Cytologic Histotyping of Gynecologic Malignancies in Peritoneal Fluids Is Reliable When Compared to Its Corresponding Surgical Specimen.

Thomas Sabljic: data curation, formal analysis. Si Kei (Sandy) Lou: conceptualization, data curation, formal analysis, visualization, writing – original draft, review, and editing.

The project was conducted in accordance with our institutional Research Ethics Board–approved protocol. A retrospective search of the laboratory information system at University Health Network (UHN) was performed to identify malignant PTFL with a GM diagnosed on SS, during the period of 2017 and 2022. A total of 899 malignant PTFL were identified. GM affected 55.8% ( n  = 502) of malignant PTFL, of which 15.7% ( n  = 79) was the only diagnostic specimen, 22.3% ( n  = 112) was the initial diagnostic sample (with a subsequent SS) and 1.4% ( n  = 7) had a preceding nondiagnostic SS. Cases with a preceding or concurrent diagnostic SS were excluded from analysis ( n  = 304; 60.6%) since these tended to defer to SS for definitive subtyping. All fluids were examined using the ThinPrep (Cytyc Corporation; USA) method. All CBs were generated from fresh fluid or residual PreservCyt material (if there was insufficient fresh fluid remaining) after formalin fixation and processing using the Histogel method (ThermoFischer Scientific; Waltham, MA, USA). Automated processing was done for CBs and SS using the program settings for small and large histologic specimens, as appropriate. Cytohistologic correlation was performed by comparing the diagnosis of CS to its corresponding SS (from the pathology report) and deemed concordant when the cytologic diagnosis was equivalent to the SS diagnoses—including low‐grade serous neoplasm (LGSN), serous carcinoma, mucinous adenocarcinoma (MAC) and identification of one component of a mixed/synchronous malignancy. CS that were not further subtyped (e.g., adenocarcinoma; AC) or incorrectly subtyped as compared to the SS were deemed discordant. Slides from discordant and selected concordant cases were reviewed. Specimen characteristics including CS volume submitted, CB cellularity (assessed visually on hematoxylin–eosin‐stained (HE) slides at 4‐μm thickness) and immunophenotype were recorded in order to deduce potential factors leading to cytohistologic discordance. Low CB cellularity is deemed < 100 tumor cells. Limited IHC panel refers to the lack of WT‐1 and/or p53 performed. High‐grade cytomorphologic features include the presence of mitotic figures and/or pleomorphism and/or prominent nucleoli. Classic features of CCC refer to the presence of intranuclear inclusions and/or hobnailing and/or raspberry bodies. Features compatible with HGSC include complex hierarchical branching, significant pleomorphism, brisk mitotic activity, and prominent vacuolations. Statistical analysis was performed using Prism 9 (GraphPad Software; San Diego, CA, USA). Cohen's kappa was used to measure cytohistologic concordance. The two‐tailed Fisher test was used to compare cytohistologic concordance categorical variables. The correlation between concordance and CS volume was performed using logistic regression.

