Ovarian serous tumors of low malignant potential

ORIGINAL ARTICLE Lymph Node Involvement in Ovarian Serous Tumors of Low Malignant Potential (Borderline Tumors) Pathology, Prognosis, and Proposed Classification Jesse K. McKenney, MD, Bonnie L. Balzer, MD, PhD, and Teri A. Longacre, MD Abstract: The occurrence of regional lymph node involvement (LNI) in patients with primary ovarian serous tumors of low malignant potential (S-LMP), although well described in the literature, continues to be problematic. Most studies indicate that LNI is not associated with an adverse prognosis, but there has not been a comprehensive study addressing the histologic patterns of LNI, the importance, if any, of classifying the type of LNI (ie, as either noninvasive or invasive in analogy to peritoneal implant classification), or the presence and significance of associated endosalpingiosis. To further evaluate LNI in S-LMP, 74 patients with ovarian S-LMP and a lymph node biopsy or sampling were studied. Thirty-one of 74 patients had LNI in pelvic (18; 58%), mesenteric/omental (9; 29%), paraaortic (8; 26%), or supradiaphragmatic (2; 6%) lymph nodes. The number of involved nodes ranged from 1 to 20 (mean, 11.1). Four patterns of LNI were identified: individual cells, clusters of cells, and simple, nonbranching papillae (28 of 31; 90%); intraglandular (21 of 31; 68%); cells with prominent cytoplasmic eosinophilia ( eosinophilic cell pattern) (16 of 31; 52%); and micropapillary pattern (5 of 31; 16%). LNI was diffuse in at least one lymph node in 13 patients (42%) and formed nodular aggregates greater than 1 mm in 6 patients (19%). Nodal endosalpingiosis was present in 58% of cases with LNI compared with 35% without LNI (P = 0.06). There was no significant difference in survival for patients with LNI compared with patients without LNI. However, the presence of discrete nodular aggregates of epithelium greater than 1 mm in linear dimension without intervening lymphoid tissue was associated with a statistically significant decreased disease-free survival when compared with other patterns of LNI (P = 0.02). Nodular aggregates were strongly associated with desmoplastic fibrous stromal reaction (P = 0.001) and micropapillary architecture (0.02). There was also a trend for decreased survival among patients with LNI without associated endosalpingiosis (56%) From the Department of Pathology, Stanford University School of Medicine, Stanford, CA. The current address of Dr. McKenney is University of Arkansas for Medical Sciences, Little Rock, AR. The current address of Dr. Balzer is Memorial Sloan-Kettering Cancer Center, New York, NY. Supported by American Cancer Society Grant No. 96-50 (to T.A.L.). Presented in part at the 94th meeting of the United States and Canadian Academy of Pathology, San Antonio, TX, February 2005. Reprints: Teri A. Longacre, MD, Department of Pathology, Stanford University School of Medicine, Room L235, 300 Pasteur Drive, Stanford, CA, 94305 (e-mail: longacre@stanford.edu). Copyright r 2006 by Lippincott Williams & Wilkins compared with patients with LNI associated with endosalpingiosis (85%) and those with endosalpingiosis only (93%). This study suggests that patients with ovarian S-LMP may be further substratified into risk categories by the presence of nodular aggregates of S-LMP in lymph nodes, a feature that is more common in cases with micropapillary architecture and associated stromal reaction in the intranodal tumor. This high risk pattern of LNI may have a predictive value similar to invasive peritoneal implants and deserves independent evaluation in future studies of S-LMP. Key Words: ovary, serous borderline tumor, low malignant potential, atypical proliferative, lymph node, implants (Am J Surg Pathol 2006;30:614 624) Ovarian serous tumors of low malignant potential (S-LMP) may be associated with lymph node involvement (LNI) in 21% to 29% of patients who undergo a formal lymph node sampling at the initial staging surgery. 17,20,24,27 Most studies of ovarian S-LMP with clinical follow-up data indicate that lymph node status is not an independent prognostic factor for patient survival, but the presence of lymph node involvement by S-LMP continues to be problematic for several reasons. First, most studies have simply evaluated lymph nodes as either positive or negative ; the specific histologic patterns of nodal involvement by S-LMP and their distinction from nodal involvement by serous carcinoma have only rarely been described. 5,27 Second, although criteria have been proposed for invasive and noninvasive implants in the peritoneum and outcome data have been linked to these subtypes, criteria for classifying LNI into prognostic groups have not been addressed. 6,8,12,13 For example, little is known about the clinical significance of the volume of tumor present in lymph nodes, the site of the tumor involvement within an individual lymph node (ie, subcapsular or parenchymal), or the anatomic location of the involved lymph node. Third, because endosalpingiosis (also referred to as benign mu llerian inclusions) is frequently associated with S-LMP in lymph nodes, it has been proposed that nodal involvement by S-LMP may in some cases arise independent of the ovarian tumor via neoplastic transformation from preexisting endosalpingiosis. 2,5,14,21,26,27 Given these two potential developmental pathways, the 614 Am J Surg Pathol Volume 30, Number 5, May 2006

