Prior appendectomy does not protect against subsequent development of malignant or borderline mucinous ovarian neoplasms

Background— Due to concern that mucinous malignant or borderline ovarian neoplasms (MON) may represent metastatic deposits from appendiceal primaries, gynecologic oncologists routinely perform appendectomy in these cases. However, a multidisciplinary critique of this practice is lacking. Methods— The New England Case-Control study database was utilized to compare the effect of prior appendectomy against known risk factors for MON. Pathology and operative reports of local cases of MON were reviewed to estimate the frequency of microscopic mucinous lesions in the appendix. Protein expression patterns among mucinous ovarian, colorectal, and appendiceal cancers were compared by immunohistochemistry. Results— From the New England Case-Control study, 287 cases of MON were compared against 2,339 age-matched controls. Prior appendectomy did not reduce the risk of MON (OR 1.28, 95% CI 0.83–1.92, p=0.23), while prior tubal


Introduction
Epithelial ovarian carcinomas (EOC) are the leading cause of death among gynecologic tumors [1]. EOC are histologically classified into four major subtypes: serous, clear cell, endometrioid, and mucinous [2,3]. Mucinous ovarian carcinomas have been the least studied of these, probably because of their relative rarity, comprising about 3% or less of EOC [4]. Mucinous tumors can exist as both invasive and borderline tumors, here collectively referred to as mucinous ovarian neoplasms (MON). Although it has been argued that MON bear some relationship to the endocervix, the mucinous epithelium that characterizes MON more frequently resembles gastrointestinal (GI) epithelium [5]. Even when excluding cases of pseudomyxoma peritonei, which are now generally accepted to occur almost exclusively in association with appendiceal primaries, most pathologists still maintain that the diagnosis of primary MON requires consideration and exclusion of metastases from other GI carcinomas [6,7]. Indeed, the epidemiology, histology, and molecular biology of MON are routinely compared to GI mucinous carcinomas, in particular those arising in the colon [8]. Coupled with rare case reports of goblet cell carcinoids ("adenocarcinoid" tumors) presenting as isolated adnexal masses, these reports have advanced the notion that a significant proportion of MON are subsequently found to have arisen from an occult appendiceal or other GI primary, and therefore that the appendix should be routinely removed at the time of surgery for any malignant or borderline MON [9][10][11][12]. In addition, other authors have advocated routine appendectomy in all EOC cytoreductive surgeries regardless of histology to exclude isolated metastases from the ovary to the appendix [13][14][15]. As a result, routine appendectomy at the time of surgery for a suspected or confirmed (by frozen section) MON of malignant or borderline potential has become common.
In this study, we examine the relationship between malignant and borderline MON and mucinous appendiceal tumors. We test the idea that some seemingly isolated MON are actually derived from the appendix by using a large regional case-control study to compare the effect of prior appendectomy against established risk factors for EOC. We then report our recent clinical experience with regards to the issue of occult appendiceal primaries at the time of surgery for suspected malignant or borderline MON and microscopic metastases from MON to the appendix. Finally, we compare the immunohistochemical (IHC) pattern of mucinous ovarian carcinomas (MOC) to mucinous appendiceal and colorectal carcinomas to test the ability of pathologists to discriminate among these clinical entities.

New England Case-Control Study
Data derived from four phases of a case-control study of ovarian cancer, the New England Case-Control (NECC) study, were used [16,17]. Cases were enrolled from 7/1984 -9/1987 (NECC2), 5/1992 -3/1997 (NECC3), 8/19988/ -4/2003, and 10/2003and 10/ -11/2008. Data from an earlier phase between 1978 and 1981 (NECC1) were no longer available electronically and not included. NECC2 identified ovarian cancer cases from ten hospitals in Boston; NECC3, 4, and 5 used statewide cancer registries and tumor boards to identify cases diagnosed in Eastern Massachusetts and the State of New Hampshire. The four phases enrolled 2,475 cases including 2,274 with epithelial ovarian cancers, of which 287 were mucinous. Controls for NECC3 were identified by random-digit dialing supplemented with residents' lists for older controls. About 10% of households dialed had an eligible control and of these, 421 (72%) agreed to participate. All controls for NECC2, 4, and 5 were identified through town residents' lists in Massachusetts and Driver License Registries in New Hampshire. Of 5,151 potential controls identified through town books in all phases, 1,671 were ineligible due to bilateral oophorectomy, 1,562 declined participation, and 1,918 were enrolled. In total, 2,339 controls were enrolled. This study is approved by the Brigham and Women's Hospital and Dartmouth Medical Center Institutional Review Boards.