The patient demographics and specimen characteristics are summarized in Table  1 . The patient age range was 29–90 years (median: 61.5 years). The CS included had a median volume of 180 mL (range: 5–4000 mL). The SS included consists of 59 biopsies (38 omental, 10 peritoneal, two endometrial, one lymph node, one ovarian, and 7 other biopsies) and 53 SRs (two hysterectomy only, seven salphingo‐ophorectomy (SO) ± omentectomy, 43 hysterectomy with SO ± omentectomy and one retroperitoneal lymph node dissection). Clinical and specimen characteristics of cases. Note : Anatomical site based on surgical specimen. Abbreviations: BSO, bilateral salpingo‐ophorectomy; CS, cytologic specimen; TAH, total abdominal hysterectomy; USO, unilateral salpingo‐ophorectomy. Of the 112 initial diagnostic CS, 103 (91.9%) were subtyped with a concordance rate of 91.2% (almost perfect agreement, K  = 0.842; 95% confidence interval [CI] of 0.719–0.965). The overall subtyping concordance rate (including cases that were not further subtyped—i.e., AC) was 83.9% (substantial agreement, kappa = 0.697; 95% CI of 0.568–0.825) (Table  2 ). In terms of anatomic site, ovarian carcinomas have the highest concordance rate of 87.1%, followed by endometrium (61.5%) and endocervix (0%). HGSC was most correctly subtyped as EOC (91.1%), followed by CCC (80.0%) and LGSN (55.6%; LGSC = 57.1% and SBT = 50.0%). Cytohistologic correlation by diagnosis and anatomical site on surgical specimen. Note : Diagnosis based on surgical specimen. Nonspecific diagnosis includes high‐grade carcinoma and poorly differentiated malignant neoplasm. A total of 18 cytologic diagnoses (16.1%) were discordant with their corresponding surgical diagnoses (Table  3 ); of these, six EOC cases (5.9%) were incorrectly subtyped. These included one HGSC misclassified as clear cell carcinoma (CCC; Case 1); one serous borderline tumor (SBT) with microinvasion and noninvasive implants misclassified as HGSC (Case 2); three LGSC misclassified as HGSC (Cases 3–5); and one HGSC misclassified as LGSC (Case 6). In Case 1, cytomorphology demonstrates features of both HGSC and CCC; in addition, interpretation of IHC was confounded by heterogeneous expression of WT‐1 and HNF‐1beta in the CS. This is compared to the SR status post‐NAC, showing features more suggestive of HGSC and diffuse positivity for WT‐1 and HNF‐1beta. In Case 2, clusters of cells typical of LGSC were seen on CS (Figure  1 ) in a background of vacuolated cells that could have been misinterpreted as a higher grade lesion. In addition, no IHC was performed. In Case 3, due to specimen limitations (i.e., low volume and low cellularity), low‐grade cytomorphology was not represented, and further ancillary work‐up was not attempted. Similarly, in Case 4, focal high‐grade features (including atypia and mitotic figures) were seen on the biopsy, but the low‐grade features were not well presented in the CS. In both Cases 4 and 5, despite an IHC profile supportive of the ultimate SS diagnoses (i.e., p53 wildtype expression), the sign‐out cytopathologist relied more on the high‐grade cytomorphologic features for diagnosis. In Case 6, the absence of high‐grade features and limited IHC characterization may have contributed to the incorrect grading of the serous carcinoma (e.g., no p53). Characteristics of peritoneal fluid that were discordant with the surgical diagnosis. Abbreviations: AC, adenocarcinoma; CCC, clear cell carcinoma; G, grade; HGSC, high‐grade serous carcinoma; IHC, immunohistochemistry; SBT, serous borderline tumor. High‐grade features include the presence of mitotic figures and/or pleomorphism and/or prominent nucleoli. Features compatible with HGSC include complex hierarchical branching, significant pleomorphism, brisk mitotic activity and prominent vacuolations. Classic features of CCC refers to presence of intranuclear inclusions and/or hobnailing and/or raspberry bodies. Cell block cellularity was visually assessed, and low cellularity is < 100 tumor cells. Limited IHC panel refers to the lack of WT‐1 and/or p53 performed. Peritoneal fluid diagnosed as high‐grade serous carcinoma (HE cell block 400×, left). There are epithelial clusters with monotonous nuclei which is morphologically consistent with serous borderline tumor (arrowhead), displayed against a background of vacuolated cells. These vacuolated cells could