Am J Surg Pathol Volume 30, Number 5, May 2006 Serous Borderline Tumors definition and the relative prognostic and therapeutic implications, if any, of primary versus metastatic S-LMP in lymph nodes is unknown. In this study, we report a detailed series of 74 patients with ovarian S-LMP who underwent some form of lymph node sampling with emphasis on patterns of involvement by S-LMP, possible predictors of adverse outcome, and association with endosalpingiosis. MATERIALS AND METHODS The study group, drawn from a database containing approximately 540 S-LMP tumors accessioned into the Stanford University Medical Center Division of Surgical Pathology or the Stanford Pathology Consultation Service from the years 1958 to 1998, 20 consisted of 74 patients who underwent a surgical staging procedure that included retrieval of lymph nodes. Slides of the primary ovarian tumor, staging biopsies, lymph node biopsies, and when applicable, resections of recurrent tumor were reviewed. Demographic, intraoperative, and clinical follow-up data were obtained from hospital and clinic charts under the guidelines of the Stanford University Institutional Review Board (Protocol No. 79606). The primary ovarian tumors were classified according to the current World Health Organization criteria into two types: typical S-LMP (serous borderline tumor) and S-LMP with micropapillary features (serous borderline tumor with micropapillary features). 18 All tumors with micropapillary or cribriform features were classified on the basis of the modified Burks criteria. 9,11 The presence of stromal microinvasion was determined using the criteria of Bell and Scully. 4 Ovarian autoimplants were not classified as stromal invasion. 20,23 Extra-ovarian epithelial implants were classified as invasive, noninvasive, or indeterminate as described. 6,20 Lymph nodes were classified as positive or negative based on the presence or absence of involvement by S- LMP. Glands that had complex, branching papillae arising from the surface, frequently with associated epithelial tufting, and/or single cells, papillae, or clusters of epithelial cells within the lymph node sinuses or parenchyma were classified as S-LMP (ie, positive lymph node involvement). Endosalpingiosis, defined as a well-formed glands lined by a single layer of tubal-type epithelium, was not counted as a positive lymph node but was scored separately. Simple, nonbranching papillae protruding into the lumen of the gland did not preclude classification as endosalpingiosis, and was recorded as atypical endosalpingiosis. Intranodal mesothelial cells were not scored as positive in the absence of associated S-LMP. The positive lymph nodes were further evaluated for the following features: architectural pattern and degree of cytologic atypia of the S-LMP epithelium, mitotic figures, presence of tumor in sinuses or parenchyma, greatest linear dimension of epithelial aggregates without intervening lymphoid tissue, number of foci involved in an individual lymph node, presence of diffuse nodal involvement, associated stromal reaction, and extranodal extension. Sinus involvement was defined as the presence of S-LMP within subcapsular and/or medullary sinuses. Parenchymal involvement was defined as S-LMP intimately admixed with extrasinusoidal lymphoid tissue anywhere within the lymph node. Four morphologic patterns were identified: 1) individual cells, cell clusters, and simple papillae; 2) intraglandular aggregates; 3) cells with prominent cytoplasmic eosinophilia ( eosinophilic cells ); and 4) micropapillary. Micropapillary architecture was scored on the basis of criteria provided by Bell et al for extra-ovarian implants. 8 Cytologic atypia was scored as absent, mild, or moderate. Tumors exhibiting nuclear features within the spectrum of a typical ovarian S-LMP (ie, minimal nuclear enlargement and no prominent nucleoli) were scored as having mild atypia, whereas those exhibiting nucleomegaly, typically with prominent nucleoli, were scored as having moderate atypia. None of the cases had the degree of nuclear pleomorphism and hyperchromasia typical of high-grade serous carcinoma. Mitotic figures, if present, were noted. For each case, the presence or absence of associated endosalpingiosis was recorded, and when present, any morphologic continuity between S-LMP and endosalpingiosis was noted. S-LMP stage was determined according to the International Federation of Gynecology and Obstetrics system. 1 When an initial conservative or limited surgical procedure was followed within a brief interval of time (not to exceed 12 months), by an additional surgical staging procedure, both operations were considered a single procedure for determination of stage, and the follow-up interval was determined from the initial procedure. Patient age, associated pregnancy, treatment, and disease status at last known follow-up were obtained from the treating physician or by chart review when available. Disease status was defined as follows: a) dead of disease (DOD), patient died as a result of persistent, progressive, or recurrent serous disease; b) alive with disease (AWD), patient alive with clinical and/or radiographic evidence of persistent, progressive, or recurrent serous disease at last follow-up visit; c) dead of intercurrent disease (DID), patient died of an unrelated cause with no clinical or radiographic evidence of persistent, progressive or recurrent serous disease; or d) no evidence of disease (NED), no clinical or radiographic evidence of persistent, progressive, or recurrent serous disease at last follow-up visit. Recurrent or progressive disease was determined as previously described. 20 The endpoints selected for analysis included disease-free survival and overall survival measured in months from the date of initial diagnosis to the date of death or date of last follow-up. Statistical analysis was performed by means of Fisher exact test (two-tailed). The significance level was set at P<0.05. RESULTS Clinical and Pathologic Features of S-LMP With Lymph Node Sampling (n = 74) Seventy-four patients with S-LMP who had lymph nodes obtained during a staging procedure for their r 2006 Lippincott Williams & Wilkins 615

