Adenocarcinomas metastatic to the liver

The value of cytokeratins 20 and 7 in the search for unknown primary tumors


  • Tibor Tot M.D., Ph.D.

    Corresponding author
    1. Department of Pathology and Clinical Cytology, Central Hospital, Falun, Sweden
    • Central Hospital, Department of Pathology, S 79182 Falun Sweden.
    Search for more papers by this author



Metastatic adenocarcinoma in the liver with an unidentified primary tumor site is a common clinical problem. Pathologists often are asked to identify the primary tumor site. The histologic picture itself usually is not helpful, because the histology may be similar in the metastases of tumors with different primary localizations. Immunohistochemistry can be helpful, but the previously recommended antibody panels are too complicated for everyday use.


A simple immunohistochemical algorithm with two monoclonal cytokeratin (CK) antibodies, CK20 and CK7, was tested on 93 autopsy cases of adenocarcinomas metastatic to the liver. Sections of the liver metastases were stained automatically and evaluated as negative (no staining), focally positive, or diffusely positive. Statistical comparison of the staining results for a single antibody was calculated as an odds ratio.


Thirty-six of 93 (39%) metastases proved to be CK20 positive (+). In this group, the CK20+/CK7 negative (−) pattern was highly characteristic for colorectal localization of the primary tumor, having been observed 17 of 21 of the cases (81%). The CK20+/CK7+ pattern of the metastatic liver adenocarcinomas was highly suggestive of primary localization in the pancreas or biliary tract (11 of 14 cases; 79%). Exclusion of the tumors originating in the stomach raised these values to 94% and 92%, respectively. The statistically calculated predicted probability of primary tumor site being in the colon or rectum for CK20+/CK7− metastasis was 78,41%, the probability of a primary tumor being located in the pancreas or biliary tract was 74,85%, if calculated for the whole study group.


The tested simple algorithm proved to be useful in CK20 positive (+) cases, predicting a primary tumor localization in the colon, rectum, pancreas, or biliary tract with high accuracy. The CK20− group was too heterogeneous to be classified adequately by these two antibodies. Cancer 1999;85:171–7. © 1999 American Cancer Society.


The author thanks Dr. H.H. Chen, Graduate Institute of Epidemiology, College of Public Health, National Taiwan University, for help with statistics and Gösta Andersson for the photography.

Metastatic adenocarcinoma is the most common type of liver tumor by far. The most common sources of these metastases are primary lung, breast, and colorectal carcinomas.1 As many as 60% of all metastatic adenocarcinomas are of unknown origin at the time of initial clinical presentation.1, 2 Consequently, the search for an unknown primary tumor is a common clinical problem. Extensive radiologic and endoscopic examinations are expensive, inconvenient, time consuming, and often unsuccessful.2 Any clue about the possible primary tumor site that the pathologist can report will be beneficial.

Several studies of the immunohistologic phenotypes of the primary adenocarcinomas have shown the site specificity of some single tumor markers, such as prostate specific antigen3 and thyroglobulin.4 Other single markers, (for example, carcinoembryonic antigen [CEA]), fail to express the expected high site specificity.5 On the other hand, there are very few studies dealing systematically with the immunophenotyping of metastases. Brown et al.6 identified a panel of eight markers as a useful diagnostic adjunct in the determination of the primary site. However, in everyday clinical practice, a simpler approach is needed.

The development of monoclonal antibodies against various cytokeratin (CK) polypeptides has opened new perspectives in studying the normal and neoplastic epithelial cells. There are extensive reports of the patterns of reaction in almost all types of primary carcinomas.7–9 CK20 seems to be a very useful marker for distinguishing gastrointestinal adenocarcinomas from adenocarcinomas of other primary tumor sites.10 On the other hand, CK7 has been shown to be present in a large number of adenocarcinomas but not in colonic carcinomas.1 To test a simple, two-step algorithm using these two antibodies, we studied 93 liver metastases in autopsy cases with verified primary tumor sites. Our working hypothesis was to group the liver metastases according to CK20 positivity/negativity as the first step of the single algorithm and, in the second step, to distinguish the tumors of the CK20 positive (+) group by CK7 reactivity. The same approach has been used by Wauters et al. in a study of carcinomas metastatic to the ovary,12 by Loy and Calaluce13 studying pulmonary and colorectal adenocarcinomas, by Loy et al.14 in a study of a large series of primary ovarian tumors and ovarian metastases from colonic carcinoma, by Berezowski et al.15 in a study of ovarian and colonic carcinomas, and by Sack and Roberts16 on fine-needle aspirates and effusions. To our knowledge, expression of CK20 and CK7 in liver metastases from autopsy-verified primary adenocarcinomas has not been studied previously.