represent reactive mesothelial cells or macrophages, which may benefit from immunohistochemical work‐up (Case 2). The subsequent resection specimen was diagnosed as serous borderline with microinvasion and noninvasive implants (HE 400×, right). [Color figure can be viewed at wileyonlinelibrary.com ] In 12 CS (10.7%), there were no further subtyping beyond AC or carcinoma (seven ovarian, four endometrial, and one endocervical AC). Five cases were subsequently diagnosed as ovarian HGSC (Cases 7–11) in which the cytomorphology can be compatible with HGSC, but due to specimen limitations (Case 8) and insufficient IHC characterization (Cases 7, 9–11), no further subtyping was attempted. Two cases were subsequently diagnosed/favored as CCC of the ovary (Case 12) and of the endometrium (Case 14) in which the CS lacked classic cytomorphologic features of CCC, and no further subtyping was performed despite a supportive IHC profile, or IHC was simply not performed, respectively. Three cases were subsequently diagnosed as endometrial endometrioid AC (Cases 13, 15, 16). One of these cases demonstrated evidence of mucinous differentiation on SR (Case 13) and was mistakenly diagnosed as AC with intestinal differentiation on CS. Two others were diagnosed as high‐grade endometrioid AC (Cases 15 and 16) in which ancillary testing was not pursued at the pathologist's discretion. One case was diagnosed as endometrial endometrioid AC, Grade 3, and undifferentiated carcinoma (Case 17), in which the CB was low in cellularity, thus precluding further work‐up. In the remaining case, HPV‐associated endocervical AC was diagnosed on resection (Case 18), in which further work‐up was precluded on the CS by specimen limitations (low volume and low CB cellularity). Next, we evaluated the correlation between different CS characteristics and cytohistologic concordance. There is a statistically significant difference in the usage of IHC (94.0% versus 73.7%; p  = 0.018) and the volume submitted for cytologic examination ( p  = 0.038) between concordant and discordant cases. There is no significant difference in CS volume submitted between the various anatomical sites ( p  = 0.847). The rate of concordance between CS and its corresponding biopsy (88.3%) versus SR (80.8%) was not statistically significant ( p  = 0.300). Due to known limitations of small samples in diagnosing LGSN (includes LGSC and SBT) and AC with mucinous/intestinal differentiation or mucinous AC, these cytologic diagnoses were considered concordant with their surgical counterpart. Identification of one component of a mixed or synchronous carcinoma, as well as carcinosarcoma, was also considered cytohistologically concordant. In six cases (5.4%), the CS was correctly diagnosed as serous carcinoma (Table  4 ); however, grade was not specified. These were subsequently diagnosed on SS as ovarian HGSC (Cases 19–23) and ovarian LGSC (Case 24). In four cases, the responsible pathologist chose not to further subtype despite a confirmatory IHC profile (Cases 19 and 22) or chose not to perform further IHC characterization despite a sufficient CB (Cases 21 and 24). In one case, the p53 IHC was difficult to interpret due to high background staining; thus, further subtyping was not attempted (Case 23). In Case 20, further work‐up was hampered by specimen limitations. In three cases that were diagnosed as LGSN (with a differential of LGSC and SBT), these were subsequently diagnosed as LGSC ( n  = 2) and SBT ( n  = 1). No HGSC was misclassified as LGSN on CS. Characteristics of peritoneal fluid diagnosed as serous carcinoma or mucinous adenocarcinoma. Abbreviations: AC, adenocarcinoma; CS, cytologic specimen; GI, gastrointestinal; HGSC, high‐grade serous carcinoma; IHC, immunohistochemistry; LGSC, low‐grade serous carcinoma. High‐grade features include the presence of mitotic figures and/or pleomorphism and/or prominent nucleoli. Cell block cellularity was visually assessed, and low cellularity is < 100 tumor cells. Limited IHC panel refers to the lack of WT‐1 and/or p53 performed. Two cases were diagnosed as AC with intestinal phenotype/differentiation (Table  4 ) on CS. In one case, it was subsequently diagnosed as ovarian mucinous AC, moderately differentiated, intestinal type (Case 25). In the other case, both the CS and SS did not specify a potential site of primary based on limited sampling (Case 26). In seven cases (6.3%), CS identified one of the carcinomatous components of a mixed/synchronous malignancy. These include five carcinosarcomas in which the serous carcinoma component ( n  = 4) (Figure  2 ) and AC with gastrointestinal (GI) differentiation ( n  = 1) were identified in the PTFL. The associated sarcomatous components on resection include chondrosarcoma ( n  = 3), rhabdomyosarcoma ( n  = 1) and unspecified ( n  = 1). In one case, the HGSC component was identified in the PTFL of a patient with synchronous ovarian HGSC and an incidental Grade 1 endometrial endometrioid AC on resection. In one case of mixed endometrial serous and CCC, only the clear cell component was identified in the PTFL. Peritoneal fluid showing only the carcinomatous component—high‐grade serous carcinoma (HE cell block 400×, left). This is characterized by clusters of tumor cells with marked nuclear atypia and macronuclei. Isolated tumor cells are evident. The corresponding surgical resection shows a carcinosarcoma—high‐grade serous carcinoma and chondrosarcoma (HE 200×, right). [Color figure can be viewed at wileyonlinelibrary.com ] In two cases (1.8%), even the biopsies were unable to further subclassify the tumor beyond the cytologic diagnosis. This includes a case of high‐grade carcinoma on SS with ambiguous morphology. The other case was called a poorly differentiated malignant neoplasm. The tumor remained difficult to precisely subtype on subsequent SR and was ultimately diagnosed as a malignant neoplasm with steroidogenic differentiation after an extensive IHC and molecular work‐up. In seven patients (1.4%), the PTFL provided a more definitive/specific diagnosis than its preceding surgical biopsy (Table  5 ). These include two omental biopsies with only fibro/adipose tissue (Cases I and II) whereas the corresponding CS was able to render a diagnosis of serous carcinoma and CCC. These were confirmed on subsequent SR. In three surgical biopsies (one omentum, one liver, one vaginal nodule), only rare atypical cells were seen, and these were subsequently diagnosed on CS as HGSC (Cases III and VI; Figure  3 ) or AC (Case VII). Similarly, a definitive diagnosis of HGSC was not possible in an endometrial biopsy, given its fragmented nature and paucity of lesional cells (Case V). Nondiagnostic surgical biopsy with subsequent diagnostic cytologic specimens. Abbreviations: CCC, clear cell carcinoma; HGSC, high‐grade serous carcinoma; NA, not available. Omental biopsy with rare atypical cells (arrowhead, HE 200×, left) highlighted by PAX8 (200×, left insert) (Case 3). Corresponding peritoneal fluid showing abundant tumor diagnosed as high‐grade serous carcinoma characterized by tumor cells with marked nuclear atypia and macronuclei scattered singly and in clusters (HE cell block 400×, right). This is confirmed on subsequent surgical resection. [Color figure can be viewed at wileyonlinelibrary.com ]

The study of serous fluids in the context of GM has focused on the risk of malignancy, morphologic features, and IHC investigations of common EOCs [ 21 , 22 ]. There is no study to date examining the reliability of PTFL in GM histotyping, as well as the root cause analysis of discordant cases in a clinical setting. In our experience, histotyping of GM in serous fluids is not performed universally. Our current practice at UHN is to subclassify any CS, including serous fluids, to the best of our ability, similar to any other histologic specimen. This is reflected in a higher subclassification of 92% compared with only 55% previously reported in our institution over a decade ago [ 3 ]. GM typically constitutes the largest proportion of positive PTFL and may represent the only or initial diagnostic specimen, as we have seen in our cohort [ 6 , 23 ]. Hence, the ability to subclassify GM in CS minimizes the need for additional, more invasive sampling, thereby decreasing associated costs and risks, as well as waiting time for definitive management. Reliable diagnoses and histotyping are clinically relevant as type‐specific management is being increasingly used. For example, since LGSC do not typically respond to platinum‐based chemotherapy, unlike HGSC, primary surgical approaches are preferred [ 9 ]. In addition, therapies geared toward the unique pathogenesis of different histotypes are being utilized more often; such as the