Interobserver Variability in Assessing High-Grade Dysplasia and Architecture in Colorectal Adenomas: A Multi-Centre Canadian Study
Allison Osmond, Hector Li Chang, Richard Kirsch, Dimitrios Divaris, Vincent Falck, Dong Feng Liu, Celia Marginean, Ken Newell, Jeremy Parfitt, Brian Rudrick, Heidi Sapp, Joanna Walsh, Sharyn Smith, Fasahat Wasty, David Driman. Western University, London, Canada; University of Toronto, Toronto, Canada; Grand River Hospital, Kitchener, Canada; University of Calgary, Calgary, Canada; Woodstock General Hospital, Woodstock, Canada; University of Ottawa, Ottawa, Canada; Grey Bruce Health Services, Owen Sound, Canada; Dalhousie University, Halifax, Canada; Huron Perth Health Alliance, Stratford, Canada; St. Thomas Elgin General Hospital, St. Thomas, Canada
Background: Colorectal cancer screening programs have been introduced throughout Canada. Standardized diagnosis of colorectal adenomas (CRAs) is required for risk stratification, establishing surveillance intervals, and analysis of epidemiologic data. Guidelines were developed by a pan-Canadian expert panel to standardize reporting of colorectal polyps. The aims of the present study were a) to assess interobserver variability in the classification of dysplasia and architecture in CRAs, and b) to determine if interobserver variability could be improved by the adoption of criteria specified in the guidelines.
Design: An a priori power analysis was used to determine an adequate number of cases and participants. Twelve pathologists, 6 in community practice and 6 in academic practice, independently classified 40 whole slide images of CRAs according to architecture (tubular, tubulovillous, villous) and dysplasia grade (low (LGD), high (HGD)). After a washout period of 30 days, participants were provided with the consensus guidelines and asked to reclassify the study set.
Results: At baseline there was moderate interobserver agreement for both architecture [K= 0.4700; 95% CI 0.4427, 0.4972] and dysplasia grade [K= 0.5680; 95% CI 0.5299, 0.6062]. Following distribution of the guidelines, there was improved interobserver agreement in assessing architecture [K = 0.5403; 95% CI 0.5133, 0.5674)]. For dysplasia grade, overall interobserver agreement remained moderate but decreased significantly [K = 0.4833; 95% CI 0.4452, 0.5215], with the decrease in agreement being greater amongst pathologists in community practice. Half of the cases in the study contained HGD. Six pathologists diagnosed HGD in 25-40% of cases, 4 diagnosed HGD in 41-50% of cases, and 2 diagnosed HGD in ≥75% of cases.
Conclusions: CRAs can be classified with only a moderate level of interobserver agreement. Adoption of the consensus guidelines resulted in improved interobserver agreement in assessing polyp architecture, but decreased agreement in grading dysplasia. The variability in the diagnosis of HGD, even following guideline review, suggests the need for more concerted knowledge-transfer exercises.
Tuesday, March 5, 2013 1:00 PM
Poster Session IV # 136, Tuesday Afternoon