McKenney et al Am J Surg Pathol Volume 30, Number 5, May 2006 disease were identified from the database. The lymph node status of 52 of these patients was previously reported in brief. 20 The patients ages ranged from 12 to 72 years (mean, 45.4 years; median, 45 years). One patient was pregnant at initial presentation. Twenty S-LMP were International Federation of Gynecology and Obstetrics stage I, 14 stage II, 36 stage III, and 4 stage IV. The number of lymph nodes sampled ranged from 1 to 52 (mean, 10.6; median, 6). Lymph nodes were submitted in their entirety, which in some cases required several blocks, but level sections were not routinely performed. Thirty-one (42%) patients had LNI by S-LMP, and 43 (58%) patients either had no LNI or LNI by endosalpingiosis (or atypical endosalpingiosis) only. Fourteen (19%) patients were upstaged as a result of the lymph node findings. Four of 24 patients with intraperitoneal disease confined to the ovary and 8 of 22 patients with disease confined to the pelvis were upstaged to stage III as a result of LNI, including 1 patient previously reported as stage II. 20 Two patients were upstaged to stage IV on the basis of axillary and supraclavicular lymph node involvement, respectively. The incidence of micropapillary architecture and stromal microinvasion in the primary ovarian tumor did not differ significantly between the two groups (Table 1). However, LNI was strongly correlated with invasive peritoneal implants (P = 0.02). The demographic and pertinent histologic features of the primary ovarian tumors and implants for the two groups of patients are presented in Table 1. Follow-up data, ranging from 8 to 230 months (mean, 65 months; median, 64 months), was available in 63 of the 74 (85%) patients. The overall survival and disease-free survival for all patients was 92% and 82%, respectively (Table 2). Five patients died of disease at intervals ranging from 8 to 106 months. An additional 10 patients developed recurrent disease (5 with LNI and 5 without LNI) with an interval from initial surgery to TABLE 1. Demographic Features and Pathology of Ovarian Tumors and Implants in Patients With S-LMP and Lymph Node Sampling LNI ( ) LNI (+) Total Patients (no.) 43 31 74 Age (mean) (yr) 43.6 48 45.4 FIGO stage Stage I 20 (46.5%) 0 20 (27%) Stage II 14 (33.5%) 0 14 (19%) Stage III 8 (19%) 28 (90%) 36 (49%) Stage IV 1 (2%) 3 (10%) 4 (5%) Ovary Micropapillary 10 (23%) 5 (16%) 15 (20%) Stromal microinvasion 4 (9%) 6 (19%) 10 (14%) Extraovarian implants Total 23 (54%) 27 (87%) 50 (68%) Noninvasive 18 (78%) 19 (70.5%) 37 (74%) Indeterminate 5 (22%) 2 (7.5%) 7 (14%) Invasive 0 6 (22%) 6 (12%) Lymph nodes (mean no.) 10.3 11.1 10.6 LNI, lymph node involvement. TABLE 2. Follow-up Status for S-LMP Patients With and Without Lymph Node Involvement (LNI) Patients Without LNI (n = 43) Patients With LNI (n = 31) Total (n = 74) NED 34 16 50 AWD 2 4 6 DOD 3 2 5 DID* 3 0 3 LFU* 1 9 10 Overall survival 92% 91% 92% Disease-free survival 87% 73% 82% NED, no evidence of tumor; AWD, alive with disease; DOD, dead of disease; DID, dead of intercurrent (unrelated) disease. *Patients lost to follow-up or who died of other causes were censored from analysis. recurrence ranging from 7 to 158 months (mean, 49 months; median, 35.5 months). Five of the patients with recurrent disease were alive with disease at last follow-up. One patient with stage IV disease and axillary lymph node involvement had persistent, but otherwise stable, disease at 11 months of follow-up. Follow-up intervals did not differ significantly between patients with and without LNI, and there was no significant difference in overall survival or disease-free survival between the two groups (Table 2). Anatomic Site of Lymph Node Involvement by S-LMP Thirty-one of 74 patients had LNI in pelvic (18; 58%), mesenteric/omental (9; 29%), paraaortic (8; 26%), and supradiaphragmatic (2; 6%) lymph nodes. Eight patients had mesenteric or omental lymph node involvement alone. There was no correlation between anatomic site of involvement and overall or disease-free survival. There was no significant difference in histologic pattern, extent, or location of LNI in the various lymph node groups. Histologic Patterns of Lymph Node Involvement by S-LMP Several different histologic patterns of LNI were often intermixed in individual patients as well as in the individual lymph nodes and included: individual cells, clusters of cells, and simple papillae (28 of 31; 90%), intraglandular pattern (21 of 31; 68%), prominent cytoplasmic eosinophilia (16 of 31; 52%), and micropapillary architecture (5 of 31; 16%). Most cases of LNI, regardless of histologic pattern, exhibited mild cytologic atypia similar in appearance to the primary ovarian tumor. In 9 cases, LNI exhibited moderate atypia, which was characterized by nuclear enlargement, mild nuclear irregularity, and prominent nucleoli. Mitotic figures were present in 7 cases of LNI, 4 of which also exhibited moderate cytologic atypia. 616 r 2006 Lippincott Williams & Wilkins

Am J Surg Pathol Volume 30, Number 5, May 2006 Serous Borderline Tumors Individual Cells, Clusters of Cells, and Simple Papillae Twenty-eight of 31 (90%) patients with LNI had an admixture of single epithelial cells, clusters of epithelial cells, and/or simple, nonbranching papillae in the lymph node sinus or parenchyma and this constituted the most common pattern of LNI by S-LMP (Fig. 1). Five of these patients had stromal microinvasion and four had micropapillary architecture in the primary ovarian tumor. Sixteen patients with this pattern of LNI had noninvasive peritoneal implants, whereas 6 had invasive implants and 2 had implants that were indeterminate for invasion. Follow-up information was available in 20 of the 28 (71%) patients as follows:2dodat8and74months,3awd(11,61,and 230 months), 15 NED, and 8 with no available follow-up data. Disease-free survival for patients with individual cells, clusters, and simple papillae versus LNI without this pattern was 75% and 50%, respectively (P =0.42). Intraglandular Pattern Twenty-one of 31 (68%) patients with LNI had an intraglandular pattern characterized by a complete glandular space lined by serous type epithelium associated with a complex intraglandular papillary proliferation showing secondary and tertiary branching and epithelial tufting (Fig. 2). Three cases consisted entirely