An Appraisal of Immunohistochemical Stain Use in Hepatic Metastasis Highlights the Effectiveness of the Individualized, Case-Based Approach: Analysis of Data From a Tertiary Care Medical Center Open Access

The pathology database was searched for liver biopsies performed for hepatic metastasis from 2015 to 2018. Cases of primary liver tumors, including cholangiocarcinoma, were excluded when the diagnosis was clinically suspected, known (confirmed via prior pathologic review), or diagnosed based on clinicopathologic grounds. However, cases that presented without a clinical suspicion and those for which cholangiocarcinoma was diagnosed after clinicopathologic workup to rule out metastasis were included. The pathologic diagnosis, IHC stains, and patient demographics were obtained from the pathology report.

Our institutional practice of IHC workup involves selection of the markers based on patient demographics, clinical history, review of available clinical workup, and tumor morphology in each case. Besides age, gender, and any prior history of known malignancy, the morphology of the tumor and any clinically suggestive findings are taken into account for choosing the IHC markers in each case. For this study the tumors were divided into 6 morphologic patterns (Table 1; Figure 1, A through F): adenocarcinoma (Ad-Ca), poorly differentiated carcinoma (PD-Ca), squamous cell carcinoma (SCC), neuroendocrine neoplasm (NEN), nonepithelial tumors (N-EpiT) (melanoma, lymphoma, and spindle cell tumors), and carcinoma, not otherwise specified (Ca-NOS). NENs included well-differentiated neuroendocrine tumors as well as neuroendocrine carcinomas, including small cell carcinoma. If a primary malignancy was already known or highly suspected based on clinical information, and/or prior material was available for comparison, very few IHC markers (2 or 3, and sometime none) were applied in each case. In other cases, with no known primary, the tumor morphology was used to choose IHC markers; common markers used with each morphologic pattern are shown in Table 1. However, it needs to be emphasized that none of the panels are preset panels in our practice, and the final choice is still morphology and pathologist dependent.

Cases that presented with no known primary site prior to the liver biopsy were considered initially CUP (i-CUP), and those where the primary site could not be determined despite extensive clinicopathologic workup were designated true CUP (t-CUP). The clinical workup varied in these cases and included serologic workup and various imaging studies. The slides of all cases were reviewed by 2 experienced gastrointestinal pathologists (X.Z., D.J.).

The IHC markers used in each case, including the total number, were recorded. The mean and SD for number of markers used in each category were calculated. We also examined the setting (generalist versus specialist) and the level of experience of the pathologist signing out the case. Because the number of years of practice was not easily available for each pathologist signing out the cases during the study period, the academic rank at the time of signing out the case was used as an indirect measure of experience. For further analysis, the faculty were divided into junior (instructor and assistant professor) and senior (associate professor and professor). The pathologists signing out the cases were also categorized as generalists or specialists based on their practice locations.

The final clinical diagnosis and therapeutic decisions made in each case were recorded from the electronic medical record system and were reviewed by 2 gastrointestinal oncologists in the study (T.S., J.L.). Institutional review board approval was obtained before the start of the study.

A total of 406 patients with liver biopsies of metastatic tumors met our study criteria. The median overall patient age was 65 years (range, 16–90 years), with an approximate male to female ratio of 1 to 1 (193 men, 213 women). The median age of patients with t-CUP was also 65 years (range, 46–79 years). The patient demographics were 80.5% (327 of 406) White, 12.8% (52 of 406) Black, 4.4% (18 of 406) Hispanic, 1.7% (7 of 406) Asian, and 0.5% (2 of 406) other.

Among all the liver biopsies with metastatic tumors, 61% (248 of 406) had pathologic material from the prior known tumor for comparison (core biopsy, fine-needle aspiration, or excision). In these cases, the pathologists used an average (±SD) of 2.6 (±2.6) IHC markers/case (Figure 2). In 39% (158 of 406) of cases, the primary site was considered unknown at the time of liver biopsy, and these constituted i-CUP cases, although in some cases there was some clinical or radiologic suspicion of a primary site. The pathologic examination and IHC workup of the i-CUP cases suggested the likely primary/differential diagnosis, which on further clinical workup led to a confirmation of the primary site in 93% (147 of 158) of the cases (Figure 2). After pathologic analysis and IHC, 97% (395 of 406) of the entire study cohort had a primary site determination. Thus, the primary site remained unknown after clinical, radiologic, and pathologic analysis in only 3% of all cases (11 of 406), which were designated t-CUP (Figure 2). Once N-EpiTs were excluded, the number of CUP cases was 2.9% (11 of 382).

