Secondary Logo

Journal Logo

Original Article

Can pathologists reliably establish ampulla of Vater carcinoma histologic subtype on H&E alone? Concordance of subtype and comparison to immunohistochemistry-established subtype

Tanager, Kevin S. MDa; Li, Yueying MD, PhDb,c; Alpert, Lindsay MDa; Setia, Namrata MDa; Weber, Christopher MD, PhDa; Hart, John MDa; Xiao, Shu-Yuan MDa,b,c,∗

Author Information
doi: 10.1097/JP9.0000000000000032
  • Open



The ampulla of Vater, also known as the hepatopancreatic ampulla, is a complex anatomic structure located in the second part of the duodenum, formed by the union of the common bile duct and the main pancreatic duct, and surrounded by the sphincter of Oddi muscle.[1] This structure provides a conduit for the secreted products of the exocrine pancreas and bile to enter the intestinal (INT) lumen and thereby facilitate digestion.

Periampullary tumors are defined as tumors arising within 2 cm of the ampulla in related structures including head of pancreas, common bile duct, and duodenum. Ampullary carcinomas (ACs) are defined as tumors arising in the ampullary complex, located distal to the bifurcation of the distal common bile duct and pancreatic duct.[2] Given this specific anatomic location at the junction of pancreatic and INT tissues, ACs are considered a distinct histologic entity.

AC is a somewhat rare disease accounting for 0.2% of gastrointestinal cancers, with an approximate incidence less than 1 per 100,000 individuals.[2,3] Given that even small ACs can cause symptoms of biliary obstruction, they are usually diagnosed at a relatively earlier stage in contrast to non-obstructing tumors of the periampullary region. Such early detection may contribute to the more favorable prognosis for AC compared to periampullary pancreatic or duodenal cancer.[2]

Management of AC typically involves surgical resection via either ampullectomy or pancreaticoduodenectomy (Whipple procedure), which is performed with curative intent except in cases of obvious clinically advanced stage. Following pathologic diagnosis and staging, adjuvant chemotherapy is often also employed, especially in cases with lymph node involvement (>pN0), residual disease (non-R0), or large tumor size (advanced pT).[4]

Pathologic staging (pTNM) of the surgical resection specimen provides valuable prognostic information in cases of AC, but additional prognostic factors have been proposed, including histologic phenotype. Of note, ACs exhibit 2 major histologic subtypes: INT and pancreatobiliary (PB); and sometimes occurs as a mixed (MIX) pattern of these 2 subtypes. Histologic subtyping of AC as a prognostic factor has not yet been widely validated, but emerging evidence suggests that the presence of PB differentiation independently predicts poor prognosis after pancreaticoduodenectomy.[5–9] Furthermore, histologic subtype has recently begun to influence the selection of adjuvant chemotherapeutic agent in applicable cases.[8,10] If histologic subtype does indeed portend differing prognosis between PB and INT, and also increasingly affects chemotherapy selection, it therefore becomes imperative for pathologists to make this distinction upon pathologic diagnosis of the pancreaticoduodenectomy specimen.[2,11]

Histologic subtyping of AC involves integration of morphologic and immunophenotypic features. The morphologic features of the INT subtype include well-formed tubular glands composed of pseudostratified tall columnar cells often containing mucin and exhibiting basally located oval- or cigar-shaped nuclei. The PB subtype is characterized by simple or branching glands and small solid nests of cells surrounded by dense desmoplastic stroma, with cuboidal to low columnar cells arranged in a single layer without nuclear pseudostratification and often with pleomorphism.[12,13] While some ACs can be readily subtyped based on this scheme, whether exhibiting a single subtype (PB or INT) or multiple clearly defined subtypes (MIX), many cases exhibit ambiguous histologic features, posing difficulty to the subtyping.[7] Therefore, immunophenotype has been explored to facilitate this task, particularly in ambiguous cases. Whether or not to use immunostains has mostly been at the discretion of individual pathologists.

