Do General Pathologists Assess Gastric and Duodenal Eosinophilia?: A Community-Based Study

Community-based nonacademic pathologists from across the United States were asked to participate in a study described as aiming to determine how gastric and duodenal biopsies are evaluated in general practice. Prospective participants had to meet the following requirements: (1) being in practice for 5 or more years after residency; (2) being board certified in anatomic pathology but not having had subspecialty training in gastrointestinal pathology; and (3) having a diagnostic workload that included at least 20% endoscopic mucosal biopsies. Among those who expressed an interest in the study and met the criteria, 31 pathologists were selected and offered a small honorarium to evaluate the cases, select the best diagnosis, and return slides and answer sheets within 1 week. Participants were then randomly assigned to 3 groups: 2 that included 10 pathologists each, and 1 group that included 11 pathologists, designated as groups A, B, and C, respectively. All participants were part of multiple practice groups varying in size from 5 to more than 50 pathologists from 12 states spanning all geographic areas of the United States. They were selected to represent community-based practices unaffiliated with academic centers that saw significant numbers of gastrointestinal biopsies. These pathologists and their referring gastroenterologists are likely to represent the first line in the diagnosis of EGID.

In the absence of consensus-based diagnostic guidelines for either EoG or EoD, we used a working definition used in previous studies.^12,21 EoG required the presence of 5 nonoverlapping HPFs (as defined above) each containing 30 or more eosinophils; EoD required the presence of 3 nonoverlapping HPFs each containing 30 or more eosinophils.

Cases were selected by 2 of the authors (R.M.G and K.O.T.) from the archives of a study originally designed to identify adult subjects with EoG or EoD.¹² As part of the original protocol, each subject had a set of biopsies from the gastric corpus, the gastric antrum, and the duodenum. Each set included 4 fragments of tissue from each site embedded in a single paraffin block from which 5-μm sections stained with hematoxylin-eosin were prepared. Eosinophils had been previously quantified in each of the biopsy fragments by 2 pathologists, as detailed elsewhere.^8,12,21 Briefly, we used microscopes with ×40 0.75 ∞/0.17/FN 20.5 lenses and 22-mm oculars. Each HPF in which eosinophils were counted measured 0.237 mm². For the present study, the sets comprised biopsies from 16 patients. The original diagnoses, as well as mean eosinophil counts and distribution of eosinophil infiltrates in the different tissue fragments, are outlined in Table 1 for the gastric biopsies and Table 2 for the duodenal biopsies. Participants were blinded to this information. Representative photomicrographs from 1 case of EoG and 1 of EoD are depicted in Figure 1, A and B, and Figure 2, A and B, respectively.

Three sets of slides were prepared from each case, stained with hematoxylin-eosin, and verified by 2 pathologists to be essentially identical to the originals and have eosinophil counts per HPF within ±10% of the original counts.

A form, depicted in Figure 3, was provided with each case. The form included a selection of gastric and duodenal histopathologic diagnoses and a space for additional comments. Participants were asked to select a diagnosis by choosing one or more of the boxes on the form and adding free text if they felt it was necessary to communicate additional information. Each case was accompanied by a succinct clinical history similar to the limited clinical data usually provided with gastrointestinal biopsies in clinical practice. Pathologists in group A received a brief clinical history with no mention of the possibility of allergic or eosinophilic conditions (eg, “30-year-old woman with dyspepsia and vomiting”). Participants in group B received an identical history, but with a suggestion of allergic or eosinophilic condition (eg, “history of atopic dermatitis” or “peripheral eosinophil count = 1500 eosinophils/μL”). Pathologists in group C were provided a history identical to group B, but with added context that the referring gastroenterologist suspected an eosinophilic gastrointestinal disorder, and that gastric and duodenal biopsies were obtained specifically to either confirm or refute their clinical suspicion.

Subjects who participated in the original study had signed a broad consent to have their biopsies examined for the purpose of research studies.¹² Cases were selected by the 2 pathologists who had previously evaluated the biopsies for the original study. They were unaware of any demographic or clinical information about the subjects, except for age and sex, and had no access to the patients’ medical history. Slide sets prepared for this study were assigned random numbers with no cross-reference to any original patient identification. Once prepared and assigned a study number, all cases became completely unidentifiable to all study participants, including the investigators. Therefore, no additional informed consent was necessary. Participants were asked whether they wished to be acknowledged in publications that reported this study: 12 pathologists agreed to have their names disclosed and 19 declined.

This observational study included a small number of participants and a small number of cases, with 14 possible diagnostic selections for the stomach and 10 for the duodenum. Therefore, the diagnoses returned by the participant pathologists were assessed with descriptive statistics and reported in tables. To compare the performance of the different groups of pathologists with respect to the diagnoses of eosinophilic conditions, we calculated odds ratios (ORs) and their 95% CIs. When the CI did not cross the unit, results were considered significant.

