Application of the New Proposed Adenocarcinoma Classification – A Reproducibility Study
Jason M Wells, Sanjay Mukhopadhyay, Haresh Mani. Penn State Milton S. Hershey Medical Center, Hershey, PA; SUNY Upstate Medical University, Syracuse, NY
Background: The proposed IASLD/ATS/ERS adenocarcinoma classification document has stressed the need for further validation. The aim of this study was to assess whether the determination of tumor pattern is reproducible among pulmonary pathologists in routine daily practice.
Design: All slides (median 4, range 1-9) from 75 consecutive resected T1 adenocarcinomas were assigned a primary and secondary pattern by 3 pathologists with pulmonary subspecialty training and at least 5 years' experience, practicing at different institutions. A pre-study training set was not used. Results were analyzed for reproducibility.
Results: The most common predominant patterns were acinar (31%) and solid (27%), as were the most common secondary patterns (acinar 40%, solid 13%). The most common combination of predominant and secondary patterns was acinar/solid (25%). The kappa value for interobserver agreement was 0.32 for predominant pattern, 0.26 for secondary pattern and 0.26 for the combination of primary/secondary patterns. All 3 readers agreed on a predominant pattern in only 26/75 (33%) cases; there was complete disagreement in 19/75 (25%) cases. The most common area of discordance was in assigning acinar vs. lepidic predominant pattern (15 cases). Reproducibility was best for solid pattern (k=0.65), followed by micropapillary (k=0.35), lepidic (k=0.28), papillary (k=0.2) and acinar (k=0.08) patterns. The number of slides per case did not significantly impact reproducibility (k=0.33 for cases with up to 4 slides and 0.28 for cases with more than 4 slides). Kappa values were similar (k=0.3 to 0.39) when comparing each pair of reviewers.
Conclusions: Inter-observer reproducibility in the determination of predominant and secondary patterns in resected pulmonary adenocarcinomas is poor, even among pulmonary pathologists, when applied to evaluating actual slides of resected tumors. Previously reported high kappa correlations based on representative images may not apply to routine everyday practice. The inability to reproducibly distinguish lepidic vs. acinar patterns, in particular, adversely impacts the reliable diagnosis of adenocarcinoma in situ.
Monday, March 4, 2013 2:45 PM
Proffered Papers: Section D, Monday Afternoon