use of PARP inhibitors in tubo‐ovarian HGSC patients with somatic or germline BRCA1/2 mutations and the use of trastuzumab in HER2 overexpressed uterine serous carcinoma [ 11 , 24 ]. IHC is an important adjunct to morphologic examination in order to correctly subclassify a GM. In our study, we demonstrated a significantly higher rate of IHC utilization in CS that were cytohistologically concordant. In Köbel et al. [ 25 ], they demonstrated the use of six markers (WT‐1, p53, p16, HNF‐1beta, ARID1a, and progesterone receptors) significantly improved the diagnostic accuracy of EOCs in a tissue microarray study. Similar IHC panel approaches have been reported in a more recent study by Zelisse et al. [ 26 ], showing the use of IHC predicted the final EOC histotype in over 90% of cases in which morphology alone cannot. Studies have found reliable IHC correlation between CS and SS in commonly used IHC markers (e.g., WT‐1, PAX8, p53) used for the subtyping of GM [ 15 , 27 ]. In a root cause analysis of indeterminate diagnosis in serous fluids by Gokozan et al. [ 28 ], they demonstrated the most common contributing factors were low cellularity and low volume. Similarly, we found a significant difference in the volume submitted for cytologic examination and its resultant cytohistologic concordance. At times, when tumor cellularity is low, having additional volume allows for additional processing, which may yield a more definitive diagnosis. In three of our discordant cases, low volume and low CB cellularity precluded further ancillary work‐up. Interestingly, in the remaining discordant cases, despite having sufficient material, the responsible pathologist has chosen to either only do a limited IHC panel or chose not to further subtype even when the IHC profile was supportive of the ultimate surgical diagnosis, resulting in a lower concordance rate (83.9%, kappa = 0.697). However, when a pathologist does subtype, the concordance reaches over 91% (kappa = 0.842). In SR and to some degree in biopsies, various parts of the tumor are sampled by representative sections; however, in CS, tumor cells are shed into the peritoneal space. It is unclear whether there is preferential shedding from various parts of the tumor, which may explain why at times focal high‐grade features on SS are overrepresented in CS. This also affects IHC interpretation, as different parts of the tumor may express IHC differently, leading to differences in IHC expression patterns in CS versus SS. For example, in one case, heterogeneous WT‐1 and HNF‐1beta were seen on the CS but not on the subsequent SR. Typically, SS are treated as the gold standard. However, at times, even the surgical biopsies can be scant, making precise diagnosis difficult or nearly impossible. In two of our cases, the surgical biopsies were unable to render a more specific diagnosis than its preceding CS due to limited sampling. In one case, definitive histotyping was not possible even on the SR as the tumor was very difficult to subtype even after an exhaustive work‐up. In seven patients, a diagnosis was rendered in the subsequent PTFL (two of which was confirmed on SR) even though the initial biopsy was nondiagnostic. The distinction between HGSC and LGSN—that is, LGSC, SBT—can be difficult, especially in small samples, but can be mitigated with the use of IHC [ 25 , 29 , 30 ]. In a number of discordant cases, the use of IHC (e.g., p53) may have raised suspicion for HGSC or LGSC even when there was an absence or presence of high‐grade features on cytomorphology alone. In contrast, the precise subclassification of LGSN is more challenging due to overlapping morphologies and IHC profiles [ 31 , 32 ]. In addition, there are no cytologic features that can distinguish SBT and noninvasive implants from LGSC [ 22 ]. In one of our SBT cases, it is unclear whether the presence of abundant vacuolated cells was the result of microinvasion and/or noninvasive implants which can be misinterpreted as a higher grade lesion. IHC use in this scenario may have identified these cells to be either a low‐grade lesion, reactive mesothelial cells, or macrophages, which are common causes for false positive results in serous fluids [ 6 ]. In our remaining two cases of SBT, one was misclassified as HGSC based on morphology and limited IHC evaluation, whereas the other did not go beyond the diagnosis of