All participants returned the slide sets with the diagnostic forms completed. The average number of reported years in practice for all pathologists was 12, with minimal differences among the 3 groups. All pathologists were part of community-based practices with varying workloads that included approximately 20% gastrointestinal biopsy specimens. Thus, there were no notable differences in the overall diagnostic experience and practice settings of pathologists in groups A, B, and C.

Seven biopsy sets from the stomach consisted of normal mucosa. Overall, 28 of the 31 participants (90%) appropriately designated these as normal/minimal changes. Alternative selections included clinically inconsequential diagnoses such as reactive gastropathy and mild chronic inactive gastritis. Five gastric biopsy sets showed Helicobacter pylori gastritis: correct or suspected diagnoses (in some cases with the appropriate suggestion to perform a special stain to confirm) were made by 9 of 10 pathologists in group A, 8 of 10 in group B, and 9 of 11 in group C. There were 2 cases of autoimmune atrophic gastritis. A correct diagnosis (corpus-restricted metaplastic atrophy) was made by only 5 of 10 observers in group A, 3 of 10 in group B, and 5 of 11 in group C. The remainder of observers made descriptive diagnoses, mostly reactive gastritis in the antrum and chronic inactive gastritis with focal intestinal metaplasia in the corpus. A normal duodenum was included in 3 sets; these were diagnosed correctly by 9 of 10 pathologists in groups A and B and 10 of 11 in group C. The one case of celiac disease was correctly identified by 29 of the 31 pathologists. In contrast, only 2 of 10 pathologists in group A, 1 of 10 in group B, and none in group C commented on the marked duodenal intraepithelial lymphocytosis (Marsh IIa) present in one of the cases.

Two of the 16 gastric cases met the criteria for EoG (≥30 eosinophils/HPF in 5 nonoverlapping HPFs); in addition, both cases of corpus-restricted autoimmune atrophic gastritis and 2 cases of H pylori gastritis had elevated eosinophils, to an extent that a comment could have been added. Thus, the 10 pathologists in groups A and B had 60 individual opportunities to mention eosinophils, whereas the 11 pathologists in group C had 66 opportunities. As illustrated in Figure 4, eosinophils were mentioned in only 1 of 60 possible instances in group A (1.7%), 2 of 60 in group B (3.3%), and 10 of 66 of in group C (15.2%). The difference between groups A and B was not significant (OR, 2.04; 95% CI, 0.17–23.05). In contrast, when the responses of group C were compared with those of group A (OR, 10.53; 95% CI, 1.31–85.00; P = .02) and those of group B (OR, 5.18; 95% CI, 1.01–24.70; P = .05) there was a moderate statistical significance. No additional comments regarding eosinophil infiltrates were made by any pathologists in either group A or B. In contrast, 3 pathologists from group C used the provided space to comment on increased eosinophils: one wrote the peak eosinophil counts in 2 cases, one counted eosinophils in a single case, and the third described increased eosinophils without providing counts.

Eleven of the 16 duodenal cases met the criteria for EoD (≥30 eosinophils/HPF in 3 nonoverlapping HPFs); in addition, the case with duodenal intraepithelial lymphocytosis also had elevated eosinophils. Thus, there were 120 opportunities to mention eosinophils in groups A and B and 131 in group C. Figure 4 depicts the percentage of instances in which eosinophils were mentioned: pathologists in group A mentioned eosinophils in 1 of the 120 cases (0.8%), those in group B in 4 cases (3.3%), and pathologists in group C in 14 of 131 cases (10.7%). The difference between groups A and B was not significant (OR, 4.44; 95% CI, 0.45–35.27). In contrast, when the responses of group C were compared with those of group A (OR, 14.24; 95% CI, 1.84–110.03; P = .002) and those of group B (OR, 3.47; 95% CI, 1.02–10.85; P = .04) there was a moderate statistical significance. None of the pathologists in either group A or B made comments on the eosinophil content. In group C, 4 pathologists added comments (in 1, 4, 4, and 6 cases, respectively). In all these comments except 1, peak eosinophil counts were provided. No pathologist provided counts in more than 1 HPF.

The results of this study show that a group of general pathologists in private practice rarely reported prominent eosinophilic infiltration in gastric and duodenal biopsies. Pathologists in group A, who received no historical clinical information or requests regarding eosinophilia, and those in group B, who were given a brief history suggesting the possibility of an eosinophilic disorder, responded in an essentially identical manner, with only rare mentions of the eosinophilic infiltration in the mucosal biopsies they were observing. When a third group of pathologists (group C) were told that the clinicians submitting the biopsy specimens suspected an EGID and were specifically asked to evaluate for this possibility, a comment was added in only 9 of 60 possible instances (15%) of gastric and in 15 of 131 possible instances (11%) of duodenal biopsies. Eosinophil counts were not requested, but peak counts were provided in a total of 18 cases. Even when counts were given, a diagnosis of either EoG or EoD was not consistently selected.