Among morphologic patterns, Ad-Ca (62%; 253 of 406) was the most common, followed by NEN (13%; 54 of 406), PD-Ca (11%; 43 of 406), N-EpiT (6%; 24 of 406), SCC (5%; 20 of 406), and Ca-NOS (3%; 12 of 406) (Figure 3). For the Ad-Ca category, the most common primary sites included the large bowel (24%; 60 of 253), breast (22%; 56 of 253), pancreas (19%; 47 of 253), and lung (11%; 28 of 253). The distributions of primary sites for all cases and by each morphologic category are shown in Figures 4 and 5, respectively. Overall, the most frequent primary sites for hepatic metastasis were as follows: lung (19%; 76 of 406), breast (17%; 69 of 406), large bowel (16%; 63 of 406), and pancreas (14%; 58 of 406). Among metastatic t-CUPs, 9% (1 of 11) were Ad-Ca, 55% (6 of 11) were NEN, 18% (2 of 11) were PD-Ca, and 18% (2 of 11) were SCC (Table 2). The frequencies of morphologic categories of the metastases and associated primary sites are listed in Table 2. Of note, 21 cases of cholangiocarcinoma presented as i-CUP. One case of cholangiocarcinoma was excluded from the analysis, as it already had a prior confirming tissue diagnosis.

For liver biopsies with an unknown primary (36%; 147 of 406) at the time of biopsy (i-CUP), the pathologist used an average (±SD) of 5.9 (±3.2) IHC markers/case. In cases that remained of unknown primary despite extensive workup (t-CUP) (3%; 11 of 406), the pathologist used an average (±SD) of 9.5 (±3.0) IHC markers per case. Among the morphologic patterns, the maximum number of IHC markers on average was used for PD-Ca (5.9), followed by Ca-NOS (5.5). The lowest number of markers was used for Ad-Ca (3.3). Among the 5 most common primaries, the average IHC markers/case were 1.8 for large bowel metastasis, 3.4 for breast metastasis, 4.2 for pancreatic metastasis, 5.6 for lung metastasis, and 7.2 for biliary tumors (Table 3). The 5 most used IHC stains were CK7 (12%), CDX2 (10%), CK20 (10%), TTF1 (9%), and GATA3 (7%). For each morphologic category, the 5 most used IHC stains are also listed in Table 3.

Of 406 liver biopsies with metastasis, 82% (334 of 406) were diagnosed by a pathologist specialized in gastrointestinal and liver pathology, and 18% (72 of 406) were diagnosed by a pathologist practicing in a community-based general sign-out setting (Table 4). Specialist pathologists used fewer IHC stains per case (3.7 ± 3.6) compared with generalist pathologists (5.0 ± 2.2), which was statistically significant (P = .007). Among specialized pathologists, junior faculty used on average more stains (4.1 ± 3.6 IHC stains/case) compared with senior faculty (3.7 ± 3.6 IHC stains/case), but this was not statistically significant (P = .44).

Liver biopsies for workup of metastasis are commonly encountered in pathology practice, but standardized practice guidelines are lacking. Prior studies have looked at the sensitivity and specificity of various IHC markers for predicting tumor type/primary site, and many algorithms have been suggested for workup of hepatic metastases from an unknown primary. The list of such markers and suggested algorithms has evolved over the years and continues to do so.

An early study on liver metastases showed that a primary site could be identified in only 27% of cases based on pathologic workup.¹⁴  A subsequent study on metastatic Ad-Ca from unknown primary sites showed better results with a standardized diagnostic panel of 4 IHC markers that correctly identified primary sites in 66% of cases. ¹⁵More recently, another study using a panel of 10 IHC markers on metastatic Ad-Ca was helpful in identifying the primary site in 88% of cases.⁷  Another study showed IHC accurately predicted the primary site in 83.3% of cases from a wide spectrum of differential diagnoses using an average of 8.3 stains.¹⁶  These studies suggest that with increasing numbers of IHC markers, especially when they are used as a panel, the chances of identifying the possible primary site in liver metastases increase, and hence many laboratories use preset panels, which often include more than 15 IHC markers. However, in current practice, the need to conserve tissues for various ancillary prognostic and molecular markers has become critical, and it is important to use strategies that are not only cost-effective but also tissue preserving. Arguably, standardized panels remove subjectivity for choosing appropriate markers and expedite the workup of cases. On the other hand, an individualized, case-based approach requires medical judgement, but also tends to reduce the cost while helping to conserve tissue in the blocks. However, to our knowledge there are no published data on IHC use and the cost-effectiveness of this approach compared with the use of preset standardized IHC panels. Our study, which is a large retrospective analysis of liver biopsies with metastasis from a tertiary care academic medical center, provides data on patterns of hepatic metastases and IHC use in such cases and analyzes some factors responsible for variations within the practice. The study also provides evidence that the individualized, case-based approach is highly effective in the workup of liver metastasis, with only very few cases that remain t-CUP.