serous neoplasm. Since the differentiation of SBT from LGSC is not entirely reliable on CS, studies recommend simply diagnosing these cases as LGSN while providing a differential diagnosis of LGSC and SBT [ 22 , 32 ]. No HGSC was misclassified as LGSN. Since our study selected for CS with a malignant surgical follow‐up, we cannot comment on the diagnosis of LGSN from benign mimickers like endosalpingosis and endometriosis. High‐grade cytomorphologic features are evident in both HGSC and CCC. HGSC is characterized by complex hierarchical branching, significant pleomorphism, prominent nucleoli, and brisk mitotic activity [ 22 ]. In contrast, CCC is characterized by moderate nuclear pleomorphism and hobnailing. Uncommonly, it can display distinct, but not entirely specific features, like intranuclear inclusions and raspberry bodies [ 33 , 34 ]. In both cases of CCC that were misdiagnosed as HGSC or AC, the CS showed high‐grade nuclear features but lacked the classic features of CCC. In one case, features suggestive of CCC and HGSC were seen on the CS, but only HGSC features were evident in the SR post‐NAC. It is unclear whether NAC may have altered the morphology. In addition, since treatment data was not available, whether NAC was the culprit in some of our discordant cases is unknown. Another pitfall to the diagnosis of GM in CS is mucinous AC. The intestinal phenotype was usually apparent immunophenotypically, if not readily on morphology, in our cohort. However, differentiating primary ovarian tumors from GI primaries may be difficult, given that many parameters (such as tumor unilaterality, tumor size, clinical stage, and growth pattern) are usually not available in small specimens, and immunophenotypes can be nonspecific [ 35 ]. This is evident in Case 26, in which both the biopsy and CS did not commit to a gynecologic primary, given the limited sampling. Cytologic diagnosis of carcinosarcomas is challenging because of the rare presence of the sarcomatous component in serous fluids. In a study by Wang et al. [ 36 ], they demonstrated that only 2 of 14 positive fluid samples from Mullerian carcinosarcoma patients had spindled sarcomatous cells. In Kundu et al. [ 37 ], the sarcomatous component was not apparent in all 10 cases of serous fluids. In both studies, the heterologous elements were not represented in the CS. Similarly, we did not identify any sarcomatous components in our positive PTFL from our carcinosarcoma patients. Furthermore, the diagnoses of cervical and endometrial AC in PTFLs can be difficult, given the low reported positivity rates of 9% and 15%, respectively. The incidence becomes higher in patients with advanced stage and ovarian metastasis [ 18 ]. This is in contrast to reported rates of 17% (in early stages) to 89% (in advance stages) in ovarian malignancies [ 1 ]. In our only case of endocervical AC on SR, because the cytomorphology was not entirely specific, and due to low cellularity, no further work‐up was triggered on the CS. In the case of synchronous ovarian HGSC and incidental low‐grade endometrial endometrioid AC, only the HGSC is identified. In addition, endometrioid AC, especially high grade, may be difficult to diagnose on CS, and even on biopsies, as the morphologic and IHC features are not entirely distinct from other common high‐grade EOCs [ 26 ]. In our cohort, three endometrioid AC were not further subtyped, even after IHC evaluation on two of them. In addition, the variant morphologies of endometrioid AC may pose additional diagnostic challenges. For example, an endometrioid AC of the ovary was diagnosed as AC with intestinal differentiation on CS because of the presence of mucinous differentiation.

GM affects a large proportion of malignant PTFL and often represents the initial or only specimen prior to treatment initiation. Accurate histologic subtyping is paramount to the clinical management of GM. Cytologic histotyping of GM is reliable as compared to its surgical counterpart. In many instances, the inability to further subtype CS was a due to insufficient immunophenotyping because of specimen limitations or at the pathologist's discretion. On occasion, the surgical biopsy may be nondiagnostic given the paucity of lesional cells, and the diagnosis may rely on PTFL sampling. This highlights an opportunity to further improve the diagnosis/subtyping of GM in PTFL.