Chart review
The medical charts of 106 patients from the NECC study population were available for review by virtue of being operated on at either Brigham and Women's Hospital or Massachusetts General Hospital. These were combined with an additional 64 patients operated on at Brigham and Women's from 2006-2011 not enrolled in NECC. Operative reports and pathology reports from these cases were read to determine the frequency of appendectomy at the time of surgery and the incidence of microsopic metastases to the appendix from the ovary. In addition, the medical charts of an additional 26 patients operated on by the gynecologic oncology service at Brigham and Women's Hospital for an adnexal mass with the subsequent finding of a GI primary cancer were assessed by a member of the gynecologic oncology division (KME) for the frequency of a microscopic GI primary that would have been diagnosed only by routine appendectomy and not by examination of the other pathologic specimens obtained via oophorectomy or other cytoreductive procedures. Chart review was approved by the Partners Healthcare Institutional Review Board.

Statistical methods
For the case control study, continuous variables were categorized based on quartiles of the control distributions. Subjects with missing exposures were excluded on an exposurespecific basis. Unconditional logistic regression models were used to assess the associations between exposures and MON. All models were adjusted for the matching factors (age, study site, study phase), as well as parity, breastfeeding, OC use, genital talc exposure, Jewish ethnicity, and tubal ligation. All analyses were performed with SAS (SAS Institute, Cary, NC). Immunohistochemical staining results were compared using Fisher's exact test (GraphPad Software, La Jolla, CA).

Pathology samples
A total of 65 cases of mucinous tumors were selected from the surgical pathology files of the Brigham and Women's Hospital including 21 mucinous ovarian carcinomas, 18 mucinous colon carcinomas (MCCs), and 26 mucinous appendiceal carcinomas (MACs). Sites of origin of all tumors were known before this study on the basis of clinical and radiologic information in addition to surgical specimen examination. The tissues had been routinely fixed in 10% neutral formalin and embedded in paraffin. At least one paraffin tissue block with tumor was selected from each case by a pathologist with expert training in gynecologic or gastrointestinal pathology (MSH or JLH and LAD, respectively).

Risk factors for mucinous ovarian neoplasms
The New England Case-Control (NECC) study included 287 cases of MON and 2,339 agematched controls (Table 1). Among the MON, 169 cases were borderline tumors and 118 cases were invasive adenocarcinomas. There was no protective effect against MON conferred by appendectomy either overall (OR 1.28, 95% CI 0.83-1.92, p=0.23) or in the borderline (OR 1.41, 95% CI 0.83-2.39, p=0.2) or invasive adenocarcinoma (OR 1.12, 95% CI 0.63-1.99, p=0.69) patient subsets (Tables 1, S2, and S3). Among the invasive carcinomas, there was no relationship between prior appendectomy and stage (Table S4). In contrast, prior tubal ligation (OR 0.65, 95% CI 0.43-0.99, p=0.05), parity (OR 0.56, 95% CI 0.40-0.79, p=0.0009), and breastfeeding (OR 0.62, 95% CI 0.47-0.83, p=0.002) were each associated with a lower risk of MON. Compared to controls, MON cases were more likely to be current smokers (OR 2.04, 95% CI 1.48-2.80, p<0.0001). A dose-dependent effect was seen with smoking (p-trend <0.0001), with the highest risk of MON occurring among women with a smoking history of more than 28 pack years (OR 2.68, 95% CI 1.85-3.89, p<0.0001). In contrast to other types of epithelial ovarian neoplasm, there was no association between MON and either a personal or family history of breast cancer [18]. Similarly, the protective effect of oral contraceptive use against MON only reached borderline significance (p-trend=0.06).

Frequency of microscopic appendiceal GI tumors as isolated ovarian neoplasms
All gynecologic oncology cases from 2006-2011 were reviewed, and 26 cases were surgeries for a pelvic mass subsequently identified to be a mucinous GI tumor metastatic to the ovary (Table 2). Among these, 8 were appendiceal tumors, 8 were colorectal tumors, 8 were of unclear GI origin, and 1 tumor each was of gastric or small bowel origin. All 26 cases presented with widely metastatic disease. Among these, only 1 tumor (3.8%) was felt to have potentially arisen from a microscopic GI primary. In that case, the appendix had been removed seven years prior. Both on the original pathology and on review seven years later, a mucinous cystadenoma was seen in the appendix but no evidence of carcinoma. Even so, the patient's widely metastatic tumor was histologically consistent with an appendiceal primary, not a MON. Additionally, among the 26 cases there was one case of an incidental carcinoid tumor, not related to the patient's colorectal primary, and one case of appendiceal cancer where the appendix was fused to the ovary and thus not grossly identifiable at the time of surgery. There were 9 cases (34.6%) classified as GI primaries in part due to signet-ring cell morphology in either the adnexal lesion or metastatic implants from elsewhere in the pelvis.

Frequency of microscopic metastases from MON to the appendix
As another possible indication to remove the appendix at the time of surgery for MON would be to exclude unsuspected metastases from the ovary to the appendix, we also evaluated the frequency of microscopic, isolated appendiceal metastases from the ovary (Table 3). 170 cases of MON were reviewed, 106 from the NECC study population and 64 additional patients operated on at Brigham and Women's Hospital but not enrolled in the NECC study. 91 cases involved borderline mucinous tumors and 79 cases involved invasive mucinous carcinomas. Pathology reports for the appendix were available in 72 (42.4%) cases. Among the appendectomy specimens, there was 1 subcentimeter carcinoid tumor and 3 cases of superficial metastases from the ovarian lesion to the appendix, all in the setting of gross metastatic disease. All other appendectomy specimens were benign. No isolated microscopic metastases to the appendix or microscopic mucinous appendiceal adenocarcinomas in appendectomy specimens were observed.

