Automated FISH Analysis for the HER2 Gene Status Assessment in Breast Cancer: A Validation Study Based on 306 Cases
G DeMaglio, G Falconieri, S Pizzolitto. General University Hospital, Udine, Italy
Background: Automatic analysis of HER2 fluorescence-in-situ-hybridization (FISH) samples have been recently developed as a potential alternative to conventional gene assessment based on manual signal enumeration. However, specific studies addressing the evaluation of automated systems and their clinical validation are limited. We present a comparative study of traditional/manual versus software-assisted HER2 FISH scoring applied to routine breast cancer samples.
Design: HER2 FISH was investigated by analysis of 356 breast tumor samples obtained from routinely handled tissues, including core needle biopsies (n=188; 52.8%) and surgical specimens (n=168, 47.2%). FISH was performed using commercial probes (PathVysion®) and analyzed according to ASCO/CAP guidelines. The commercial software Metafer4 (Metasystems, Altlussheim, Germany) was used for automated analysis. For conventional evaluation, a tumor was considered to be amplified when the ratio of HER2 gene signals to chromosome 17 was greater than 2.2, non-amplified when the ratio was less than 1.8, and equivocal when the ratio was in the range 1.8-2.2. The Metafer4 PathVysion V2 classifier defines the range as 1.5-3.0 for borderline cases, and the ratio as less than 1.5 and greater than 3.0 for non-amplified and amplified cases, respectively.
Results: An average of 42 cells and 385 tiles for each sample was analyzed with manual and automated counting, respectively. Fifty tissue samples (50/356, 14.0%) with excess tissue autofluorescence and weak hybridization signals were excluded from the analysis. Manual and automated analysis was concordant for 286/306 cases (93.4%): discrepancies were retrospectively explained with suboptimal tumor area selection, HER2 tumoral heterogeneity or neoplasms with a ratio close to the equivocal range. The concordance rate was 97.7% and 91.8%, respectively, for non-amplified and amplified cases, whereas for equivocal cases, it was 57.1%. Borderline cases were 14/306 (4.5%) for manual counting and 21/306 (6.9%) for automated scoring. The concordance rate was near total (99%) for manual and automated scoring if equivocal cases were excluded.
Conclusions: Automated analysis of FISH samples for HER2 testing allows a faster and accurate study of a larger number of cells compared with that obtained by visual counting, especially for non-amplified and amplified cases. Although manual counting is still needed for equivocal cases, our data indicate that an automated approach to HER2 FISH scoring may be implemented in routine practice.
Tuesday, March 23, 2010 1:00 PM
Poster Session IV # 13, Tuesday Afternoon