To properly interpret these results, one consideration is whether a low level of experience in gastrointestinal pathology could explain the findings. However, the 3 groups were very similar in general and gastrointestinal diagnostic experience, with an average reported percentage of gastrointestinal mucosal biopsies of approximately 20% of their practice volume. With an average time in practice of 12 years, all participant pathologists can be assumed to have had considerable experience with gastrointestinal diagnoses during their professional life, which was reflected in the diagnoses of non-EGID cases we introduced into the slide sets specifically to assess the participants’ familiarity with gastrointestinal conditions. For gastric biopsies, a normal mucosa was correctly diagnosed in almost all cases. A weaker performance was found in the evaluation of H pylori gastritis, but this was still diagnosed between 78% (in 31 of 40 possible instances by pathologists in group A) and 89% (in 39 of 44 possible instances by pathologists in group C) of the time. This lower rate may be related to the absence of either special or immunohistochemical stains for the detection of H pylori, on which many pathologists routinely rely. Atrophic corpus-restricted metaplastic gastritis, which should normally evoke the probability of autoimmune gastritis, was diagnosed as such in only about one-half of the cases, with the other reports giving a variety of descriptive diagnoses that, although not incorrect, would likely fail to alert a clinician about the nature of the condition and would not trigger an adequate management plan. Pathologists, even gastrointestinal pathologists, seem to have a certain reluctance to diagnose atrophy. This attitude was documented in the preparatory phases of the updated Sydney system,²²  when gastric biopsies circulated to the 16 participant gastrointestinal pathologists yielded a dismal agreement on atrophy, and it was confirmed in subsequent efforts specifically designed to reach a consensus on the diagnosis and the grading of gastric atrophy.^23,24  Thus, the relatively poor performance of our study groups with respect to the diagnosis of atrophic gastritis may not be notable and would not impact an assessment of their competence as diagnostic pathologists.

Therefore, the question remains why there is little evaluating or commenting on mucosal eosinophil infiltrates. A possible answer may be found by revisiting the history of EoE. After the early publications suggesting that EoE could be an independent nosologic entity different from reflux esophagitis,^25,26  the emerging evidence was not rapidly accepted into practice. Furthermore, when histopathologic criteria based on counting eosinophils were published and then modified and updated, skepticism still remained.²⁷  Clinicians eventually embraced the concept of EoE and began asking pathologists to count eosinophils in esophageal biopsies, and the dogma that esophageal eosinophils are synonymous with reflux began to crumble. Now, most pathologists are familiar with the histologic parts of the EoE consensus guidelines^18,19  and have EoE as one of their primary differential diagnoses when approaching an esophageal biopsy. With nonesophageal EGIDs, we may be 2 decades behind EoE. The volume of literature seems to support this. Excluding case reports, from 2002 to April 2022 PubMed lists 2422 English-language publications on human EoE, with a peak in 2018 of 302 papers; during the same period, there were only 335 publications on nonesophageal human EGIDs, with a peak of 32 in 2020. Currently, very few clinicians express the suspicion of EoG, EoD, or EoC when they submit nonesophageal gastrointestinal biopsies. Rarely, if ever, do they ask for an eosinophil count.

The importance of nonesophageal EGIDs is becoming increasingly recognized by expert groups,^2,3,10,11  the information is being disseminated to gastroenterologists, and new treatment modalities are emerging that could benefit this patient population,^12,13  which appears to be larger than previously appreciated. Thus, it seems appropriate to attempt enhancing pathologists’ interest in considering and, when appropriate, diagnosing (or at least mentioning) EGIDs or eosinophilic infiltrations. This is unlikely to happen without consensus guidelines published not only in gastroenterology journals, but also, in more targeted forms, in journals that are read and well regarded by pathologists. To be widely used, such criteria must be relatively simple and realistic. We have recently presented a proposal for a simplified way to diagnose EoG and EoD.²⁸  However, it will need to be validated by other groups. Meanwhile, as busy clinical pathologists will probably have neither the time nor the inclination to count eosinophils in 8 or 12 HPFs in a gastroduodenal biopsy set, we suggest that those with access to large databases of systematically collected clinicopathologic data go through a reduction process and devise criteria that, although sacrificing specificity to sensitivity, can be used in the general practice of surgical pathology.

The authors wish to thank M. Anreder, MD; A. Campbell, MD; L. E. De Fanti, MD; A. De Vito, MD; J. A. Kozel, MD; R. Leonard, MD; E. Marinas, MD; J. O’Dell, MD; I. Shahab, MD; A. K. Sotelo, MD; J. Thibodeaux, MD; Z. Webb, MD; and the other 19 pathologists who chose to participate in this study anonymously.