Case Selection

All 133 cases of adenocarcinoma with liver metastases from the autopsy registry of the Department of Pathology and Clinical Cytology in Falun during the 10 calendar years 1987–1996 were reviewed retrospectively. Those with unequivocal autopsy findings and sufficient paraffin-embedded tissue both from primary tumor site and from liver metastasis were included in this study. Any doubt about the site of the primary tumor in the original autopsy report or on review was a criterion for exclusion of the case. Cases with more than one possible primary localization and a few cases with very advanced autolytic changes also were excluded. The remaining 93 cases formed the study material. The primary tumors were located in the colon (17 cases), rectum (6 cases), stomach (10 cases), pancreas (23 cases), gallbladder or extrahepatic bile ducts (17 cases), breast (10 cases), prostate (7 cases), and ovary (3 cases). Primary tumors were defined according to tumor site. Type of adenocarcinoma or tumor grade was not analyzed in relation to immunohistochemical findings.


Three-micrometer-thick sections from the selected original paraffin blocks were cut and rehydrated as usual. Immunohistochemical staining was carried out in the Ventana ES automated Immunohistochemistry System (Ventana Medical Systems Inc., Tucson, AZ) using original Ventana reagents, with the exception of the primary antibodies (CK20, clone Ks 20.8, and CK7 clone OV-TL 12/30; both from DAKO A/S, Glostrup, Denmark). Antigen retrieval was performed for both primary antibodies: overnight incubation in 0.1 M citric acid, pH 7.2, at 65°C (a standard procedure in our laboratory) for CK20 and 8 minutes of protease digestion in the Ventanas system using the original protease 1 for CK7. The Ventana system uses biotinylated secondary goat antimouse antibody for the detection system and streptavidin-horseradish peroxidase conjugate for visualization of diaminobenzidine (DAB) solution. Endogenous biotin activity was blocked with a solution of streptavidin. CK20 was used at a dilution of 1:100 with a reaction time of 32 minutes, whereas CK7 was used at the same dilution but with a reaction time of 28 minutes. The slides were weakly counterstained with hematoxylin and were mounted routinely. For CK20, a positive control was used (a section of rectal adenocarcinoma), and the hepatocytes that were present in all of the slides (as selected) served as an intrinsic negative control. Intrahepatic bile ducts served as an intrinsic positive control for CK7, and, in the majority of cases, hepatocytes could be used as a negative intrinsic control for this antibody as well.

Evaluation and Statistics

All of the slides were evaluated by the author, and the immunostaining was assessed semiquantitatively. No staining in the tumor cells was considered to be negative (−), staining in less than 10% of tumor cells was considered to be focally positive, and staining in more than 10% of tumor cells was considered to be diffusely positive staining (+). For statistical analysis, one group (I) with any degree of positive staining (diffuse or focal) and another group (II) with negative staining were formed, and the tumors were grouped as follows: metastases from the colon and rectum (group A), from the stomach (group B), from the pancreas and gallbladder (group C), and the rest (group D) (Fig. 1).

Figure 1.

A chart showing the distribution of the primary sites of cytokeratin 20 positive (CK20+) liver metastases. CK7, cytokeratin 7.

Statistical comparison of the staining results for a single antibody was calculated as an odds ratio, with group A as a reference group for CK7 and group D as a reference group for CK20. The statistical significance was tested with a 95% confidence interval. The predicted probability of the liver metastases originating from groups A, B, C, and D was calculated as follows:

equation image

P values were calculated for antibodies, and for their combination with regard to positive (1) and negative (−1) staining.


Assessment of the immunostaining for CK20 was easy because of the distinct, high quality staining and because of the negative reaction of the liver structures. A nondistinct, weak staining of the hepatocytes was present in only three cases. CK7 staining usually was even more distinct, but the intrahepatic bile duct epithelium, which was often proliferating and intermingled with the tumor cells, caused some difficulties in interpretation (Fig. 2). Occasionally, periportal hepatocytes were stained positively. Only a distinct staining in the cells of the tumors (often not as intense as in bile ducts) was interpreted as positive.

Figure 2.

Proliferated bile ducts were stained distinctly with cytokeratin 7 (CK7). There was no staining in the tumor tissue.

