부산대학교 도서관

8008 Background: Morphology remains the best biomarker for chemotherapy selection in NSCLC. Both bevacizumab and pemetrexed (in first line) are FDA indicated for non-SC histology only, making precise diagnosis key, although data are few on the reproducibility of their categorization. The standard of care, consensus-based World Health Organization diagnostic classification system, Ed. 4 (WHOC) includes 44 diagnostic categories (DC), does not include immunohistochemical (IHC) data, and is not independently validated. In this study, we estimate the reproducibility of SC versus non-SC classification to be only moderately precise.Methods: Pathologists (P) assigned WHOC diagnoses to virtual H&E slides from an incident surgical cohort of 96 primary lung tumors. At a power of 80% we estimated 24 P reviewing 48 slides each would allow for a reproducibility measure in the clinically useful range kappa=0.7, standard error=0.09. In addition to the 44 DC, we also considered the reproducibility of a simplified 10 category system and a chemotherapy-directed 2 class system.Results: 12 self-identified "lung experts" (E) and 12 community (C) P scored slides based on the 44 possible WHOC DC totaling 222 pathologist-pairs and 7130 slides-pairs. P rated confidence in their choices as: high 53%, somewhat 38%, and not confident 8%. 92% of slides were rated as having sufficient quality. Distribution of DC included SC 30% and adenocarcinoma 36%.Conclusions: To our knowledge this is the largest study ever performed of the WHOC, the only one to consider all 44 DC and to distinguish E vs C P. Although this model tests the reproducibility of the WHOC, not clinical lung cancer diagnoses, it shows that reproducible diagnosis of SC based on H&E morphology alone is inadequate. In the era of histology guided therapy, it highlights the need for strict diagnostic criteria and confirmatory IHC stains in the diagnosis of SC. [Table: see text] No significant financial relationships to disclose.