Primary mucinous ovarian tumors vs. ovarian metastases from gastrointestinal tract, pancreas and biliary tree: a review of current problematics

Both primary and metastatic mucinous ovarian tumors are characterized by a number of macroscopic and microscopic features, which may be helpful in their differential diagnosis (Figs. 1, 2 and 3). However, these are not entirely specific and a subset of metastatic tumors may sometimes mimic primary ovarian tumors [3, 8, 55, 63, 64]. In general, most primary ovarian mucinous carcinomas are large, unilateral tumors that are cystic, multilocular, with intact smooth surface, without any nodularities. Cystic and solid areas in these tumors are evenly distributed throughout the tumor. Areas with features of benign mucinous cystadenoma or MBT are common. The predominant type of invasion is usually expansile, with the infiltrative type of invasion being absent or only focal in most cases. On the contrary, metastatic tumors are often smaller, bilateral, and involve the ovarian surface or superficial cortex. The nodularity is usually visible macroscopically, however, even cases without macroscopically distinct nodules usually show these at a microscopic level, where they are characterized by aggregates (groups, nodules) of tumor cells surrounded by normal ovarian stroma, especially in the superficial cortex. Another feature which seems to be relatively specific for ovarian metastasis of LAMN is gross mucinous multinodular appearance [65]. Metastatic tumors also often show a predominantly infiltrative type of invasion. However, some metastases (especially of pancreatobiliary tract malignancies) are known to often mimic primary ovarian tumors [8, 17, 66, 67]. These tumors commonly show areas of benign or MBT appearance. Some tumors even consist only of these areas and do not show any areas of invasive growth (so-called paradoxical maturation). Metastatic tumors of pure MBT appearance are typically characterized by more pronounced nuclear atypia, and the lesions mimic an MBT with intraepithelial carcinoma. Moreover, even metastatic tumors can be large, unilateral, and even without surface involvement. According to our review, in studies in which this data was available, unilateral tumors ≥10 cm represent 15% of all metastatic tumors (29/190 cases; range 5–19.8%). These tumors are particularly difficult to diagnose as metastatic, especially in cases when they represent the first manifestation of a hitherto clinically occult primary extraovarian disease.

A summary of the morphological features is given in Table 2, although there is no single feature which would allow for a clear distinction between a primary and a metastatic tumor. For example laterality, which is cited as one of the key features when distinguishing between primary (mostly unilateral) and metastatic (more often bilateral) tumors, is by itself a highly unreliable feature. As with other parameters, historical data concerning mucinous carcinomas should be viewed with caution. For example, in a study published in 2005 analyzing the data from the Surveillance Epidemiologic and End Results (SEER) program (which included data from 22,328 women with ovarian cancer without central evaluation from the period between 1992 and 2000), MC was reported as bilateral in 355/1665 cases (21.3%) regardless of the disease stage [68]. The high percentage of primary bilateral tumors suggests that some of these tumors were probably of a metastatic origin. Based on our analysis of the data, which we performed using the data mined from 26 studies dealing with this issue, bilaterality was found in 9.4% of all primary mucinous ovarian tumors (104/1105), in 10.3% of MC (95/918) and in 49.7% of metastases (1176/2365), regardless of the primary source [3, 17, 20, 23, 34, 38, 46, 47, 55, 60‐62, 66, 67, 69‐80]. However, when evaluating colorectal cancer metastases separately, bilaterality was found in only 25.8% of cases (25/97). Still, the percentage of bilateral primary ovarian tumors (10.3% of primary MC) seems to be too high and the possibility that some studies may be biased by metastatic ovarian tumors misclassified as primary tumors cannot be excluded with certainty. Tumor laterality alone is not sufficient enough to distinguish between primary and metastatic tumors, and a careful evaluation which takes into consideration other tumor features is necessary. However, laterality is still one of the key parameters evaluated in the algorithmic approaches (usually in combination with the size of the tumor), which are discussed in detail below.

