Crypt Apoptotic Count Reproducibility in Small Bowel Allograft Biopsies.
Geoffrey Talmon, Radwa El Behery, Stanley Radio, James Wisecarver, Valerie Shostrom. University of Nebraska Medical Center, Omaha
Background: Enumeration of crypt apoptotic bodies (AB) is one feature in the schema for the histologic diagnosis of acute cellular rejection (ACR) in small bowel allografts, with ≥6 AB in the proper background qualifying as mild ACR. Similar cases with fewer AB are often called indeterminate. When a quantified variable is used for diagnosis, clear definition and reproducibility are important. Some consider only well-formed “classical” exploding crypt cells as AB while others are more “liberal” in their descriptions, including clusters of pyknotic debris. Due to differing definitions, there is a potential for variation in AB numbers between pathologists that may impact diagnosis. To our knowledge, this is first study specifically examining intra- and interobserver reproducibility in AB counts.
Design: Thirty mucosal biopsies from small bowel allografts were obtained, selected by original diagnosis: 10 screening biopsies negative for ACR, 10 indeterminate, and 10 mild ACR. Three transplant pathologists blindly reviewed each case twice with randomization before each read. They were given written and pictorial descriptions of classical and liberal AB and counted the maximal number per 10 epithelial cells. A difference of >2 AB qualified as discrepant. “Diagnoses” (assuming the proper histologic background) were characterized by AB count as follows: negative-0-2, indeterminate-3-5, and mild ACR-≥6.
Results: Intraobserver kappa values for overall classical and liberal AB counts ranged from 0.749-0.914 and 0.814-0.933, respectively and interobserver kappa values were 0.478 and 0.686 (P<.05, each). Intraobserver diagnosis agreement based on classical AB number occurred in 79 of 90 instances (88%) and in 57 of 90 using the liberal definition (63%, P<.01). Total interobserver diagnosis agreement using an average of both individuals' counts based on the classical AB description was reached in 9 of the 10 negative, indeterminate, and mild ACR cases. Using the liberal definition, total diagnostic agreement occurred in 9 negative, 6 indeterminate, and 7 mild ACR cases (P=0.13).
Conclusions: There was substantial intraobserver and moderate to substantial interobserver agreement in overall crypt AB counts using classical and liberal definitions. Slightly better agreement was seen using the latter. However, both intra- and interobserver variability would have led to diagnostic discrepancies in a small subset of cases. Better diagnostic agreement was achieved when counting classical AB.
Tuesday, March 1, 2011 1:00 PM
Poster Session IV # 58, Tuesday Afternoon