Abstract
Background: Celiac disease clinical trials employ histologic endpoints to demonstrate drug efficacy. Once therapeutics enter the market, insurance payors may require demonstration of persistent histologic injury in duodenal biopsies before approving payment for life-long medications. In that scenario, pathologists may be asked to compare the injury between diagnostic and follow up biopsies, with the result determining coverage decisions. While interobserver agreement among pathologists on the diagnosis of celiac disease is high, agreement on grade of injury is modest to low in the literature. This study was undertaken to establish the minimal perceptible difference in villous architecture between two celiac biopsies that can be perceived by pathologists, using routine stains without software tools. A secondary aim was to explore the histologic factors that contribute to pathologists’ grading decisions.
Methods:10 GI pathologists from 10 academic medical centers compared histologic findings between 20 pairs of digitally scanned hematoxylin and eosin (H&E) stained celiac duodenal biopsies. Previously measured quantitative villous/crypt (VhCd) measurements served as ground truth, with differences in VhCd between the two biopsies within the pairs ranging from 2.81-0.46um. Prior to scoring, a virtual live training session occurred to align on how to recognize well oriented VhCd units. Based on villous architecture in each biopsy, pathologists were instructed to designate which biopsy in the pair had more enteropathy. A designation of no difference was allowed. A free text field for comments was included.
Results: In 12/20 (60%) pairs at least 8/10 pathologists correctly identified the biopsy with lower VhCd (Table 1). While smaller differences in VhCd were significantly associated with fewer correct responses (p<0.01), all 10 pathologists correctly assessed the biopsy pair with the smallest VhCd difference. Comments revealed that in equivocal pairs, IELs differences were used to select the biopsy with more enteropathy, and the finding of lower IELs in the biopsy with lower VhCd was associated with more incorrect answers. Further analysis revealed that 22/40 (55%) of biopsies had only 1 well-oriented VhCd unit.
Conclusion:Using routine H&E evaluation without quantitation, the minimal detectible VhCd difference that pathologists can detect between two celiac duodenal biopsies was not reached in this series. Disagreements between pathologists’ assessment and software measured VhCd suggest that other factors, including IEL density and poor biopsy orientation may confound the ability to determine which of two biopsies shows more enteropathy. A larger study is underway to clarify this challenge in the era of emerging celiac therapeutics in which pathologists may play a role in drug eligibility.
Authors
- Robert et al.