of the intraglandular pattern, one of which formed nodular aggregates with associated stromal reaction. Three patients with intraglandular LNI had stromal microinvasion in the primary ovarian tumor and three had micropapillary architecture. All patients with intraglandular LNI had intraperitoneal implants (3 invasive and 18 noninvasive). Follow-up information was available in 14 of 17 patients (11 NED and 3 AWD at 11, 38, and 230 months). Disease-free survival for patients with the intraglandular pattern of LNI compared with that for patients without the intraglandular pattern of LNI was 79% and 71%, respectively. Prominent Cytoplasmic Eosinophilia (Eosinophilic Cells) Sixteen of 31 (52%) patients with LNI had intranodal epithelium with prominent eosinophilic cytoplasm similar to that commonly seen in ovarian stromal FIGURE 1. Lymph node sinus involvement by S-LMP. Individual cells, clusters of cells, and small, simple papillae of S-LMP are intermixed with psammoma bodies. Note the endosalpingiosis at top right. FIGURE 2. Intraglandular pattern of LNI by S-LMP. The glands contain papillae with secondary branching and tufting. This pattern, in conjunction with that depicted in Figure 1, constitutes the most common histologic patterns of lymph node involvement by S-LMP. r 2006 Lippincott Williams & Wilkins 617

McKenney et al Am J Surg Pathol Volume 30, Number 5, May 2006 FIGURE 3. Lymph node sinuses and parenchyma contain aggregates of single cells and glands with ample eosinophilic cytoplasm ( eosinophilic cells ). Note the stromal response associated with a glandular pattern at right. microinvasion (Fig. 3). Three patients had lymph nodes diffusely infiltrated by eosinophilic cells with a sinus and parenchymal distribution, one with nodular aggregates. Five patients with LNI featuring eosinophilic cells had stromal microinvasion in the primary ovarian tumor and another patient had micropapillary architecture in the primary ovarian tumor; all others had typical S-LMP histology (not significant). Eleven patients with eosinophilic cell LNI had peritoneal implants. Two had invasive peritoneal implants and nine had noninvasive implants. There were no significant associations between eosinophilic cell LNI and peritoneal implant status or subtype. Follow-up was available in 10 (62.5%) patients with eosinophilic cell LNI with intervals of 20 to 87 months (mean, 41 months). Six were NED with no evidence of recurrent disease, two were NED following recurrence at 12 and 74 months, 1 was AWD at 11 months, and 1 was DOD at 8 months. There was no significant difference in overall survival or disease-free survival for patients with eosinophilic cell LNI compared with other patterns of LNI. Micropapillary Architecture Micropapillary architecture in nodal S-LMP was present in 5 of 31 (16%) patients with LNI (Fig. 4). The FIGURE 4. Micropapillary pattern of LNI by S-LMP. Micropapillary architecture forms nodular intraparenchymal aggregates with a stromal response (A). The individual micropapillae are associated with moderate nuclear atypia and rare mitotic figures (B). epithelium generally showed a mixture of two patterns: exophytic with delicate, filiform papillae without fibrovascular cores and endophytic with a gland-like space containing an epithelial proliferation resulting in a complex maze of small slit-like spaces. The number of nodes involved by S-LMP with micropapillary architecture ranged from 2 to 11 (mean, 4.6), compared with a mean of 2.1 (range, 1 12) involved nodes in patients without micropapillary architecture in the intranodal tumor. Three of the 5 (60%) lymph nodes with micropapillary architecture also had an associated stromal reaction, compared with 2 of 26 (8%) without micropapillary architecture (P = 0.02). One had extranodal extension, and this was the only case in the series with that finding. Four of the five (80%) micropapillary cases had nodular aggregates of epithelium ranging in size from 2 to 8 mm, whereas only 2 of the 26 (8%) nonmicropapillary cases formed a nodular aggregate that was 1 mm in size (P<0.001). Three of the 5 (60%) micropapillary cases diffusely replaced at least one lymph node, compared with 10 of 26 (38%) nonmicropapillary cases (not significant). Of the 5 micropapillary cases, 3 (60%) 618 r 2006 Lippincott Williams & Wilkins

Am J Surg Pathol Volume 30, Number 5, May 2006 Serous Borderline Tumors were associated with endosalpingiosis, compared with 16 of 26 (62%) nonmicropapillary cases. One patient with micropapillary LNI had stromal microinvasion in the primary ovarian tumor, and another patient had micropapillary architecture in the primary ovarian tumor; all others had typical S-LMP histology (not significant). All 5 patients with micropapillary LNI had peritoneal implants. Two had invasive peritoneal implants with micropapillary architecture, 2 had noninvasive implants, and 1 had indeterminate implants with micropapillary architecture. There were no significant associations between micropapillary LNI and peritoneal implant status or subtype. Follow-up data were available in 4 patients (80%) with micropapillary architecture in nodal S-LMP with follow-up interval of 11 to 230 months (mean, 101 months). Two patients were AWD at 11 and 230 months, 1 was NED at 87 months following an abdominal wall recurrence at 74 months, and 1 was NED with no evidence of recurrent disease at 77 months. Disease-free survival for nodal S-LMP with micropapillary features was decreased compared with all other patterns of lymph node involvement (50% vs. 82%), but the difference did not reach statistical significance (P = 0.22). Stromal Response in Lymph Node Involvement Five of 31 (16%) cases with LNI had an associated stromal reaction. In all of these cases, the stroma consisted of reactive desmoplasia with focal areas of dense fibrosis showing cleft-like spaces surrounding the epithelium (Fig. 5). Each of the 5 cases with intranodal stromal reaction had nodular aggregates (P = 0.0001), 3 had