Validation of morphometric analyses of small-intestinal biopsy readouts in celiac disease

Assessment of the gluten-induced small-intestinal mucosal injury remains the cornerstone of celiac disease diagnosis. Usually the injury is evaluated using grouped classifications (e.g. Marsh groups), but this is often too imprecise and ignores minor but significant changes in the mucosa. Consequently, there is a need for validated continuous variables in everyday practice and in academic and pharmacological research. We studied the performance of our standard operating procedure (SOP) on 93 selected biopsy specimens from adult celiac disease patients and non-celiac disease controls. The specimens, which comprised different grades of gluten-induced mucosal injury, were evaluated by morphometric measurements. Specimens with tangential cutting resulting from poorly oriented biopsies were included. Two accredited evaluators performed the measurements in blinded fashion. The intraobserver and interobserver variations for villus height and crypt depth ratio (VH:CrD) and densities of intraepithelial lymphocytes (IELs) were analyzed by the Bland-Altman method and intraclass correlation. Unevaluable biopsies according to our SOP were correctly identified. The intraobserver analysis of VH:CrD showed a mean difference of 0.087 with limits of agreement from -0.398 to 0.224; the standard deviation (SD) was 0.159. The mean difference in interobserver analysis was 0.070, limits of agreement -0.516 to 0.375, and SD 0.227. The intraclass correlation coefficient in intraobserver variation was 0.983 and that in interobserver variation 0.978. CD3(+) IEL density countings in the paraffin-embedded and frozen biopsies showed SDs of 17.1% and 16.5%; the intraclass correlation coefficients were 0.961 and 0.956, respectively. Using our SOP, quantitative, reliable and reproducible morphometric results can be obtained on duodenal biopsy specimens with different grades of gluten-induced injury. Clinically significant changes were defined according to the error margins (2SD) of the analyses in VH:CrD as 0.4 and in CD3(+)-stained IELs as 30%.