Counting intraepithelial lymphocytes (IELs) is a key part of the assessment of duodenal biopsies. Immunohistochemistry (IHC) for CD3 can aid identification of lymphocytes in this context, but it is not evident that counts on hematoxylin and eosin (H&E) and CD3 are comparable. This study aimed to compare the IEL counts in duodenal biopsies using H&E stains and CD3 IHC, and to examine the interobserver variability. Thirty-five paired H&E and CD3 sections were reviewed by 6 pathologists who counted the number of IELs per 100 enterocytes. The counts were categorized into groups: normal (<25 lymphocytes), mildly raised (25-40 lymphocytes), and markedly raised (>40 lymphocytes). CD3 IHC was associated with significantly higher IEL counts than H&E. Four cases with normal H&E counts had raised counts with CD3. There was moderate agreement between observers for both H&E and CD3. Lack of concordance between CD3 and H&E IEL counts suggests that counts derived from the 2 methods may not be comparable to each other and should not be considered equivalent. There was no significant improvement in interobserver variability with CD3 IHC.

There is an increase in T cells within the epithelium of the small intestinal mucosa in celiac disease, and an important diagnostic marker of this condition is an increased intra-epithelial lymphocyte (IEL) count. Standard practice is to use hematoxylin and eosin (H&E) to identify IELs in histologic sections. Immunohistochemical stains such as CD3 may make the identification and counting of IELs easier, and it has been suggested that CD3 could be particularly valuable in borderline cases. However, this raises the question of the extent to which IEL counts made using H&E and CD3 agree with each other. Duodenal biopsies from 79 patients were analyzed. For each case, two slides were prepared: one was stained with H&E and another was immunostained for CD3. Four well-oriented villus tips with a high IEL density were selected on each slide and the IELs per 200 enterocytes counted with results expressed as IELs/100 enterocytes. Although there was a very strong, positive correlation between IEL counts made using H&E and CD3 (Spearman r = 0.96, P < .001), the level of agreement as evaluated using Bland Altman plots was poor. CD3 immunostaining resulted in higher counts of IELs/100 enterocytes than H&E (mean difference = 8 IELs) with wide limits of agreement (-10 to 26). For IEL counts greater than 30, the mean difference between counts made using CD3 and H&E increased to 12 IELs (limits of agreement -10 to 35). In conclusion, agreement between H&E and CD3 is too poor for the stains to be considered interchangeable, particularly at higher IEL counts. Counting IELs using CD3 immunohistochemistry could overdiagnose healthy patients as having Marsh grade 1 lesions if the criteria for sections stained with H&E are used.