The racial demographics of our study are similar to US Census Bureau national statistics, with slightly fewer African American, Hispanic, and Asian persons, and the overall distribution of primary sites is also similar to that in existing literature.¹⁷  In our study, among various histologic patterns, Ad-Ca comprised the largest group (253 of 406; 62%), and NENs (54 of 406; 13%) were the second most common, with half originating from the lung (28 of 54; 52%). Epidemiologically, primary lung NENs have the highest incidence among all NENs.¹⁸  Among metastatic PD-Cas, the majority originated from breast (13 of 43; 30%) and lung (12 of 43; 28%), followed by urothelial carcinomas (5 of 43; 12%) and pancreatic carcinomas (3 of 43; 7%). Thus, when encountering a metastatic PD-Ca of the liver, one may consider staining for these entities or performing a clinical and radiologic workup aimed at these sites.

The choice of IHC markers in our study varied among cases, depending upon the clinical setting and the pathologist evaluating the case. For example, in an elderly man with a metastatic carcinoma showing Ad-Ca with dirty necrosis and elevated serum carcinoembryonic antigen levels suggestive of a colorectal primary, many in our practice performed only a site-specific marker, for example, a CDX2 or SATB2 stain, whereas some performed a panel of CK7, CK20, CDX2, and/or SATB2. Similarly, in another setting of metastatic Ad-Ca in the presence of a lung mass, some performed only TTF1 or napsin A, whereas others used a broader panel of CK7, CK20, napsin A, and/or TTF-1. Overall, in this setting, where a primary tumor was already known and/or histology of the primary tumor was available for comparison and compatible, often very few and sometimes no IHC markers were used. Thus, it is not surprising that the mean number of markers used in this setting was very low (<3). In the other clinical setting, when there was no clinical suggestion of a primary site and/or a very nonspecific or poorly differentiated histology, a more extensive panel of IHC markers was used up front by most pathologists based on the tumor morphology, although the combination of markers was still pathologist dependent in our practice. For example, in a female patient with a metastatic Ad-Ca but no known or suspected primary, the initial panel often included CK7, CK20, CDX2, TTF1, GATA3, and PAX8. On the other hand, a set of neuroendocrine and/or hepatocellular markers was frequently added to this panel if the carcinoma was composed of polygonal cells without any specific pattern (Ca-NOS). Despite more markers required in this clinical setting, the mean number of markers in this category was still small (<6). Besides the clinical setting, tumor morphology played a major role in deciding the choice of IHC markers.

Of particular interest is the i-CUP category in our study, where the biopsies came to the pathologist without a known primary and relied heavily on the pathologist's judgment for further workup. Pathologists in our practice used an average of 5.9 IHC stains/case (range, 6–15 IHC stains/case) in this subset to help identify the primary site. With the use of IHC, and clinical and radiologic information where available, pathologic diagnosis helped in the identification of the primary site in 97% of all and 93% of i-CUP cases. Thus, after workup, only 3% of all cases remained as t-CUP cases, which is similar to the reported incidence of CUP (3%–5%).^3,4  As expected, these t-CUP cases often required more extensive IHC markers (mean, 9.5; range, 0–17).

Interestingly, many t-CUPs are historically noted to be poorly differentiated tumors; however, in our study the majority (55%; 6 of 11) were NENs. This may be because in the group of NENs (well-differentiated tumors or carcinomas), although IHC markers are excellent in the identification of neuroendocrine differentiation, the IHC markers for localization of primary site (eg, TTF-1, CDX2, PAX8) are not very sensitive or specific in practice. This may not matter much from a clinical management point of view, yet these represent tumors of unknown primary. Similarly, for certain other tumor types (eg, melanoma, lymphoma, sarcoma), the specific histologic type is often more important than the primary site of origin. Correlation with clinical findings on subsequent workup is also critical in determination of the primary site. In a prior study in which pathologists were blinded to clinical and radiologic information, they determined the correct primary site in only 83.3% of overall cases, which is similar to our own experience.¹⁶  Correlation with the adjunct clinical and radiologic data when needed allowed primary site determination in 97% of our cases.

The incidence of CUP in prior studies has varied widely and is dependent on the extent of workup and variation in practices across centers. In the most recent analysis based on comprehensive workup, it remains about 2% to 5%,⁴  and in our study, despite use of overall fewer IHC markers, the incidence was 3% of all cases and 2.9% of all carcinomas, which is comparable. The role of molecular pathology in this area is still evolving. Advanced diagnostics (molecular/genomic analysis) are supposed to help identify the tissue of origin in CUP cases, but the issue remains controversial, and such tests have not yet gained wide acceptance in clinical practice. Gene expression profiling (GEP) studies on tumors have demonstrated sensitivities of up to 97.1%.^16,19–23  For poorly differentiated and undifferentiated carcinomas, GEP identified the tumor origin in only 92% of cases.^16,23  A subset of our cases was subjected to molecular testing, but with a goal of finding therapeutic targets and not for the site of origin. Although developments in GEP remain exciting for identifying potential targets for treatment and personalized medicine, recent literature has shown that site-specific treatments based on GEP have not resulted in significant improvements in survival.²⁴  The molecular tests are also expensive, and the widespread clinical utility of these assays await further studies.