micropapillary architecture (P = 0.02), and 1 had a diffuse eosinophilic cell pattern (not significant). Two were associated with micropapillary features in the ovarian primary and three were associated with invasive peritoneal implants. Follow-up was available in 3 patients with LNI and an associated stromal reaction (60%). Two were AWD at 230 and 38 months, and 1 was NED at 77 months. Disease-free survival for cases with stromal reaction was 33% compared with 68% for cases without a stromal reaction (not significant). Extent and Location of Lymph Node Involvement The number of lymph nodes involved by S-LMP ranged from 1 to 20 (mean, 3). Seventeen of 31 (55%) patients had LNI in only one sampled node. Disease-free survival for patients with only one involved node versus patients with greater than one involved node was 73% and 80%, respectively. Fourteen of 31 (45%) patients had at least one lymph node diffusely involved by S-LMP. Diffuse involvement was defined as S-LMP present throughout the lymph node, but containing intervening lymphoid tissue between clusters of epithelium. Follow-up data were available for 57% of these patients. There was no significant difference in overall survival or disease-free FIGURE 5. Fibrous stromal response to S-LMP with intraglandular eosinophilic cell pattern. survival in patients with diffuse LNI versus patients without diffuse LNI. Six patients (19%) had nodular aggregates ranging from 1 to 8 mm in greatest linear dimension (Table 3). A nodular aggregate was defined as a collection of epithelium without intervening lymphoid tissue measuring greater than 1 mm in linear dimension (Fig. 6). By this definition, nodular aggregates were typically also scored as diffuse LNI; however, most cases with diffuse LNI did not exhibit nodular aggregates. To qualify as a nodular aggregate, the epithelium had to be arranged in confluent sheets or separated only by desmoplastic stroma. In contrast, diffuse LNI was characterized by epithelial cells of any morphologic pattern scattered throughout the lymph node (ie, not focal), but with intervening lymphoid tissue. Nodular aggregates were strongly associated with desmoplastic fibrous stromal reaction (P = 0.001) and micropapillary architecture (P = 0.02). Disease-free survival for cases with nodular intranodal aggregates versus no aggregates was 25% and 87%, respectively (P = 0.02). The intranodal S-LMP was located within sinuses (Fig. 7) in 26 of 31 cases (84%) (11 with sinus involvement only) and was parenchymal (Fig. 8) in 20 of 31 (65%) (5 with parenchymal involvement only). Disease-free survival for LNI with parenchymal involvement compared with no parenchymal involvement was 69% and r 2006 Lippincott Williams & Wilkins 619

McKenney et al Am J Surg Pathol Volume 30, Number 5, May 2006 TABLE 3. Patients With S-LMP and Lymph Node Involvement by Nodular Aggregates Follow-up Status (mo) Adjuvant Therapy Mitotic Figures Cytologic Atypia Associated Findings Intranodal Pattern Nodular Aggregate Size (mm) ES LN Location (no. positive) Implant Status FIGO Stage Age (yr) Ovary 70 S-LMP III Invasive Mesenteric (3) 2 + MP; IG; ICP SR Mild Present None NED (77) 39 S-LMP III Noninvasive Mesenteric (4) 8 MP; IG SR; ENE Moderate Present None AWD (230) 58 MP III Invasive Paraaortic (2) 4 MP; IG; ICP SR; diffuse Moderate Present LFU 44 S-LMP; MI IV Indeterminate Axillary (3) 2 + MP; focal EC; IG Diffuse Moderate Present Tamoxifen AWD (11) 67 MP III Noninvasive Pelvic (1) 1 + IG SR Mild Absent None AWD (38) 57 S-LMP IV Invasive Mesenteric (1) 1 + EC Focal SR; diffuse Mild Absent LFU MP, micropapillary; MI, microinvasion; NI, noninvasive; I, invasive; ES, endosalpingiosis; EC, eosinophilic cell; ICP, isolated cells, clusters, and simple papillae; IG, intraglandular; SR, stromal reaction; ENE, extranodal extension; DOD, dead of disease; AWD, alive with disease; LFU, lost to follow-up. FIGURE 6. Nodular aggregate of S-LMP in parenchyma of lymph node. Individual simple and branching papillae and cell clusters form an aggregate within the lymph node, measuring at least 1 mm in linear extent. This type of LNI is associated with a statistically increased risk of disease recurrence in patients with ovarian S-LMP and LNI. 77%, respectively (P = 0.61). There was also no significant difference in disease-free survival for patients with sinusoidal LNI versus LNI without sinusoidal involvement (76% and 75%, respectively). Association of Lymph Node Involvement With Endosalpingiosis Eighteen of 31 (58%) cases with LNI had associated endosalpingiosis (Fig. 9), whereas 15 of 43 (35%) cases without LNI had nodal endosalpingiosis (P = 0.06). In most cases, the endosalpingiotic foci without associated LNI exhibited no significant cytologic atypia; however, 4 cases had mild atypia and 1 exhibited moderate atypia. In the 18 cases with both intranodal S-LMP and endosalpingiosis, 3 cases had atypical endosalpingiosis and 3 had foci reminiscent of ovarian stromal microinvasion. These microinvasive-like foci were characterized by a glandular structure with internal papillary architecture (intraglandular pattern) in which the lining cells focally developed prominent cytoplasmic eosinophilia, discohesion, and development of underlying stromal retraction (Fig. 9). 620 r 2006 Lippincott Williams & Wilkins

