Interobserver Variability in Determining Carcinoma Extent Based on Percentage in Radical Prostatectomy (RP) Specimens
O Hameed, VL Dailey, HA Al-Ahmadie, MB Amin, TA Bismar, L Cheng, DE Hansel, PA Humphrey, JK McKenney, GJ Netto, GP Paner, J Ro, RB Shah, S Shen, K Trpkov, M Zhou. University of Alabama at Birmingham, Birmingham; Memorial Sloan-Kettering Cancer Center, New York; Cedars-Sinai Hospital, Los Angeles; University of Calgary, Calgary, Canada; Indiana University, Indianapolis; Cleveland Clinic, Cleveland; Washington University, St. Louis; Stanford School of Medicine, Stanford; Johns Hopkins University, Baltimore; Loyola University, Maywood; The Methodist Hospital, Houston; University of Michigan, Ann Arbor
Background: The International Society of Urological Pathology recommends that some measurement of tumor extent be provided when reporting carcinoma in RP specimens. Percentage of tumor involvement by visual estimation is one such assessment that has been found to correlate well with outcome. However, there are no systematic studies that adequately evaluate the interobserver variability in reporting tumor extent based on such percentages.
Design: Replicate slides (n=108) of 10 partially-embedded RP specimens (9-15 slides/case) were sent to 14 genitourinary pathologists. Participants estimated the percentage of prostatic tissue involved by carcinoma in each section and used the average to derive the total percentage of prostate involved by tumor. These percentages, as well as another set derived from image analysis of the same slides, were then broken down into 4 categories (≤5%, 6% to 20%, 21% to 50% and > 50%) which have been found to correlate well with outcome (Urol. 2008;180:571). Data were then used to evaluate agreement.
Results: All 14 (100%) participants agreed on the extent category in 5 cases, 13 (93%) agreed in 2 cases, and 12 (86%), 11 (79%) and 10 (71%) agreed on each of the remaining 3 cases, respectively. Disagreements were more frequent at both ends of the spectrum; nevertheless, the overall agreement rate was 86% and the overall kappa was 0.82. There was also excellent agreement between the participants' interpretations and those derived from image analysis with 3 (21%), 5 (36%), and 6 (43%) participants achieving 80%, 90%, and 100% agreement rates, respectively, and pair-wise kappa values ranging from 0.73 to 1.0.
Conclusions: Percentage of prostate gland involvement by carcinoma can be converted into reproducible and accurate extent categories. Additional data to follow.
Category: Genitourinary (including renal tumors)
Monday, March 22, 2010 1:00 PM
Poster Session II # 127, Monday Afternoon