The results are presented in Tables 1–4. Most of the metastases from the colon and rectum (17 of 23; 74%) showed a diffuse positivity for CK20. At the same time, only one metastasis from these tumor sites stained positively for CK7.

Table 1. Cytokeratin 20 Reactivity in Metastatic Adenocarcinomas of the Liver
Primary tumor siteDiffusely positive % (no.)Focally positive % (no.)Negative % (no.)
Colon65 (11 of 17)6 (1 of 17)29 (5 of 17)
Rectum83 (5 of 6)0 (0 of 6)17 (1 of 6)
Stomach30 (3 of 10)20 (2 of 10)50 (5 of 10)
Pancreas22 (5 of 23)17 (4 of 23)61 (14 of 23)
Gallbladder12 (2 of 17)12 (2 of 17)76 (13 of 17)
Breast0 (0 of 10)0 (0 of 10)100 (10 of 10)
Prostate0 (0 of 7)0 (0 of 10)100 (7 of 7)
Ovary0 (0 of 3)33 (1 of 3)66 (2 of 3)
Total28 (26 of 93)11 (10 of 93)61 (57 of 93)
Table 2. Cytokeratin 7 Reactivity in Metastatic Adenocarcinomas of the Liver
Primary tumor siteDiffusely positive % (no.)Focally positive % (no.)Negative % (no.)
Colon6 (1 of 17)0 (0 of 17)94 (16 of 17)
Rectum0 (0 of 6)0 (0 of 6)100 (6 of 6)
Stomach60 (6 of 10)0 (0 of 10)40 (4 of 10)
Pancreas87 (20 of 23)0 (0 of 23)13 (3 of 23)
Gallbladder82 (14 of 17)18 (3 of 17)0 (0 of 17)
Breast70 (7 of 10)30 (3 of 10)0 (0 of 10)
Prostate29 (2 of 7)0 (0 of 7)71 (5 of 7)
Ovary100 (3 of 3)0 (0 of 3)0 (0 of 3)
Total57 (53 of 93)6 (6 of 93)37 (34 of 93)
Table 3. Immunophenotype of the Metastatic Adenocarcinomas According to Primary Site
  1. CK: cytokeratin.

Table 4. The Predicted Probability of Primary Localization of the Adenocarcinomas According to the CK20/CK7 Staining Pattern of the Liver Metastases
Staining patternGroup A (%)Group B (%)Group C (%)Group D (%)
  1. CK: cytokeratin; group A: metastases from the colon and rectum; group B: metastases from the stomach; group C: metastases from the pancreas and gall bladder; group D: metastases from the breast, prostate, and ovary.


Metastatic adenocarcinomas from the stomach showed a variable immunostaining pattern: Half of them (5 of 10; 50%) were stained positively for CK20 (three diffusely and two only focally), 6 of 10 (60%) were stained positively for CK7, and 4 of 10 (40%) were negative. Although the large majority (20 of 23; 87%) of metastases from the pancreas were CK7+, there was a considerable variation in the CK20 staining of these metastases: Five of 23 (22%) were diffusely positive, 4 of 23 (17%) were focally positive, and 14 of 23 (61%) were negative. Only 4 of 17 (24%) of the gallbladder tumors were CK20+ (half of them only focally), but all of them showed CK7 positivity (most of them diffusely). Metastatic breast, prostatic, and ovarian adenocarcinomas were generally negative for CK20 staining, with the exception of one of the ovarian carcinomas, which was stained weakly and focally. The breast and ovarian metastases stained diffusely and strongly for CK7, whereas 5 of 7 (71%) prostate tumor metastases showed no staining for this antibody.

Statistical testing of the association between CK7 positivity and the origin of metastases gave odds ratios of 66.00 for group B, 271.35 for group C, and 33.00 for group D, with the staining results for group A regarded as 1.00. This means, for example, that a CK7+ liver metastasis is 271 times more likely to originate in the pancreas or gall bladder than in the colon or rectum. All of the associations reached statistical significance. A similar calculation for CK20, with group D used as the reference group, resulted in the following odds ratios: 53.83 for group A, 19.00 for group B, and 9.15 for group C (all statistically significant).