Several studies have employed immunohistochemistry (IHC) panels to further characterize AC histologic subtype, which have included cytokeratins (CK7, CK17, and CK20), MUCs (MUC1, MUC2, MUC4, MUC5AC),[2,14] and the transcription factor CDX2.[6] Broadly, PB subtype has been associated with CK7, CK17, MUC4, and MUC1 staining, while INT subtype has been associated with CK20, MUC2, and CDX2.[6,15,16]

A definitive IHC panel for determining AC has not yet been established, so at present subtyping may be done based on the histologic features of an H&E slide alone, or in conjunction with IHC information. Given that IHC can be a costly addition to routine pathology diagnosis, some studies have sought to assess interobserver concordance of histologic subtyping based on H&E alone, which have overall yielded only modest concordance among pathologists when assigning histologic subtype on solely H&E.[7,10]

The study sought to assess both interobserver concordance of assigning histologic subtype on H&E alone, as well as percent correctness of this subtyping in respect to a known IHC gold standard subtype identity for each case. This information was gathered to assess how well pathologists can assign histologic subtype based on morphology alone, or if IHC should be performed routinely to ensure proper subtyping of AC.

Materials and methods

Case selection was performed first by searching for pancreaticoduodenectomy (Whipple procedure) with a final diagnosis of invasive AC in our surgical pathology archives. This resulted in a total of 34 such cases spanning a 3-year period from July 2014 to June 2017. Institutional Review Board (IRB) approval was granted for this study. Next, the final reports for each case were reviewed for several parameters, including: final diagnosis (AC, excluding tumors arising from adjacent but non-ampullary sites such as common bile duct or duodenal non-ampullary), histologic subtype (if reported in top line diagnosis, synoptic report, or comment), pathologic stage (≥pTN), and any applicable IHC studies. Slides for each case were reviewed, including IHC slides if previously performed. For those cases which had not undergone IHC, the IHC stains for CK7, CK20, and CDX2 were performed on a block corresponding to 1 representative H&E slides. Each case was then single-blinded and de-identified.

Next, the H&E slides for each case set were given to 5 gastrointestinal pathologists for review. Review consisted of tasking each pathologist with assessing the H&E slides for histologic subtype, with answer options including INT, PB, and MIX. The pathologists were prompted to assign histologic subtype based on their overall impression of the tumor for each case. Further instruction was provided by which MIX subtype was defined as cases with a secondary subtype comprising 30% to 50% of the tumor. At this point, with a histologic subtype assigned to every case by all 5 pathologists, interobserver concordance was calculated (kappa).

Beyond interobserver concordance, an assessment of correctness of subtyping by H&E alone was performed by comparing the H&E subtype to an IHC-determined subtype for each case. To do this, the IHC studies for each case were then reviewed by 1 gastrointestinal pathologist in order to determine and assign an IHC-defined subtype identity to each case, which was termed the “IHC gold standard.” This pathologist was also one of the reviewers of the H&E subtyping, but this individual performed the IHC review after offering an H&E subtype for each case, to avoid bias. IHC review was performed by assessing each slide for staining strength on a 0, 1+, 2+, and 3+ scale, with 0 and 1+ scored as weak, and 2+ and 3+ scored as strong. IHC gold standard groupings were determined as follows: strong CK7 with weak CK20 and weak CDX2 was defined as PB; strong CK20 or strong CDX2 with weak CK7 was defined as INT, and either strong CK20 or strong CDX2 with concurrent strong CK7 was defined as MIX (Figs. 1–3).

Figure 1
Figure 1:
Representative case of pancreatobiliary (PB) histologic subtype, verified by IHC including CK7 (3+), CK20 (0), and CDX2 (0).
Figure 2
Figure 2:
Representative case of intestinal (INT) histologic subtype, verified by IHC including CK7 (0), CK20 (3+), and CDX2 (3+).
Figure 3
Figure 3:
Representative case of mixed (MIX) histologic subtype, verified by IHC including CK7 (3+), CK20 (3+), and CDX2 (3+).

Once each case was assigned an IHC gold standard subtype, the H&E subtyping results as assigned by the 5 pathologists were compared to the IHC subtype. Percent correctness was then defined as the percentage of cases for which the histologic subtype was correctly identified by H&E alone when compared to the IHC gold standard subtype, for 0, 1, 2, 3, 4, and 5 pathologists (out of 5 pathologists). The cases for which either 3, 4, or 5 out of 5 pathologists were correct were considered to have had a majority of pathologists (at least 3 out of 5) correctly assign subtype on H&E alone.

Lastly, information regarding adjuvant chemotherapy regimen and overall survival was collected for each patient via chart abstraction in the hospital electronic medical record software Epic. This information including initial adjuvant chemotherapy agent or agents, additional adjuvant chemotherapy agents implemented upon disease progression (if applicable) and survival in months following pancreaticoduodenectomy. For each case, when able, this information was viewed in light of the histologic subtype at time of diagnosis, as well as IHC gold standard subtype; and indirectly, in light of pTNM stage at time of surgery.


