Grading of Lung Adenocarcinoma: Architectural Versus Nuclear Approach
CG Deshpande, K Geisinger, I Petersen, A Moreira, M Zakowski, R Shen, J Suh, N Motoi, W Gerald, M Kris, V Rusch, WD Travis. Memorial Sloan-Kettering Cancer Center, New York, NY; Wake Forest University, Winston-Salem, NC; Universittsklinikum Jena, Germany; Stonybrook Stony Brook University Medical Center, Stonybrook, NY
Background: Well established grading systems exist for carcinomas arising in the breast, prostate & kidney. However, there is no widely accepted detailed grading system for non-small cell lung carcinoma including adenocarcinoma.
Design: We reviewed 98 lung adenocarcinomas to evaluate architectural & nuclear (size, nucleoli, pleomorphism & mitoses) approaches to grading in tissue sections. Architectural grading was performed using semiquantitation of histologic subtypes and with classification by major subtype. Bronchioloalveolar (BAC) was compared to solid and a category of glands & papillae with a combination of acinar/papillary/micropapillary patterns. A subset were evaluated by 5 pathologists for reproducibility analysis with Fleiss kappa. Statistical correlations were made between grading features and survival, EGFR and KRAS mutation data & RNA gene expression profiling data (Affymetrix, HG-U133) using R software and SPSS v 16.
Results: Architectural grading showed moderate reproducibility (0.54, N=44) and significant prognostic correlation with solid vs BAC/acinar/papillary/micropapillary tumors demonstrating a 46% vs a 64% 5-year survival (p=0.014). Nuclear grading showed slight reproducibility for nuclear size (0.06, N=44), pleomorphism (0.09, N=45) and nucleoli (0.14, N=45). Mitotic count showed fair reproducibility (0.36, N=45). No survival correlations were found with nuclear features. All 15 EGFR mutations were found in acinar/papillary/micropapillary tumors 15/67 (22.4%) rather than BAC (0/8) or solid (0/22) tumors (p=0.014, Fisher exact test). Solid pattern was associated with gene expression profiling Cluster 3 (p=0.001) and BAC was associated with Cluster 2 (p=0.005). No correlation was found with KRAS mutation.
Conclusions: Architectural grading largely based on histologic subtyping shows moderate reproducibility and is useful for predicting prognosis. It also demonstrates significant correlations with EGFR status & gene expression profiling clusters. Nuclear grading shows only slight reproducibility & we could not find any survival or molecular correlations. Further work is needed to determine if grading (eg. combined architecture & nuclei) provides any advantage to histologic subtyping with predicting prognosis.
Monday, March 9, 2009 1:00 PM
Poster Session II # 224, Monday Afternoon