occasional outliers. Conventional serous carcinoma most commonly showed diffuse CK7, WT1, and PAX8 expression and absence of p63 and GATA3. CK7 immunoreactivity was present in all carcinomas evaluated, although a subset of cases showed patchy CK7 immunostaining, especially in more basaloid phenotypes. No case was CK7 immunonegative. WT1 and PAX8 immunoreactivity rates were in line with a prior study that evaluated immunohistochemical markers in micropapillary carcinomas of unknown origin, with WT1 positivity in 91% of cases and PAX8 positivity in 100% of cases reported in ovarian carcinomas.²³  In our analysis, all ovarian serous carcinomas showed immunoreactivity for both WT1 and PAX8, although some cases showed more focal and/or weak staining. WT1, for example, has been proposed to be reduced in serous carcinomas after administration of chemotherapy,²⁴  although most other studies show equivalent expression of WT1 and other immunohistochemical markers before and after chemotherapy in serous carcinoma specimens.^25,26  Least useful was PAX8, which showed the greatest variability in staining patterns and which has also been reported in both urothelial and serous carcinomas.^27,28 

Most serous carcinomas interrogated in this study were immunonegative for p63 and GATA3, which represent 2 major immunohistochemical markers expressed in many urothelial carcinomas. In reported series, p63 is expressed in up to 83% of primary urothelial carcinomas and is frequently negative in many serous carcinomas.²⁹  However, mixed data regarding p63 status in serous carcinoma has been reported. Whereas some studies have shown that p63 expression may occur in malignant serous carcinomas of the ovary,³⁰  other studies have suggested that p63 immunoreactivity may actually decrease with malignant progression in serous lesions.³¹  In our analysis, only 2 of 17 immunostained cases (12%) of serous carcinoma demonstrated focal p63 expression—one nonproblematic case associated with conventional serous morphology and diffuse WT1 expression, and one more-challenging case that demonstrated nests of carcinoma cells reminiscent of urothelial carcinoma. This latter case also demonstrated diffuse WT1 expression and absence of UPII.

GATA3 expression has been reported in up to 78% of urothelial carcinomas,³²  including the micropapillary variant of urothelial carcinoma³³  but is typically negative in primary ovarian carcinoma.³⁴  We identified 2 of 17 serous carcinomas (12%) involving the bladder wall that showed focal, weak immunoreactivity for GATA3, including one case of spindled/sarcomatoid morphology and one case containing nests of carcinoma cells. Both cases were diffusely positive for WT1.

Finally, 1 of 17 immunostained cases (6%) of serous carcinoma showed focal, strong immunoreactivity for UPII, which has not previously been reported, to our knowledge. This case showed UPII expression specifically in regions that resembled conventional serous carcinoma, including features such as micropapillary architecture, desmoplasia, and psammomatous calcifications. Additional studies that further expand the number of cases analyzed in serous carcinoma may be useful to determine the frequency of UPII immunoreactivity in this setting.

In summary, we identified a set of diagnostically challenging cases of serous carcinoma of the gynecologic tract involving the bladder. Whereas many clinical scenarios may not pose a diagnostic dilemma, cases occurring after presumed remission of the primary serous carcinoma or clinical findings suspicious for a bladder cancer may confound the differential diagnosis. A subset of serous carcinomas involving the bladder wall may also demonstrate histologic features not classic for serous carcinoma, including morphologic overlap with urothelial carcinoma variants and surface colonization of the urothelium. Awareness of the prior clinical history and a high index of suspicion for secondary involvement of the bladder by serous carcinoma will help direct additional workup to confirm the diagnosis, including immunohistochemical stains for CK7, WT1, UPII, PAX8, p63, and GATA3.