Am J Surg Pathol Volume 30, Number 5, May 2006 Serous Borderline Tumors FIGURE 7. Sinus involvement by S-LMP. Eosinophilic cells with mild to focally moderate cytologic atypia fill subcapsular sinus and extend into the intranodal sinus. FIGURE 8. Intraparenchymal involvement by S-LMP. Papillae are surrounded by lymphocytes with no apparent surrounding clefts or sinus space. Individual cells and clusters of epithelial cells in the lymph node parenchyma and sinuses were commonly adjacent to these microinvasive-like foci. Follow-up data were available in 63% of patients with LNI and endosalpingiosis, 69% of patients with LNI but no endosalpingiosis, 93% of patients with endosalpingiosis only, and 93% of patients with no LNI and no endosalpingiosis. Although there was no significant difference in overall survival for patients with and without LNI, there was a trend for improved survival among patients with endosalpingiosis only (93%) compared with patients with LNI and endosalpingiosis (85%) and patients with LNI but no associated endosalpingiosis (56%). DISCUSSION As long as S-LMP (borderline) tumors have been recognized by the World Health Organization, efforts have been ongoing to identify features of the primary ovarian neoplasm and/or the extra-ovarian implants that identify patients at increased risk for clinical relapse or disease progression. These efforts are often confounded by several important clinicopathologic factors that are inextricably linked to this group of tumors, each of which have been discussed at length elsewhere. 7,15,19,20,24,25 Our study, which constitutes the largest series of S-LMP with pathologically documented LNI with clinical follow-up information to date, demonstrates that nodal involvement by S-LMP assumes at least four histologically distinct patterns, each of which may be present in various combinations, and all of which should be distinguished from lymph node involvement by metastatic carcinoma. As has been shown in previous studies, lymph node status does not appear to be an independent prognostic factor for patient survival in patients with S-LMP. 7,20 However, when patients with primary ovarian S-LMP and LNI are stratified by the type of lymph node involvement, the presence of nodular aggregates of LNI in these patients is associated with a statistically significant adverse prognosis independent of histologic pattern. Thirteen percent of the patients in this study were upstaged by LNI. This figure is lower than that of Leake and co-workers but higher than that reported in recent studies 17,22,28 and is attributed to differences in study design and to evolving staging practices. Because we were interested in identifying histologic patterns of LNI, we r 2006 Lippincott Williams & Wilkins 621

McKenney et al Am J Surg Pathol Volume 30, Number 5, May 2006 FIGURE 9. Partial endosalpingiotic gland involvement with stromal microinvasion pattern. The individual eosinophilic cells form discrete clusters in the nodal parenchyma immediately beneath the endosalpingiotic gland (A). Higher magnification demonstrates partial involvement of the endosalpingiotic gland by eosinophilic cells, similar in appearance to those in the adjacent clefts (B). included any patient with a primary ovarian S-LMP who had a lymph node sampled during initial staging, whether it was obtained during a formal surgical pelvic or paraaortic lymph node sampling procedure or incidentally during a staging procedure that included mesenteric/ omental or peritoneal biopsies. Other studies, focusing primarily on the utility of performing formal lymph node sampling in the staging procedure, confined their results to only those patients who underwent full pelvic and/or paraaortic lymph node dissections. The relatively high incidence of LNI in this study (42%) is due to the selective inclusion of a significantly higher proportion of patients with advanced-stage disease undergoing lymph node sampling over the years encompassed by this study as well as to the inclusion of these incidental lymph nodes. LNI was commonly associated with peritoneal implants (87% of cases), and this was statistically significant, as has been previously reported. 17 LNI was also associated with a higher incidence of disease recurrence, but unlike previous studies, 16,17 this difference did not meet statistical significance. Histologic Patterns of S-LMP Lymph Node Involvement The histologic patterns of LNI have been described in the literature only rarely. 5,7,27 The most common histologic pattern identified in intranodal S-LMP was an admixture of individual epithelial cells, clusters of epithelial cells, and simple, nonbranching papillae. This pattern frequently coexisted with an intraglandular pattern featuring glandular spaces lined by serous epithelium forming complex intraglandular papillary proliferations with secondary branching and epithelial tufting, similar to the pattern of typical S-LMP arising in the ovary. Although both patterns may be associated with a parenchymal as well as the more typical sinus nodal distribution, they could be distinguished from lymph node involvement by metastatic low-grade serous carcinoma on the basis of the comparatively low volume of epithelium in LNI, the presence of minimal to at most moderate cytologic atypia, and rare to absent mitotic figures. Micropapillary S-LMP in lymph nodes has not been previously studied and appears to be relatively rare (16% in this series), but the micropapillary pattern is considered to be one of the main candidate histologic markers of possible prognostic significance in S-LMP. Primary ovarian S-LMP with micropapillary epithelial overgrowth are more often bilateral, exophytic, and associated with extraovarian implants than S-LMP without this appearance. 7,9,11,20 In a recently completed study of 276 patients with ovarian S-LMP and at least 5 years of follow-up, micropapillary architecture was strongly associated with disease progression over time and decreased overall survival on univariate analysis, even though it did not have a significant adverse effect on survival independent of implant type. 20 In this study, there was also a trend for decreased disease-free survival in patients with micropapillary LNI (50% vs. 82% for all other patterns of LNI), but this association did not reach statistical significance (P = 0.22). As with primary ovarian micropapillary tumors, the micropapillary pattern was more frequently associated with several other histologic features that might be considered more aggressive, including stromal reaction (60% vs. 12%), diffuse nodal involvement (60% vs. 38%), nodular aggregates (33% vs. 12%), and extranodal extension (20% vs. 0%). A second pattern that has received attention in the recent literature is the infiltration by large numbers of individual, often eosinophilic cells or solid nests of cells in peritoneal implants and lymph nodes. 7 This eosinophilic cell pattern of LNI is similar in appearance to the eosinophilic cell pattern of stromal microinvasion in the ovary and was associated with stromal microinvasion in the primary ovarian tumor in almost one third of the cases in this series. This pattern of LNI was also associated with one of the two deaths in this series of S-LMP with LNI. In that case, the eosinophilic cell pattern was only focally present in the involved lymph nodes but was prominent throughout the peritoneal 622 r 2006 Lippincott Williams & Wilkins