Discussion
In this study, we systematically examined the relationship between malignant and borderline mucinous ovarian neoplasms (MON) and mucinous adenocarcinomas of the appendix using epidemiological, clinical, and immunohistochemical means. This investigation stemmed from the longstanding tradition of removing the appendix at the time of surgery for suspected mucinous ovarian neoplasms based on the assumption that some unknown proportion of apparent MON are due to an occult appendiceal lesion. While our study is based on histological specimens removed at the time of surgery after a final pathological diagnosis had been determined, our findings agree with the findings of three other recent studies, which together identified no cases of an occult appendiceal mucinous adenocarcinoma in a grossly normal appendix out of 476 appendectomies performed at the time of surgery for a mucinous ovarian neoplasm [20][21][22]. Even so, as indicated by the fact that 42.4% of patients in our study had an appendectomy at the time of surgery, the procedure is still frequently performed.
While the previous studies also used cases series to explore the relationship between MON and the appendix, this is the first study to our knowledge to look for an epidemiologic connection between the two. Were MON mostly or even frequently derived from the appendix, one would expect that having had a prior appendectomy would offer some protection against the development of MON. In this study we found no such association. In addition, we found no protective effect from hysterectomy either, arguing against the uterus or endocervix as other putative sites of MON precursors [23]. Instead, factors known to be protective against the development of other forms of EOC, namely parity, tubal ligation, and breastfeeding, appear to be the most protective against MON. We also confirmed previous reports which have shown a relationship between smoking and MON [24-26].
Despite the epidemiologic evidence distinguishing MON from mucinous appendiceal adenocarcinomas, routine appendectomy has been the practice at our institution as in many other centers. In this study, we show that the frequency of an occult appendiceal tumor masking as an overt MON is extremely rare. Similarly, cases of microscopic metastases from a MON to the appendix were also uncommon. Removal of the appendix is warranted when there is gross involvement of the organ clinically, or when widely metastatic disease makes the organ of origin uncertain. It is also essential in cases of pseudomyxoma peritonei as these are almost always of appendiceal origin. While it seems unlikely that removal of a grossly normal appearing appendix will reveal a microscopic focus of an unsuspected mucinous appendiceal primary metastatic to what otherwise appears to be a Stage IA mucinous ovarian neoplasm or borderline tumor, thorough evaluation of the appendix at the time of surgery is still recommended.
In cases where the appendix was not removed at the time of surgery for suspected MON, immunohistochemical staining may help distinguish ovarian from appendiceal or colorectal carcinomas. Mucinous ovarian carcinomas tend to be CK7+/CK20-/MUC2-/CDX2-, with variable expression of PAX8. In contrast, mucinous colorectal cancers and mucinous appendiceal cancers usually feature a CK7-/CK20+/MUC2+/CDX2+/PAX8immunophenotype, with nuclear staining for β-catenin potentially more common among colorectal adenocarcinomas. Even so, no single marker should be relied upon to make the diagnosis due to the overlap in staining patterns, although the presence of signet-ring cell morphology favors a primary GI rather than ovarian neoplasm [27]. In addition, molecular studies have shown that cases of pseudomyxoma peritonei are almost never of ovarian origin [28-31].
Finally, while malignant and borderline MON seem to be distinct from appendiceal tumors, MON do display several characteristics that distinguish them from other EOC. Unlike other EOC, there appears to be no association between MON and either a personal or family history of breast cancer and only a weak relationship between MON and oral contraceptive use. Moreover, PAX8 staining, which characterizes most Müllerian malignancies, is less common among MON. These findings suggest that MON has a unique pathogenesis. Future studies looking at MON precursors may help clarify whether they have a unique origin.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

Research Highlights
• Prior appendectomy is not protective against subsequent mucinous ovarian neoplasms.
• Occult mucinous tumors in the appendix at the time of surgery are rare events.
• Immunohistochemistry may help resolve the origin of some mucinous neoplasms Immunohistochemical analysis demonstrated that MOCs were typically positive for CK7 and negative for CK20, although rare cells were occasionally positive for CK20, as shown. CDX2 and MUC2 were typically negative in MOCs. PAX8 was weakly positive in a small subset (~24%) of MOCs. The majority of MCCs were diffusely positive for CK20, CDX2, and MUC2 and negative for CK7 and PAX8 (red arrows mark immunoreactive lymphocytes which act as an internal positive control). MACs demonstrated a similar immunoprofile compared to MCCs, with the exception of nuclear staining for β-catenin.    Expression of immunohistochemical markers in different mucinous carcinomas