We suspected that the experience of the pathologist and degree of specialization may impact the use of IHC markers used in such cases. Because it was difficult to reliably get the number of years in practice for all pathologists in our practice, we looked at their academic rank at the time of reporting the case as an indirect indicator of experience in the field, realizing this is not a perfect correlation. Our data show that junior faculty (clinical instructor or assistant professor) used an average of 4.1 ± 3.6 IHC stains/case compared with senior faculty's use of 3.7 ± 3.6 IHC stains/case; however, this was statistically not significant (P = .44). In addition to the small sample size in each group in our study, several other factors may have contributed, including (1) intradepartmental consultations, which are common in most practices, with junior faculty frequently seeking advice from others in complex cases; (2) variation in the level of expertise even within people of similar rank or duration in practice; (3) lack of correlation between expertise and rank or duration in practice; and (4) the fact that increase in expertise may plateau or become less impactful after the initial few years in practice. In our practice, most cases were handled at the institution's main campus, where the practice model is specialist based and cases were handled by a fellowship-trained gastrointestinal pathologist. However, a minority of cases were signed out at our affiliated community-level hospitals, where the practice model is generalist based. It was also not surprising to us that the average number of IHC markers used was fewer (3.7 ± 3.6) in the specialty-based system compared with the generalist system (5.0 ± 2.2), which was statistically significant (P = .007). Specialization in the field is an indirect marker of expertise, and as a result, the specialists are more likely to be selective in choosing their IHC panels. However, it is difficult to draw stronger conclusions in this regard, and better studies designed to address the issue of levels of experience, duration in practice, and expertise in the area are needed.

If one considers the approximate charge of $250 for 1 IHC marker, preset panels of 5, 10, and 20 IHC markers would lead to bills of $1250, $2500, and $5000, respectively. In comparison, using our individualized, case-based approach, patients with a known primary needed only 2.6 IHC markers on average, amounting to a charge of $650. For the i-CUP and t-CUP cases, the average numbers of IHC markers used per case were 5.9 and 9.5, respectively, and the charges per case would amount to $1475 and $2375, respectively. If one takes the entire cohort of 406 liver biopsies in our study, considering a standard preset panel of 10 IHC markers would lead to a charge of $1 015 000, which is more than double the $404 700 charge based on our individualized, case-based approach. Our approach is also much less costly compared with the reported actual cost of about $16 000 for pathology workup in one study,²⁵  where pathology costs accounted for 17.09% of the entire cost of workup of CUP cases. Obviously, the actual amounts billed/reimbursed vary depending on the payers, institutions, and country of practice; however, the differences between the 2 practice patterns are obvious.

Our results show that using an individualized morphology-based approach, despite lack of any standardization in the practice, uses fewer IHC markers compared with the preset panel-based approach for the workup of metastatic liver tumors and is cost-effective. Based on these results we would like to suggest that hepatic metastases could be divided into 2 broad categories: (1) cases where a primary tumor is known or suspected and (2) cases without any suspected or known primary. IHC use can certainly be minimized in the first group, and one should use the clinical setting, patient demographics, and tumor morphology to select the IHC panel in the second category.

Our study has several limitations. During the study duration, the availability of IHC markers in our laboratory and faculty changed, which could have impacted the results. Although the clinical charts were reviewed for establishing the final diagnosis, sometimes all the data were not available to us, based on the parameter on which the final diagnosis was established, and we could not analyze the value of clinical studies that helped in primary tumor site identification. In this study we did not aim to evaluate the accuracy of various markers or combination of markers useful in predicting the primary site from hepatic metastasis, and we cannot recommend any specific algorithms for clinical practice. Because new IHC markers are added to clinical practice on a regular basis, such panels also evolve continually, and the utility of such a standardized present panel becomes limited.

In summary, our study highlights the effectiveness of the integrated clinical, radiologic, and pathologic approach in determining primary sites of metastatic liver tumors and supports the use of an individualized, case-based approach for IHC workup of hepatic metastasis despite lack of standardization. The individualized, case-based approach seems to be more cost-effective compared with standardized preset IHC panels and helps preserve tissues for ancillary studies. Our study also provides some evidence that the number of IHC markers used in such cases decreases with increasing experience and subspecialization.