One of the features which deserves a special recognition because of its high specificity for metastatic origin is the presence of signet ring cells. Tumors containing signet ring cells are, in keeping with stringent diagnostic criteria regardless of their primary source, referred to as Krukenberg tumors. By definition, a Krukenberg tumor is a tumor in which, in addition to signet ring cells, the tumor stroma is often present in the form of a “sarcomatoid” stromal reaction [81, 82]. Krukenberg, after whom the tumor is named, hypothesized that it was an unusual type of a mucin-producing fibrosarcoma. However, subsequent studies have shown that these tumors are indeed metastatic, most commonly with a primary source in gastric cancer and less often in colorectal cancer, with other locations being implicated only rarely [83]. Stromal reaction is a reactive process which often takes on a sarcomatoid appearance, and as such can cause difficulties in differential diagnosis. However, literature data on this tumor is severely influenced by inconsistencies in the approach to its classification. The term Krukenberg tumor is sometimes used for any adenocarcinoma metastasizing to the ovary, regardless of the presence of signet ring cells and stromal response. In some studies the tumors are even termed as Krukenberg tumors regardless of their primary source. Sometimes the term Krukenberg tumor is used only for gastrointestinal tract metastases. Nevertheless, we should be aware that signet ring cells may rarely occur even in primary ovarian mucinous carcinomas, so this feature is also not entirely specific for metastastic origin [84, 85]. However, signet ring cell histology compared to non-signet ring cell histology showed a specificity of 99.7% for indicating metastastic origin (with the sensitivity reaching only 12.0%), with a positive predictive value for metastasis of 98.4% [62].

Another feature which is highly (although not entirely) specific for metastatic tumors is the presence of pseudomyxoma peritonei (PMP) [86]. The histogenesis of mucinous ovarian tumors is multifactorial, but current knowledge suggests that some tumors are of a teratomatous origin, whereas others may arise from mucinous metaplasia, probably in the form of Walthard’s nests or in Brenner tumors [87]. Tumors arising in a teratoma, which represents 3–8% of ovarian mucinous tumors, show somewhat different morphological, immunohistochemical and other features, and represent a heterogeneous group of tumors [45]. They include tumors which in many respects mimic intestinal (appendicular) tumors, most likely arising from teratomatous lower intestinal tract tissue, and tumors which are close in morphology and immunophenotype to tumors from the upper gastrointestinal or pancreatobiliary tract. Moreover, a possible origin in sinonasal-type teratoma tissue has been suggested for some intestinal type adenocarcinomas, characterized by an immunophenotype indicative of upper gastrointestinal tract (with CK7 positivity and CK20 positivity or negativity) [45]. In contrast to tumors of non-teratomatous origin, which typically show diffuse CK7 expression and a variable CK20 expression, teratomatous tumors are more heterogeneous and include CK7 negative and CK20 positive cases (which is an immunophenotype typical for colorectal carcinoma).

Another difference from tumors of non-teratomatous origin is the frequent association with pseudomyxoma peritonei in teratoma-associated lesions. In a study of 42 teratoma-associated mucinous tumors, 10 were associated with PMP (1 MA, 6 MBT, and 3 MC) [43]. However, 4 cases classified as PMP in this study showed only the presence of acellular mucin, the other 3 contained epithelial structures with features of a low-grade mucinous tumor, and the last 3 displayed carcinomatous characteristics. Changes evaluated as pseudomyxoma ovarii were also present in 8 of these tumors. An appendectomy was performed in 7 patients, all of which showed no tumor structures in the appendix. Compared to PMP associated with a low grade mucinous tumor of the appendix (LAMN), the prognosis of PMP in teratoma-associated lesions appears to be better (except in cases of peritoneal carcinomatosis associated with MC). However, the data is relatively limited in this respect. Macroscopically, the mucus of ovarian origin is much thinner compared to the thicker, jelly-like mucus of PMP associated with appendiceal tumors [88].

Generally, in the case of ovarian tumors showing morphology and immunophenotype characteristic for the lower gastrointestinal tract, one should consider the possibility of a minor teratomatous component which may have not been detected on sampling, and some authors recommend performing an extensive material re-examination, in addition to correlation with clinical data [45].

We have found 11 studies focusing on the algorithmic approach to differential diagnosis between primary and metastatic mucinous ovarian cancer [3, 13, 16, 55‐62]. Generally, the idea of an algorithmic approach is based on different gross features between primary and metastatic ovarian tumors, including especially the size and laterality. The most common approach is based on the assumption that a primary tumor is unilateral, in combination with a size criterion (≥ 10; 12; 13 or 15 cm), and any and all other tumors are considered as metastatic. In some studies this algorithm is modified by excluding tumors with signet ring cell component, which are in most cases metastatic. Additionally, some algorithms also take into account other features, including the age of the patients and some immunohistochemical results, such as PAX8, DPEP1 (dipeptidase 1; a zinc-dependent metallopeptidase involved in glutathione metabolism which is commonly expressed in colorectal carcinoma), CK7 and CDX2 expression [13, 58]. Table 3 shows our summary of the algorithms used in 9 studies, in which the available data allowed for a comparison between the primary and metastatic cases (including the accuracy, sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV)). Two studies were excluded from the analysis - one because it analyzed only metastatic cases, and the other one because it did not include the combination of size and laterality [16, 55]. In certain aspects the results of all studies are similar, in that irrespectively of the algorithm used they generally show the same trend – if the criteria for primary (metastatic) tumor are stricter, then the increased specificity is associated with decreased sensitivity. The summary of the results from all the included studies showed the following: accuracy (mean 83.5%, median 83.8%, range 71.3–96.1%, SD 5.8); sensitivity for the identification of a primary tumor (mean 84.1%, median 82.2%, range 56.3–100%, SD 14.3); specificity for the identification of a primary tumor (mean 82.8, median 81, range 72.2–94.7, SD 6.6).

