Establishment of digital cutoff values for intraepithelial lymphocytes in biopsies from colonic mucosa with lymphocytic colitis

Abstract

Background/IntroductionLymphocytic colitis (LC) and the incomplete form (LCi) are common causes of chronic watery diarrhea. Endoscopy is often inconspicuous, and the diagnosis relies on histopathological assessment of colonic biopsies. Digital Image Analysis (DIA) eliminates interobserver variation. The aim of this study was to establish digital cutoff values for LC and LCi on CD3 stained slides. Material and methodsOne hundred and six patients with a hematoxylin and eosin (HE) diagnosis of normal colonic mucosa (N = 19), non-specific reactive changes (N = 24), LCi (N = 23) and LC (N = 40) were eligible for analysis. The number of intraepithelial lymphocytes (IELs) reached by DIA in the total surface epithelium and in hot spots of the biopsies was compared with the diagnostic category assigned by the pathologists based on HE stained slides. The digitalized slides were analyzed for number of IELs using Visiopharm Quantitative Digital Pathology software. All digitalized slides were examined manually to identify differences in the approach to the evaluation of the biopsies by the pathologists and DIA. ResultsThe median IEL counts and interquartile range in the total surface epithelium were 3.6 (3.2–4.3), 4.4 (3.4–5.3), 19.8 (16.6–30.0) and 41.3 (37.0–47.8) in normal colon mucosa, mucosa with non-specific reactive changes, LCi and LC, respectively. Discrimination between normal mucosa and non-specific reactive changes was not possible. Digital cutoff values with the best separation between non-LC, LCi and LC were > 13 IELs/100 epithelial cells for LCi and > 36 IELs/100 epithelial cells for LC. These cutoff values resulted in an agreement between the pathologist’s and DIA that was very good with a kappa value of 0.90. ConclusionDespite differences among the approach of DIA and the pathologist’s assessment of IELs in colonic mucosa DIA is able discriminate between the HE based diagnoses of the three subgroups non-LC, LCi and LC with high accuracy.