costs associated with a pathologist’s interpretation of a gastric biopsy.
Batts et al8 provided guidelines for the use of H. pylori ancillary testing based on expert consensus among members of the Rodger C. Haggitt Gastrointestinal Pathology Society. The major point the authors emphasize is the fact that the vast majority of H. pylori infections can be diagnosed on
H&E stain with high sensitivity and specificity of 91% to 100%.9,13,14 The group recommended judicious use of IHC based on histologic findings and strongly recommends against up-front testing for H. pylori organisms. Chronic active gastritis (lymphoplasmacytic and neutrophilic inflammation) in which H. pylori organisms are not detected by routine stain is a definite indication for ancillary testing.
The authors further suggest testing in chronic inactive gastritis with at least moderate gastritis.8 Unfortunately, the distinction between mild chronic gastritis, without the need for further testing, and moderate chronic gastritis which should be tested, is incompletely defined. In contrast, the Houston update to the Sydney classification of gastritis seemed to suggest testing all gastric biopsies, to ensure a
careful examination and due to perceived efficiency benefits.
At other points in the manuscript, the authors made the recommendation somewhat more targeted by saying staining should be performed in “inflamed” biopsies in which H. pylori could not be detected by H&E.15 This group discussed the limitations in diagnosing inactive chronic gastritis, including an imperfect definition of normalcy, geographic variation in gastric immune cell concentrations, and variably applied pathologic criteria.15 Given these tremendous obstacles to a precisely defined abnormal state, it is not surprising that a practical point of distinction between “mild chronic inactive gastritis” and “moderate chronic inactive gastritis” was not offered.
At our institution, the use of ancillary testing for H. pylori is at the discretion of the attending pathologist. As a practical matter, special histochemical stains are rarely, if ever, used by our pathologists for this purpose, whereas the immunohistochemical stain (anti-H. pylori antibody; BioCare Medical, LLC, Concord, CA) is the default ancillary stain when one is considered necessary. No reflexive testing is performed and no strictly enforced rules are in place with regard to when an IHC stain should be ordered. Individual pathologists have no financial incentives to use or not use IHC. Thus, we hypothesized that there would be variability in H. pylori IHC usage rates among pathologists. We further hypothesized that since most pathologists would see a similar case mix, the difference in utilization rates would most likely be the result of varying criteria for testing. We assumed that different pathologists would have different thresholds for when an immunohistochemical test was required. We believe
this to be a reasonable assumption due to the lack of broadly accepted diagnostic criteria and testing algorithms, as discussed above. On the basis of this assumption, we determined if those pathologists who used IHC more frequently had a higher H. pylori diagnostic rate than those using IHC less frequently. We also investigated if pathologists with subspecialty practice in gastrointestinal (GI) pathology used IHC
at a different rate than general surgical pathologists

MATERIALS AND METHODS
Following approval by the Columbia University
Institutional Review Board [IRB-AAAO2353 (Y1M00)],
all gastric biopsy pathology reports from the New York Presbyterian/Columbia University Medical Center during the calendar year 2013 were retrieved and retrospectively reviewed. The initial database query selected for any cases that had a part-type text “stomach,” a term that is always included in specimen part-type assignment of stomach biopsies. Cases accessioned with multiple parts were included, as long at least 1 part of the case was from the
stomach. Gastroesophageal junction and gastric cardia specimens were excluded as biopsies at these sites are commonly performed for evaluation of reflux-related pathologies and H. pylori is less frequently diagnosed in these sites.
Resulting data consisted of the case number, case
pathologist, part-type information, final diagnosis, and
comments/notes section texts queried from the Cerner CoPath Plus database. This was then tabulated onto a spreadsheet for data review. For each case, we reviewed the final diagnosis text, comments, and addenda (if applicable) in order to tally the following: (1) the number of gastric biopsies in the case, (2) the number of H. pylori positive biopsies, (3) the number of IHC stains ordered,
and (4) the number of positive IHC stains.
These annotated data were imported as a comma separated values file into RStudio with R version 3.2.0 (64-bit) to facilitate data manipulation and statistical analyses. We calculated the diagnostic rate of H. pylori positive cases and the number of IHC stains ordered on a per case basis. For our subgroup analysis, the IHC usage rate and diagnostic rates of H. pylori of GI versus general surgical pathologists were compared. At our institution, GI biopsies are divided between a subspecialized GI pathology service and a general surgical pathology service according to a loadbalancing system that depends dynamically on several factors including daily case volume and optimization of resident education. We defined a GI pathologist (n=8) as a faculty member who has completed a GI pathology fellowship and who rotates on the subspecialized service.
Statistical Methods
Using R software, we calculated the Pearson product moment correlation coefficient to determine if diagnostic rate and IHC utilization rates among pathologists significantly correlated. The Welch t test was used to measure differences between the mean diagnostic rates and the mean IHC utilization rates within the GI versus general surgical pathologists subgroups.
RESULTS
Distribution of Cases and Pathologists
A total of 3751 cases matching our selection criteria were identified dispersed among 19 pathologists (Table 1). The median total case load was 127 and ranged from 41 to 518 per attending pathologist. GI pathologists (n=8) had a median of 311 cases and a range of 42 to 518 cases and 11 general surgical pathologists had a median of 120 cases and a range of