Gene expression microarrays are being used to develop new prognostic and predictive tests for breast cancer, and might be used at the same time to confirm oestrogen-receptor status and ERBB2 status. Our goal was to establish a new method to assign oestrogen receptor and ERBB2-receptor status to breast carcinoma based on mRNA expression measured using Affymetrix U133A gene-expression profiling.

We used gene expression data of 495 breast cancer samples to assess the correlation between oestrogen receptor (ESR1) and ERBB2 mRNA and clinical status of these genes (as established by immunohistochemical [IHC] or fluorescence in-situ hybridisation [FISH], or both). Data from 195 fine-needle aspiration (FNA) samples were used to define mRNA cutoff values that assign receptor status. We assessed the accuracy of these cutoffs in two independent datasets: 123 FNA samples and 177 tissue samples (ie, resected or core-needle biopsied tissues). Profiling was done at two institutions by use of the same platform (Affymetrix U133A GeneChip). All data were uniformly normalised with dCHIP software.

ESR1 and ERBB2 mRNA levels correlated closely with routine measurements for receptor status in all three datasets. Spearman’s correlation coefficients ranged from 0·62 to 0·77. An ESR1 mRNA cutoff value of 500 identified oestrogen-receptor-positive status with an overall accuracy of 90% (training set), 88% (first validation set), and 96% (second validation set). An ERBB2 mRNA threshold of 1150 identified ERBB2-positive status with the overall accuracy of 93% (training set), 89% (first validation set), and 90% (second validation set). Reproducibility of mRNA measurements in 34 replicate experiments was high (correlation coefficient 0·975 for ESR1, 0·984 for ERBB2).

Amounts of ESR1 and ERBB2 mRNA as measured by the Affymetrix GeneChip reliably and reproducibly establish oestrogen-receptor status and ERBB2 status, respectively.