Am J Surg Pathol Volume 30, Number 5, May 2006 Serous Borderline Tumors implants, which were otherwise indeterminate for tissue invasion. 20 Although the association between eosinophilic cell LNI and adverse outcome is not statistically significant, the numbers of reported cases with this pattern of extra-ovarian spread are few and experience is very limited. The main differential diagnosis of this pattern of LNI is intranodal hyperplastic mesothelial cells, a distinction that may be difficult on routine hematoxylin and eosin-stained sections in some cases. 10 This distinction is particularly difficult in cases showing only single, discohesive cells, although eosinophilic papillary clusters may also pose diagnostic difficulty. In problematic cases, immunohistochemistry for Ber-EP4 and calretinin can be useful. S-LMP is characterized by cytoplasmic reactivity with Ber-EP4 but no nuclear calretinin expression, whereas mesothelial cells have the opposite phenotype. 3 Cytokeratin immunostains have limited utility in this setting because both lesions will show reactivity, and, in addition, the background population of intranodal interstitial reticulum cells may also be highlighted. Nodular Aggregates of LNI Are Associated With Adverse Prognosis, Regardless of Histologic Pattern In this series, the presence of nodular aggregates of epithelium greater than 1 mm, without intervening lymphoid tissue, predicted a statistically significant decreased disease-free survival independent of all other features. Nodular aggregates were also strongly associated with micropapillary architecture and nodal stromal reaction. The association of nodular aggregates with invasive peritoneal implants in 3 of 6 cases suggests that this pattern of LNI may represent the nodal equivalent of invasive implants and should probably be classified separately from other patterns of LNI. Nodular aggregates did not occur in association with any specific lymph node group in this study, although prior studies have reported massive intraparenchymal involvement by recurrent S-LMP with preferential involvement of the extraabdominal lymph nodes. 27 In our experience, the presence of nodular aggregates is more commonly encountered in recurrent disease and delayed lymph node involvement than at initial presentation; therefore, any nodal aggregates greater than 1 mm at presentation, regardless of the associated histologic pattern, should be regarded as a high-risk lesion analogous to invasive peritoneal implants. Whether these high-risk lesions represent an early step toward transformation to lowgrade serous carcinoma is an open question that has yet to be resolved. In contrast to the nodular pattern of LNI, the disease-free survival for patients with diffuse nodal involvement (in which tumor was present throughout the node, but with intervening lymphoid tissue) and with multiple positive lymph nodes was not significantly different from that for patients with LNI without these features. At Least Some S-LMP May Arise Primarily in Foci of Endosalpingiosis Given the methodology of a morphologic review, it is difficult to address the origin of S-LMP in lymph nodes, although credible arguments can be posed for metastasis as well as independent primary origin. Theoretically, LNI could occur via primary spread from an ovarian tumor, secondarily from an extra-ovarian implant, or de novo from a benign intranodal glandular inclusion. Several authors have noted the association with endosalpingiosis and have suggested a transformation in preexisting mu llerian inclusions. 5,14,21,27 One group has reported identical K-ras mutations in S-LMP and adjacent benign mu llerian inclusions, adding molecular evidence to support origin in lymph nodes in at least some cases. 2 Our study identified areas resembling ovarian stromal microinvasion in the epithelium of the intraglandular pattern of nodal S-LMP. Individual cells and clusters of epithelial cells with prominent eosinophilic cytoplasm in adjacent parenchyma and sinuses appeared to arise from these foci of microinvasion. This finding would seem to support the concept that S-LMP may arise in endosalpingiosis in at least a subset of cases. Interestingly, there was a trend for decreased survival among patients with LNI but no associated endosalpingiosis (56%), compared with patients with both LNI and endosalpingiosis (85%) and patients with endosalpingiosis alone (93%). This trend did not meet statistical significance but suggests that nodal implants without associated endosalpingiosis are either a more advanced stage of intranodal serous epithelial proliferation or metastases. S-LMP with ovarian stromal microinvasion and LNI, another scenario that could represent metastatic disease, was not associated with a similar trend for adverse survival. CONCLUSIONS In summary, lymph node involvement by S-LMP exhibits a variety of histologic patterns, most commonly that of single cells, cell clusters, and papillae similar to that seen in typical S-LMP. Intraglandular, micropapillary, and diffuse sinus and parenchymal expansion by eosinophilic cell patterns may also occur. As with the clinical and pathologic features of primary ovarian S-LMP, no single histologic pattern of LNI is entirely predictive of adverse outcome. 20 Regardless of morphologic pattern, however, the presence of discrete, nodular aggregates of S-LMP greater than 1 mm appears to be associated with decreased disease-free survival independent of implant type. These findings suggest that patients with noninvasive implants and LNI may be further substratified into relative risk categories by the presence of nodular aggregates of S-LMP, a finding that is more common in cases with a micropapillary pattern and an associated stromal reaction in the intranodal tumor. This high-risk pattern of LNI may be analogous to invasive peritoneal implants in terms of prognostic significance and deserves independent assessment in future studies of S-LMP. r 2006 Lippincott Williams & Wilkins 623