Patients with gynecologic malignancies (GMs) often present with increasing abdominal girth and bloating due to the accumulation of ascitic fluid [ 1 ]. For symptomatic relief, patients frequently necessitate therapeutic paracentesis, which can be examined cytologically. This fluid sample usually represents the initial diagnostic specimen before any surgical biopsy is attempted. At times, this could represent the only specimen prior to neoadjuvant chemotherapy (NAC) initiation and/or surgical resection (SR) [ 2 , 3 ]. GM accounts for the majority of metastatic tumors in peritoneal effusions in females, with tubo‐ovarian primaries being the most common, followed by tumors of the endometrium, uterus, and cervix [ 4 ]. Although it is acknowledged that the risk of malignancy of a malignant serous fluid approaches 100% [ 5 , 6 , 7 ], the mere diagnosis of “positive for malignancy” or “malignant cells present” may not suffice to guide clinical management. In our experience, histologic subtyping on peritoneal fluids (PTFL) is not performed in many institutions. This may be due to pathologist preference, the variable availability of ancillary tests and the fear of misclassification. Definitive subtyping is essential to triaging patients into the correct management algorithm. For example, patients with high‐grade serous carcinoma (HGSC) will be considered for NAC followed by cytoreductive surgery, whereas low‐grade serous carcinoma (LGSC) patients will be treated with primary cytoreductive surgery [ 8 , 9 ]. Increasingly, therapies targeting the unique pathogenesis of various tumor types are being utilized, such as the use of trastuzumab in HER2 ‐overexpressed uterine serous carcinoma and the use of PARP inhibitors in ovarian carcinoma patients with somatic or germline BRCA1/2 [ 10 , 11 ]. The ability to utilize an otherwise discarded specimen from therapeutic paracentesis minimizes the need for additional more invasive sampling and its associated risks and costs. In addition, many ancillary tests including BRCA testing can be performed on cytologic specimens (CSs), thereby shortening turn‐around time and treatment delays [ 12 ]. However, there have only been a handful of studies examining the subtyping accuracy of PTFL in GM. For example, Freedman et al. [ 3 ] demonstrated that the diagnosis of epithelial ovarian cancer (EOC) on CSs was comparable to those made by histology. Schwartz et al. [ 2 ] demonstrated that cytomorphology alone was accurate in predicting a diagnosis of ovarian cancer on surgical follow‐up in greater than 90% of pretreated CS. However, both of these studies did not attempt to further subtype the EOC, making the reliability of histologic subtyping in CS largely unknown. More recent studies have utilized immunohistochemistry (IHC) as an adjunct to cytomorphology and have demonstrated reliable classification of common EOC as compared to its surgical specimen (SS) [ 13 , 14 , 15 ]. Besides diagnosis, PTFL involvement has staging and prognostic implications. In the FIGO staging of the primary tumor of the ovary and fallopian tube, the presence of malignant cells in ascites or peritoneal washing upstages a patient to stage IC3 in the absence of pelvic extension [ 16 ]. In a study on the diagnostic accuracy of pelvic and PTFL cytology specimens in ovarian clear cell carcinoma (CCC), they found ascitic fluids to have a higher detection rate compared to pelvic wash specimens (63% vs. 27.5%) [ 17 ]. However, the prognostic significance in endometrial and cervical cancer is more controversial. Some studies indicate a worse prognosis or indicator of aggressive tumor behavior associated with malignant cytology; hence, cytologic examination should be performed and recorded when paracentesis or washings are done [ 18 , 19 , 20 ]. Given the importance of correct diagnosis in GM and the frequent involvement of PTFL, the goal of this study is to establish the reliability of CS in GM histologic subtyping by comparing the diagnoses of the preceding PTFL to its follow‐up SS (biopsy or resection). Root cause analysis of discordant cases was performed by reviewing specimen characteristics (including CS volume submitted, cell block (CB) cellularity), IHC profile, and morphologic examination. In addition, we examined a small subset of cases with an initial nondiagnostic surgical biopsy followed by a diagnostic PTFL specimen.

The authors declare no conflicts of interest.