Interobserver concordance for H&E histologic subtyping was fair (kappa = 0.31). Percent correctness refers to agreement between H&E subtype and the IHC subtype on a particular case, and shown in Table 1. For example, for 5 cases (15% of the cases), none of the initial H&E subtype was correct as compared to the IHC subtype. Likewise, for 21% of cases 2 pathologists were correct, whereas on 18% of cases 3 pathologists were correct, on 9% of cases, 4 pathologists were correct, and on 24% of cases all 5 pathologists were correct. A majority of pathologists (at least 3 pathologists) were therefore correct on 50% of cases (Table 2). For those cases in which percent correctness was poor (2 or fewer pathologists being correct), the IHC gold standard subtype was more often MIX, with pathologists tending to rate either PB or INT, thereby underassessing the minor histologic component.

Table 1
Table 1:
Percent correctness based on immunophenotypic subtype.
Table 2
Table 2:
Percent correctness as either minority or majority of pathologists.

When probing the relationship between interobserver concordance and percent correctness, for the 11 cases in which at least 4 pathologists were concordant, the subtype was also correctly identified.

Regarding chemotherapy, only 16 of the 34 patients received adjuvant chemotherapy following pancreaticoduodenectomy (Table 3). Since the IHC gold standard subtyping employed above was performed for this study and was not reported at the time of final surgical pathologic diagnosis for these cases, chemotherapy regimens were chosen by the clinicians based on the diagnostic pathology report, which in some cases noted a histologic subtype, and in other cases simply stated “adenocarcinoma NOS.” Regardless, for cases signed out as “pancreatobiliary” 6 received gemcitabine as first-line, and 2 received “adjuvant chemotherapy” which was not further specified. Of “adenocarcinoma NOS” cases, 4 received gemcitabine, and 1 received “adjuvant chemotherapy” which was not further specified. Of “intestinal” cases (including those with signet ring cell features), 3 received FOLFOX as first-line. Of note, cases for which a histologic subtype was assigned at time of surgical pathology report, the IHC gold standard subtype assigned in this study nearly always matched, with only occasional cases which tended to assign MIX when only “pancreatobiliary” had been reported at time of sign-out (Table 4).

Table 3
Table 3:
Adjuvant chemotherapy regimen based on histologic subtype assigned in surgical pathology report following resection.
Table 4
Table 4:
Adjuvant chemotherapy regimen based on immunophenotypic subtype determined retrospectively in this study (based on CK7/CK20/CDX2 panel).


Treatment of malignancy, including surgical resection and chemotherapy, may provide a range of improved outcomes including some span of disease-free survival, if not outright cure. Staging of a malignancy seeks to delineate the extent of disease in a patient following detection, and these staging schemes are refined over time via large-scale analysis of outcomes, in order to better predict patient prognosis.[4,17] Surgical resection with accompanying histopathologic assessment is a fundamental part of cancer management, which is both therapeutic and diagnostic. Specifically, pathologic staging information is a critical component of assessing disease extent and planning for further cancer treatment modalities such as chemotherapy. While staging information is critically important for prognostication, other non-strictly staging-related histopathologic findings can affect prognosis, such as lymphovascular invasion, tumor grade, and, in some cases, histologic subtype.[4] Some malignancies, such as carcinoma of the urinary bladder and endometrial carcinoma, exhibit a range of histologic subtypes with differing biological behavior, with some subtypes much more aggressive than others.[13] This phenomenon compels pathologists to thoroughly characterize tumor resection specimens, including assessment of histologic subtype, whereby large-scale analysis of malignancies and their histologic subtypes may provide additional prognostic information beyond staging, even to the extent of dictating differing chemotherapy regimens.

As explored above, AC may exhibit PB or INT histologic subtype, or a MIX phenotype containing elements from both subtypes. This may be due to the cell of origin of the initial tumorigenesis since at the particular location both INT and biliary type epithelial cells exist. Alternatively, it may be due to defined differentiation in tumor progression, assuming the above phenotypes. Emerging evidence has demonstrated INT subtype to portend an overall improved prognosis over PB or MIX subtypes. Furthermore, while AC had traditionally been treated with gemcitabine regardless of subtype, INT cases have recently been treated with FOLFOX-based chemotherapy, analogous to colorectal carcinoma.[9] This emerging prognostic and therapeutic difference between pure INT vs PB/MIX ACs has increasingly driven gastrointestinal pathologists to assign histologic subtype on AC resections, in order to provide additional information to clinicians managing these patients. However, the ambiguity of histologic features in these tumors has led to difficulty in robust subtyping, with imperfect interobserver concordance especially on H&E alone, leading to efforts to support subtyping with immunophenotypic information.

