This query included any specimen labeled with the term duodenum, duodenal, small bowel, or small intestine and excluded any specimen that contained the word aspirate or aspiration so as to exclude fluid analysis from the dataset. For individuals who underwent more than one examination during this period, we included only the first chronological examination. Because the primary aim was to assess biopsy practices in patients without known CD, we excluded any patient with a known history of CD as described in the clinical indication field. To determine the number of duodenal biopsy specimens for

each biopsy set, we used a free-text search of the pathologist’s description of each sample. When present, Olaparib ic50 specimens from the duodenal bulb (identified either in the endoscopist’s report or the histopathologic interpretation) were included in the total number of specimens submitted. Cases in which the number of specimens submitted was not quantified (either not stated or characterized as multiple) were excluded. We used the chi-square test to assess the association between adherence to the recommendation of submitting ≥4 specimens and the proportion of patients with pathological findings

consistent with CD. Because this dataset did not contain information on serological findings or follow-up clinical information, we defined Selleck HIF inhibitor a priori having a result of either blunted villi (Marsh IIIA) or flat villi (Marsh IIIB/C) as meeting the IMP dehydrogenase pathological definition of CD. For assessing the relationship between ordinal categories such as year or number of specimens and the pathologic diagnosis of CD, we used the Cochran-Armitage test for trend. Given the possibility that gross endoscopic findings may be associated

with both the number of specimens submitted and the probability of CD, we investigated the relationship between adherence to submitting ≥4 specimens and CD while stratifying by gross endoscopic findings. We used the Breslow-Day test for homogeneity of odds ratios (OR) so as to assess whether gross appearance modifies this association. Generalized estimating equation multivariate logistic regression was used to determine factors associated with the submission of ≥4 specimens, adjusted for clustering by individual provider. We postulated that adherence to this practice was correlated with individual providers. Using the generalized estimating equation in this multivariate analysis takes such clustering into account when the variance of hypothesized associations is estimated. We used SAS version 9.1 (Cary, NC) for all statistical calculations. All P values presented are 2-sided. The Institutional Review Board of Columbia University Medical Center evaluated this study protocol and designated it as “non-human subject research” involving de-identified records. A total of 150,155 procedures involving a duodenal/small-bowel biopsy were submitted for histopathologic evaluation during the 4-year period.