Table 3 summarizes the two staining patterns by assessing the positivity of immunostaining without distinction between diffuse and focal staining. The liver metastases of primary colorectal adenocarcinomas showed a typical CK20+/CK7− pattern in 17 of 23 (74%) cases. Four of /6 of the CK20− colorectal adenocarcinomas were moderately differentiated, whereas two were poorly differentiated small cell adenocarcinomas. The metastases of the pancreaticobiliary tumor group showed two major patterns: Twenty-six of 40 (65%) were CK20−/CK7+, and 11 of 40 (28%) were CK20+/CK7+. Breast carcinoma metastases invariably showed a CK20−/CK7+ pattern. On the basis of the considerable variation in the patterns of staining of the metastatic adenocarcinomas from the stomach, this algorithm was not effective in their detection.

The predicted probabilities of the primary source of liver tumor metastases on the basis of the combination of the two antibodies are presented in Table 4. Testing the whole study group, the probability that a CK20+/CK7− metastasis originated from the colon or rectum was 78.41%. This probability for a CK20+/CK7+ metastasis was 3.81%. The probability that a metastasis was of pancreaticobiliary origin (group C) was 7.24% in the CK20+/CK7− cases and 74.85% in the CK20+/CK7+ cases. The calculated predicted probabilities in CK20− groups could not verify a clear association between the staining patterns and the groups of primary tumors: Relatively high probability was found for groups C and D to be associated with the CK20−/CK7+ pattern and for groups A and D to be associated with the CK20−/CK7− pattern.

In CK20+ tumors, as demonstrated in Figure 1, in the group of CK20+/CK7− liver metastases, 17 of 21 (81%) had their origin in the colon or rectum. In the CK20+/CK7+ subgroup, 11 of 14 (79%) originated in the pancreas or the gallbladder. If primary gastric carcinoma was excluded, then the CK20+/CK7− phenotype could indicate the colorectum as the primary site in 17 of 18 (94%) cases, and the CK20+/CK7+ phenotype could indicate a primary tumor in the pancreas or biliary tree in 11 of 12 (92%) of cases. Considering the CK20+/CK7− phenotype of the metastases as “colorectal” and the CK20+/CK7+ phenotype as “pancreaticobiliary,” we could correctly detect the primary site in 31 of 36 (86%) of CK20+ liver metastases. The CK20− group proved to be too heterogeneous for any successful detection.


Our autopsy material, like the material of many other pathology departments, does not reflect the actual epidemiologic situation. There is preselection by a low autopsy rate and by local differences in health care organization, and it is influenced by the professional interest of the clinicians. Many cancer cases with a definite diagnosis during the patient's life time are lost to the autopsy service, whereas other cases with tumors that are difficult to diagnose by using clinical methods are overrepresented. This explains the relatively large numbers of pancreaticobiliary carcinomas seen in our material. On the other hand, lung carcinomas are treated only preoperatively in our hospital and are not represented in this study.

CK20 labels the normal intestinal epithelium, the gastric foveolar epithelium, the umbrella cells of the urothelium, and epidermal Merkel cells.17 This antibody labels the majority of adenocarcinomas of the colon, mucinous ovarian tumors, and transitional and Merkel cell carcinomas, and it is often positive in adenocarcinomas of the stomach, bile duct, gallbladder, and pancreas.10 Adenocarcinomas of the breast, lung, and endometrium and nonmucinous tumors of the ovary are essentially negative.10 The most frequent CK20+ carcinoma type is adenocarcinoma of the colon and rectum; however, as the current study shows, a considerable proportion of colorectal carcinomas do not react with CK20. Although CK20 positivity (particularly when it is coupled with CK7 negativity) speaks strongly in favor of colorectal origin, CK20 negativity (with or without CK7 negativity) does not necessarily rule out this possibility: CK20 positivity has been reported in lung adenocarcinomas.13, 18

CK7 is known to label several types of normal and neoplastic glandular epithelia; carcinomas from the gastrointestinal tract and prostate generally are negative, whereas breast, lung, and subtypes of adenocarcinomas of the ovary generally are positive.11, 14

CK 20 does not label normal hepatocytes or normal biliary epithelium. In only three cases from the current series was weak staining observed in hepatocytes. In our opinion, CK20 positivity, especially if it is diffuse and distinct, is a marker of the metastatic nature of an adenocarcinoma in the liver. On the other hand, CK7 is positive in cholangiocellular and mixed hepatocholangiocellular carcinomas,19 and, consequently, a CK7+ staining provides no information about the metastatic or primary character of the tumor.