The study sought to assess interobserver concordance among a group of 5 gastrointestinal pathologists for subtyping 34 cases of AC based on H&E alone, and the overall accuracy of subtyping when matched with the immunophenotypic profile of each case. Interobserver concordance was fair (kappa = 0.31), which indicates that while some cases yielded agreement, other cases proved difficult to confidently subtype. Pure INT and PB cases were more often concordant than MIX cases; in these MIX cases, lack of concordance typically arose when pathologists assigned a single subtype to the dominant phenotype, but under appreciated the extent of the minor subtype. This implies that ACs should be adequately sampled at grossing, and should be thoroughly assessed microscopically to fully appreciate the extent of histologic subtypes present.

In order to assess the accuracy of subtyping, a gold standard subtype was established for each case via an IHC panel of CK7, CK20, and CDX2. Although a broad array of IHC stains have been investigated in attempts to form a panel capable of subtyping AC, no single panel has been proven ideal. We chose these 3 markers for the simplicity and anticipated reproducibility. In addition, some of the cases in this study had had these 3 stains performed at time of diagnosis to aid in subtyping and affected final diagnosis and, thereby, clinical judgment.

Once the IHC gold standard for each case had been established, percent correctness was calculated, whereby each case was assessed for how many pathologists out of 5 total arrived at the correct subtype (Table 1). Given that a majority of pathologists (at least 3 out of the total 5) were correct on 50% of cases (Table 2), this indicates that about half of cases of AC exhibit a fairly straightforward histologic appearance and can be handily subtyped. The remainder, in which only a minority of pathologists correctly assessed the subtype, more frequently occurred in cases with a MIX phenotype, with pathologists tending to assign only the predominant subtype, and underestimating the minor contributor. It is acknowledged that some lack of accuracy may arise from the admittedly imperfect IHC scheme employed to generate the gold standard subtype for each case, and one possible area of future direction would be to explore additional stains, such as the MUC family, in order to further strengthen this IHC panel.

Data on chemotherapy regimen and survival was somewhat limited. Sixteen of the 34 AC cases assessed in this study received adjuvant chemotherapy post-pancreaticoduodenectomy. Unfortunately, some of these cases were lost to follow-up without mention of a specific chemotherapy drug regimen. The 2 major chemotherapy regimens utilized were FOLFOX or FOLFIRI (leucovorin, 5-fluorouracil, and either oxaliplatin or irinotecan), vs gemcitabine and/or capecitabine, with the exception of those not specified, below designated as chemo not-otherwise-specified (NOS). A number of cases assigned MIX in this study were originally signed out as AC not-otherwise-specified (NOS), perhaps indicating that initial diagnosis was complicated by multiple histologic subtypes, and these patients tended to receive gemcitabine as well. Those cases in which a histologic subtype was assigned at time of diagnosis, PB cases tended to receive gemcitabine, while INT cases tended to receive FOLFOX-based therapy (Table 3). FOLFOX includes either FOLFOX or FOLFIRI, and gemcitabine includes gemcitabine and/or capecitabine. Of note none of these were assigned MIX subtype at time of diagnosis. Table 4 refers to the chemotherapy regimen based on the immunophenotypic subtype determined in this study based on the CK7, CK20, and CDX2 panel (therefore, reflecting retrospective information not on hand at the time of diagnosis and treatment). Of note, a fair number of cases received no chemotherapy at all, but these were typically cases diagnosed and resected at an earlier stage, for which chemotherapy was not indicated, and prognosis may be dominated by stage rather than histologic subtype.

Overall, based on these findings, the histologic subtyping of AC seems to be only confidently performed on H&E alone in straightforward histologic cases, and those with any ambiguity seem best served with further elucidation by IHC. Some clinical oncologists have demonstrated increasing interest in the histologic subtype of AC cases, in order to guide adjuvant chemotherapy regimen selection. In situations of AC with ambiguous histomorphology on the initial H&E slides (such as areas resembling both foveolar-type and INT-type epithelium in the same tumor, or tumors with overall poor differentiation), implementation of this IHC panel may allow for elucidation of histologic subtype. Given that histologic subtype also affects prognosis and dictates chemotherapy regimen, pathologists may be best served by erring on the side of performing IHC to add confidence to their histologic subtyping, in order to provide the most thorough evaluation possible to the clinician, and thereby, to the patient.