In summary, algorithms by themselves are not sufficient for the definite distinction between primary and metastatic ovarian tumors, but they do provide important information which can be used in combination with microscopic features and immunohistochemical profile in order to classify most tumors with high accuracy. In one study, the suggested algorithm was intentionally optimized for high sensitivity for metastatic tumors, as the authors emphasized that a misdiagnosis of a metastatic tumor as a primary MC has greater consequences for the patient [62]. However, the achieved high sensitivity for mOC (90.1%) was associated with low specificity (59.0%). In another study, a combination of CDX2, CK7 and DPEP1 showed an accuracy of 97% (56/58 tumors) for the detection of primary MC, and accuracy of 100% (16/16 tumors) for the detection of mCRC. However, the reported accuracy for metastases from the upper GIT reached only 56% (9/16 cases) [58]. Interestingly, when the algorithm results were stratified according to the primary source of metastases, in one study the success ratio showed much better results for pancreatobiliary tract neoplasms than for colorectal carcinoma [57].

The results for all antibodies are summarized in Table 1. The statistical significance of each antibody for the differential diagnosis is summarized in Table 4. The results show that there are in fact several antibodies, which reached statistical significance when comparing primary mucinous tumors and metastatic / extraovarian tumors. However, statistical significance alone does not necessarily equate to practical usefulness. Generally, there is no single antibody upon which the decision concerning the primary source of the tumor may be based with certainty.

In the differential diagnosis between primary and metastatic ovarian mucinous tumors it is necessary to take into account the origin of the mucinous tumor because, as it has already been mentioned, some tumors arising in a teratoma may be indistinguishable from metastatic tumors on the IHC level [43, 45]. In cases of other mucinous tumors, the use of IHC depends on the differential diagnosis in question.

In the differential diagnosis with metastases from the lower GIT (colorectal and appendiceal carcinoma) a combination of CK7, CK20, CDX2, SATB2, and PAX8 is often used. Especially for CK7 and CK20 it is appropriate to evaluate not only the presence of positivity, but also its extent and the relationship of mutual expression of these 2 antibodies. Primary mucinous ovarian tumors express CK7 in about 90% of cases, and the expression is almost always diffuse (about 85% of tumors express CK7 in more than 50% of tumor cells, however, the criteria for diffuse expression vary among studies) [32, 33, 51, 89]. CK20 expression is also relatively common in primary ovarian mucinous tumors (about 65–70% of cases), but diffuse expression is found only in about 40% of cases. When evaluating the coordinate expression of CK7/CK20, regardless of its extent, primary ovarian mucinous tumors are positive for both markers in 67% of cases, CK7 positive / CK20 negative in 26% of cases, and CK7 negative / CK20 positive in only 7% of cases. However, primary mucinous tumors arising in a teratoma are CK7 negative / CK20 positive in 50% of cases [45]. CDX2 expression occurs in 49% of cases of primary tumors, but strong expression is observed in only 26% [18, 27, 90, 91]. SATB2 is a marker which is significantly more specific for colorectal cancer than CDX2 [4‐7, 10‐12, 92]. In primary ovarian mucinous tumors SATB2 is expressed only in about 8% of cases. In contrast, PAX8 expression is reported in about 35% of primary cases, although it can be relatively weak and only focal. In comparison to primary mucinous ovarian tumors, colorectal and appendiceal carcinoma metastases show CK7 expression in 31 and 26% of cases respectively, although the expression is diffuse in only 6 and 13% of cases. In contrast, CK20 is positive in 90 and 92% of cases, and almost always diffuse. CDX2 expression is positive in about 93% of colorectal and 97% of appendiceal cancers, and is almost always diffuse. SATB2 positivity is reported in 87% of appendiceal and 75% of colorectal tumors, and is almost always diffuse as well. According to our review, PAX8 is positive in about 5% of appendiceal cancers, while in colorectal cancer the expression of PAX8 has not been reported at all. Certain studies also evaluated the coordinate expression for some antibodies, which is an approach which seems to be more beneficial for the purposes of differential diagnosis. The most commonly used combination is CK7 and CK20, but combinations of other antibodies have been reported as well, including PAX8, CK7, CK20, CDX2 and SATB2. The results concerning the coordinate expression extracted from 10 studies are summarized in Table 5 [5, 9, 16, 20, 22, 38, 43, 45, 51, 93]. However, with the exception of the CK7/CK20 combination, the data is rather limited. Briefly, the coordinate expression of CK7/CK20 in ovarian metastases is significantly different compared to primary mucinous tumors. The expression of both markers is reported in about 22% of appendiceal and 11% of colorectal cancers, and most these tumors (78 and 79%) are CK7 negative / CK20 positive. According to our review, CK20 negative / CK7 positive staining pattern occurs in 0% of appendiceal tumors and about 3% of colorectal tumors. Negativity of both markers is reported in 0% of appendiceal and 6% of colorectal tumors. In summary, immunohistochemical examination can be helpful in the differential diagnosis between primary ovarian mucinous tumors and metastases from the “lower” GIT, but the use of a panel of antibodies and the correct interpretation of their results is crucial.

