The development of Herceptin (Trazumatab) makes testing for HER2 status important for choosing optimal therapy in breast cancer. This study addresses the precision, accuracy, and reproducibility of HER2 assays. HER2 was assessed retrospectively by immunohistochemistry (IHC) with Dako 'Herceptest', by IHC with the monoclonal antibody CB11, and by fluorescence in situ hybridization (FISH, PathVysion), in a series of 216 formalin-fixed breast carcinomas including 191 for which quantitative HER2 data from radioimmunohistochemistry (Q-IHC) were available. All tests were scored independently by two observers. Positivity rates varied between Herceptest (12.6%), FISH (19.4%), and CB11 IHC (28.5%). Kappa values showed that IHC-based tests were more susceptible to inter-observer variation (kappa=0.67 and 0.74 for Herceptest and CB11, respectively) than FISH (kappa=0.973). Overall test accuracy (see the Materials and methods section) for CB11 IHC (83.8%) was lower than Herceptest (87.4%) or FISH (93.2%). FISH predicted p185 HER2 overexpression (determined by Q-IHC) better (concordance index C.Ind. 0.90) than CB11 IHC (C.Ind.=0.85) or Herceptest (C.Ind.=0.81). Of 42 cases with gene amplification by FISH, 67% were positive in the Herceptest (2+ or 3+) vs. 83% with CB11. Of 174 cases negative by FISH, 96% were negative in the Herceptest and 68% with CB11. In conclusion, FISH is the most accurate, reproducible, and precise predictor of HER2 overexpression in routine diagnostic laboratories.
Copyright 2001 John Wiley & Sons, Ltd.