McKenney et al Am J Surg Pathol Volume 30, Number 5, May 2006 ACKNOWLEDGMENTS The authors thank all the clinicians and pathologists who provided the case material and additional clinical and pathologic follow-up as well as Dr Michael R. Hendrickson and Dr. Richard L. Kempson; this study would not be possible without their invaluable contributions. REFERENCES 1. International Federation of Gynecology and Obstetrics. Annual report on the results of treatment in gynecological cancer. Int J Gynecol Obstet. 1989;28:189 190. 2. Alvarez AA, Moore WF, Robboy SJ, et al. K-ras mutations in Mullerian inclusion cysts associated with serous borderline tumors of the ovary. Gynecol Oncol. 2001;80:201 206. 3. Attanoos RL, Webb R, Dojcinov SD, et al. Value of mesothelial and epithelial antibodies in distinguishing diffuse peritoneal mesothelioma in females from serous papillary carcinoma of the ovary and peritoneum. Histopathology. 2002;40:237 244. 4. Bell DA, Scully RE. Ovarian serous borderline tumors with stromal microinvasion: a report of 21 cases. Hum Pathol. 1990;21:397 403. 5. Bell DA, Scully RE. Lymph node involvement by ovarian serous borderline tumors. Mod Pathol. 1992;5:61A. 6. Bell DA, Weinstock MA, Scully RE. Peritoneal implants of ovarian serous borderline tumors: histologic features and prognosis. Cancer. 1988;62:2212 2222. 7. Bell DA, Longacre TA, Prat J, et al. Serous borderline (low malignant potential, atypical proliferative) ovarian tumors: workshop perspectives. Hum Pathol. 2004;35:934 948. 8. 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. 2001;25:419 432. 9. Burks RT, Sherman ME, Kurman RJ. Micropapillary serous carcinoma of the ovary: a distinctive low-grade carcinoma related to serous borderline tumors. Am J Surg Pathol. 1996;20:1319 1330. 10. Clement PB, Young RH, Oliva E, et al. Hyperplastic mesothelial cells within abdominal lymph nodes: mimic of metastatic ovarian carcinoma and serous borderline tumor: a report of two cases associated with ovarian neoplasms. Mod Pathol. 1996;9:879 886. 11. Eichhorn JH, Bell DA, Young RH, et al. Ovarian serous borderline tumors with micropapillary and cribriform patterns: a study of 40 cases and comparison with 44 cases without these patterns. Am J Surg Pathol. 1999;23:397 409. 12. Gershenson DM, Silva EG, Levy L, et al. Ovarian serous borderline tumors with invasive peritoneal implants. Cancer. 1998;82:1096 1103. 13. Gershenson DM, Silva EG, Tortolero-Luna G, et al. Serous borderline tumors of the ovary with noninvasive peritoneal implants. Cancer. 1998;83:2157 2163. 14. Kadar N, Krumerman M. Possible metaplastic origin of lymph node metastases in serous ovarian tumor of low malignant potential (borderline serous tumor). Gynecol Oncol. 1995;59:394 397. 15. Kurman RJ, Seidman JD, Shih IM. Serous borderline tumours of the ovary. Histopathology. 2005;47:310 315. 16. Leake JF. Tumors of low malignant potential. Curr Opin Obstet Gynecol. 1992;4:81 85. 17. Leake JF, Rader JS, Woodruff JD, et al. Retroperitoneal lymphatic involvement with epithelial ovarian tumors of low malignant potential. Gynecol Oncol. 1991;42:124 130. 18. Lee KR, Tavassoli FA, Prat J, et al. Tumours of the ovary and peritoneum. In: Tavassoli FA, Devillee P, eds. Tumours of the Breast and Female Genital Organs. Lyon, France: IARC Press; 2003: 119 124. 19. Longacre TA, Kempson RL, Hendrickson MR. Serous tumours of low malignant potential (serous borderline tumours): moving toward detente. Histopathology. 2005;47:315 318. 20. Longacre TA, McKenney JK, Tazelaar HD, et al. Ovarian serous tumors of low malignant potential (borderline tumors): outcomebased study of 276 patients with long-term (> or = 5-year) followup. Am J Surg Pathol. 2005;29:707 723. 21. Moore WF, Bentley RC, Berchuck A, et al. Some mullerian inclusion cysts in lymph nodes may sometimes be metastases from serous borderline tumors of the ovary. Am J Surg Pathol. 2000;24: 710 718. 22. Rao GG, Skinner E, Gehrig PA, et al. Surgical staging of ovarian low malignant potential tumors. Obstet Gynecol. 2004;104: 261 266. 23. Rollins SE, Young RH, Bell DA. Autoimplants in serous borderline tumors of the ovary: a clinicopathologic study of 30 cases of a process to be distinguished from serous adenocarcinoma. Am J Surg Pathol. 2006;30:457 462. 24. Seidman JD, Kurman RJ. Ovarian serous borderline tumors: a critical review of the literature with emphasis on prognostic indicators. Hum Pathol. 2000;31:539 557. 25. Sherman ME, Mink PJ, Curtis R, et al. Survival among women with borderline ovarian tumors and ovarian carcinoma: a populationbased analysis. Cancer. 2004;100:1045 1052. 26. Shiraki M, Otis CN, Donovan JT, et al. Ovarian serous borderine epithelial tumors with multiple retroperitoneal nodal involvement: metastasis or malignant transformation of epithelial glandular inclusions? Gynecol Oncol. 1992;46:255 258. 27. Tan LK, Flynn SD, Carcangiu ML. Ovarian serous borderline tumors with lymph node involvement: clinicopathologic and DNA content study of seven cases and review of the literature. Am J Surg Pathol. 1994;18:904 912. 28. Winter WE 3rd, Kucera PR, Rodgers W, et al. Surgical staging in patients with ovarian tumors of low malignant potential. Obstet Gynecol. 2002;100:671 676. 624 r 2006 Lippincott Williams & Wilkins