HER-2/neu in breast cancer: interobserver variability and performance of immunohistochemistry with 4 antibodies compared with fluorescent in situ hybridization.

Abstract

The immunohistochemistry (IHC) performance of 4 anti-HER-2/neu antibodies was compared with fluorescent in situ hybridization (FISH) analysis of HER-2/neu gene expression in breast cancer patients considered for Herceptin (Trastuzumab) therapy. Interobserver variability in IHC interpretation was measured. Formalin-fixed tissue was received from 24 provincial hospital laboratories. The following anti-Her-2 antibodies were used: DAKO A0485 (polyclonal), Novacastra CB11 (monoclonal), Zymed TAB250 (monoclonal), and DAKO HercepTest (polyclonal). Additional sections were analyzed by FISH (Vysis). Three pathologists blinded to FISH results independently interpreted invasive tumor cell membranous staining on a scale of 0 to +3. The HER-2/neu gene was considered amplified when the FISH signal ratio of HER-2/CEP-17 was > or =2.0. Blocks from all hospitals and of all ages were suitable for IHC and FISH analysis. No interlaboratory analysis variability was noted. The interobserver agreement (kappa) for stain intensity for each antibody was good for 0 and +3 but poor for +1 and +2. Reasonable concordance between IHC and FISH was found with three of the four antibodies. TAB250 was the most sensitive antibody. For the three pathologists, the IHC sensitivities and specificities compared with FISH using 0/+1 as negative and +2/+3 as positive were as follows: A0485, 63-84/95-98; CB11, 63-66/97-98; TAB-250, 82-100/94-95; HercepTest, 59-77/91-93. The positive and negative predictive values varied by stain intensity. Stain scores of 0 and +3 were highly predictive of gene status. Stain scores of +1 and +2 were not sufficiently predictive to classify cases as amplified versus nonamplified. IHC is a reasonable first test to assess HER-2/neu status in patients with breast cancer. For most cases, DAKO A0485, TAB250, and HercepTest adequately predicted gene status. In cases with stain intensity of +1 or +2, the interobserver agreement is poor, and the predictive value is unsatisfactory for clinical use. Additional testing, preferably with FISH, is recommended.