Metastases from the pancreatobiliary tract represent probably the most problematic category with regard to possible confusion with primary ovarian tumors [8, 17, 66, 94]. Regarding the IHC examination, the antibodies listed in the differential diagnosis between primary ovarian mucinous tumors and metastases of the “lower” GIT are practically useless in the context of pancreatobiliary neoplasms. The expression of CK7 and CK20 is almost identical in both groups of tumors. Although the expression of CDX2 is slightly more common in primary ovarian mucinous tumors, in practice this difference is difficult to utilize (49% positivity in primary ovarian tumors vs. 19% in pancreatic cancers and 31% in pancreatobiliary cancers). SATB2 expression is not reported in pancreatobiliary tumors; however, its expression in primary ovarian mucinous tumors is rare (7%), which significantly limits its practical use. Of the other markers, the expression of PAX8, DPC4, and CK17 seems to be useful for the purposes of differential diagnosis. PAX8 is reported to be positive in 36% of primary ovarian mucinous tumors and in 4% of pancreatobiliary tumors. However, the sensitivity of this marker is low. A loss of DPC4 (SMAD4) expression occurs in 53% of pancreatic and only about 5–10% of primary ovarian mucinous tumors [28, 36, 37, 95]. According to the literature, cytokeratin 17 seems to be negative in ovarian mucinous tumors and positive in 27–83% of metastatic pancreatic carcinomas, but the data is very limited [94, 96].

The most problematic (on the IHC level) is the distinction between primary ovarian mucinous tumor and gastric adenocarcinoma metastasis [75, 97, 98]. Morphologically, some of the metastases have a signet ring cell component, or consist only of these elements, which is a feature strongly suggestive against the diagnosis of a primary ovarian tumor. However, it is very complicated to make the distinction between primary and metastatic tumors based on IHC results. The expression of CK7, CK20, CDX2 and SATB2 is very similar in both groups and therefore these markers cannot be used. Practically useful antibodies are mainly represented by PAX8 (positive in 35% of primary mucinous tumors vs. 0% of gastric cancers) and CA125 (positive in 24% of primary mucinous tumors vs. 0% of gastric cancers). However, these antibodies are only significant when the staining result is positive and their sensitivity is low. A minority of primary ovarian mucinous tumors (< 10%) may also weakly express ER and PR. The expression of ER/PR is not reported in gastric cancer, but again it is a marker with very low sensitivity.

The frequency of ovarian metastases from the GIT may vary, but according to the literature ovarian metastases occur during the course of the disease in 2% of patients with colorectal cancer and 2.9% of patients with gastric cancer [69, 75, 78, 97, 99‐103]. Concerning ovarian metastases, we have analyzed the data from 13 studies, which were used to mine the information on primary source [3, 28, 46, 60, 80, 104‐111]. Moreover, 5 of these studies also provided data concerning the percentage of metastatic ovarian cancer from all (primary and metastatic) ovarian tumors. Based on these 5 studies, we were able to determine that out of the 14,060 cases of ovarian cancers 656 cases (4.7%) were metastases. Based on all of the 13 studies, the most common primary source was colorectal carcinoma (32%), followed by breast carcinoma (15.4%), endometrial carcinoma (12.9%), gastric carcinoma (9.2%), appendiceal carcinoma (6.7%), uterine cervix carcinoma (2.4%), pancreatic carcinoma (2.2%), small intestine carcinoma (1.6%), and carcinoma of the gallbladder and biliary tract (1.5%). Other tumors such as lung carcinoma, skin tumors, kidney cancer, and esophageal carcinoma each accounted for less than 1%. For the entire group of metastatic tumors the primary source was unknown in 12.2%.