CK20 and CK7 have been used for immunophenotyping of metastases and primary adenocarcinomas a the study by Wauters et al.12 Their study is directly comparable with the current work because of the use of precisely the same antibodies and only minute differences in the technical procedure (different methods of antigen retrieval were used; there are no data available about the automatic immunostaining in the study by Wauters et al.). It is not clear whether the cited study was performed on autopsy-verified cases or on cases that were only biopsied. In the study by Wauters et al., CK20 positively labeled 14 of 16 colon adenocarcinomas (87% compared with our 74%), and, in their tumors, CK7 negativity was registered in 12 of 16 (75%) cases compared with our 22 of 23 (96%) cases. Both studies registered a similar variability in the staining pattern of the adenocarcinomas of the stomach. No pancreatic or biliary carcinomas were included in their study. Loy and Calaluce13 studied 151 colonic and pulmonary adenocarcinomas and found the CK20+/CK7− pattern in 77% of colonic carcinomas and 0% pulmonary carcinomas, whereas the CK20−/CK7+ pattern was not found in colonic carcinomas but was found in 86% of pulmonary adenocarcinomas. Loy and Calaluce advocated the same combined approach that was used in this study: Neither CK7 immunostaining nor CK20 immunostaining used alone could separate these tumors reliably.

The CK20+/CK7+ pattern is not unique for pancreaticobiliary adenocarcinomas. Sack and Roberts16 reported metastatic transitional cell carcinomas with this immunphenotype, and Loy et al.14 reported 27 of 45 serous and 27 of 40 mucinous ovarian tumors exhibiting the CK20+/CK7+ pattern. In the current study, we also verified a case of CK20+/CK7+ metastasis of ovarian carcinoma. Loy and Calaluce13 reported CK20+/CK7+ cases of lung adenocarcinomas, 6% (3 of 50) in primary tumors and 13% in metastatic tumors. No hepatic metastases appeared in their series. Although transitional cell carcinomas and serous papillary tumors usually are recognizable on histological basis, the differential diagnoses between a metastatic pancreaticobiliary and mucinous ovarian carcinoma can be very difficult. The current study focused on metastatic adenocarcinomas, and no cases of primary cholangiocellular liver carcinoma were included, because, as a kind of ”biliary cancer,” it can complicate further the differential diagnosis of CK20+/CK7+ adenocarcinomas.

More complex algorithms with which to search for unknown primary tumors have been developed.6, 20 Brown et al.6 reported a 66% overall accuracy using a panel of four antibodies (CEA, CA19-9, CA125, and BCA225). The two-step algorithm used in the current study is simple enough to become a routine procedure in every case of liver biopsy with adenocarcinoma, as it has become in our department. A preliminary overview of 39 prospectively collected biopsies seems to support these results, adding the important dimensions of clinical utility and a low cost-benefit ratio to laboratory perspectives. Small biopsies may not demonstrate focal staining by CK20, but cases demonstrating any CK20 positivity can be evaluated according to our results.

In conclusion, this study identifies a subset (about 40% of cases) of metastatic adenocarcinomas in the liver for which the primary tumor site is detectable with a high degree of accuracy by using a simple, two-step immunostaining algorithm. According to our results, the predicted probability that a CK20+/CK7− metastasis originates in the colon or rectum is 78%. The predicted probability that a CK20+/CK7+ metastases originates in the pancreas or biliary tree is 74%. After exclusion of the stomach as primary site, the CK20+/CK7− pattern was accompanied by primary adenocarcinoma of the colon or rectum in 94% of the cases, and the CK20+/CK7+ pattern was accompanied by with a primary pancreatic or biliary carcinoma in 92% of the cases. In everyday practice, after a confirmed adenocarcinoma in a routinely stained liver biopsy, the search for an unknown primary tumor could begin with gastroscopy and gastric biopsy. An additional gynecologic examination can rule out a primary mucinous ovarian carcinoma. The possibility of primary lung carcinoma has to be examined as well. On the other hand, breast carcinoma as a primary site was not observed among our CK20+ cases, a result that can be very useful for breast carcinoma follow-up trials. The two-step CK algorithm seems to be very helpful in CK20+ cases. Experience is needed to avoid overinterpreting CK7+ staining in the reactively proliferated bile ducts. The complete absence of CK20 staining indicates a subset of adenocarcinomas in which the primary or metastatic character of the tumor or the primary tumor site (pancreas, gallbladder, stomach, breast, prostate, ovary, lung, or colon) cannot be predicted accurately by using this method.