We would like to thank Ryan McGary and the histology team for their assistance with recuts, Dr Girish Venkataraman and Renata Kornecka for their assistance with immunohistochemistry, and Kristen Wroblewski for assistance with statistical analysis. The results of this study had been presented at the 2018 Annual Meeting of the United States & Canadian Academy of Pathology.

Author contributions


Financial support


Conflicts of interest

The authors declare no conflicts of interest.

Ethics approval

IRB approval.


1. Zheng-Pywell R, Reddy S. Ampullary cancer. Surg Clin North Am 2019; 99:357–367.
2. Williams JL, Chan CK, Toste PA, et al. Association of histopathologic phenotype of periampullary adenocarcinomas with survival. JAMA Surg 2017; 152:82–88.
3. Rostain F, Hamza S, Drouillard A, et al. Trends in incidence and management of cancer of the ampulla of Vater. World J Gastroenterol 2014; 20:10144–10150.
4. Lowe MC, Coban I, Adsay NV, et al. Important prognostic factors in adenocarcinoma of the ampulla of Vater. Am Surg 2009; 75:754–760. discussion 761.
5. Morini S, Perrone G, Borzomati D, et al. Carcinoma of the ampulla of vater: morphological and immunophenotypical classification predicts overall survival. Pancreas 2013; 42:60–66.
6. Kumari N, Prabha K, Singh RK, et al. Intestinal and pancreatobiliary differentiation in periampullary carcinoma: the role of immunohistochemistry. Hum Pathol 2013; 44:2213–2219.
7. Reid MD, Balci S, Ohike N, et al. Ampullary carcinoma is often of mixed or hybrid histologic type: an analysis of reproducibility and clinical relevance of classification as pancreatobiliary versus intestinal in 232 cases. Mod Pathol 2016; 29:1575–1585.
8. Westgaard A, Tafjord S, Farstad IN, et al. Pancreatobiliary versus intestinal histologic type of differentiation is an independent prognostic factor in resected periampullary adenocarcinoma. BMC Cancer 2008; 8:170.
9. Chang DK, Jamieson NB, Johns AL, et al. Histomolecular phenotypes and outcome in adenocarcinoma of the ampulla of vater. J Clin Oncol 2013; 31:1348–1356.
10. Ang DC, Shia J, Tang LH, et al. The utility of immunohistochemistry in subtyping adenocarcinoma of the ampulla of vater. Am J Surg Pathol 2014; 38:1371–1379.
11. Jin Z, Hartgers ML, Sanhueza CT, et al. Prognostic factors and benefits of adjuvant therapy after pancreatoduodenectomy for ampullary adenocarcinoma: Mayo Clinic experience. Eur J Surg Oncol 2018; 44:677–683.
12. Albores-Saavedra J, Schwartz AM, Batich K, et al. Cancers of the ampulla of vater: demographics, morphology, and survival based on 5,625 cases from the SEER program. J Surg Oncol 2009; 100:598–605.
13. Henson DE, Schwartz AM, Nsouli H, et al. Carcinomas of the pancreas, gallbladder, extrahepatic bile ducts, and ampulla of vater share a field for carcinogenesis: a population-based study. Arch Pathol Lab Med 2009; 133:67–71.
14. Perysinakis I, Minaidou E, Leontara V, et al. Differential expression of β-catenin, EGFR, CK7, CK20, MUC1, MUC2, and CDX2 in intestinal and pancreatobiliary-type ampullary carcinomas. Int J Surg Pathol 2017; 25:31–40.
15. Liu F, Shen D, Ma Y, et al. Identification of ampullary carcinoma mixed subtype using a panel of six antibodies and its clinical significance. J Surg Oncol 2019; 119:295–302.
16. Yun SP, Seo HI. Prognostic impact of immunohistochemical expression of CK7 and CK20 in curatively resected ampulla of Vater cancer. BMC Gastroenterol 2015; 15:165.
17. Ha HR, Oh DY, Kim TY, et al. Survival outcomes according to adjuvant treatment and prognostic factors including host immune markers in patients with curatively resected ampulla of vater cancer. PLoS One 2016; 11:e0151406.

Ampulla of Vater; Immunohistochemistry; Immunophenotyping; Prognosis; Subtype

Copyright © 2019 The Chinese Medical Association, Published by Wolters Kluwer Health, Inc. under the CCBY-NC-ND license.