Expanding the Boundaries of Digital Pathology: Application of Image Analysis as a Diagnostic Support Tool for HER2 Interpretation
Z Sun, L Dobson, KJ Kaplan. Mayo Clinic, Rochester, MN; SlidePath, Dublin, Ireland
Background: With the development of trastuzumab, accurate evaluation of HER2 status has become particularly acute in the clinical laboratory. Although the recommended primary method of evaluation is immunohistochemistry (IHC), numerous reports of variability in interpretation have cast doubt on the reliability of HER2 testing, particularly in laboratories with a low-throughput of cases. In order to address this, recent clinical guidelines have suggested that image analysis could prove to be an effective and objective tool for achieving consistent interpretation, increasing the standard and reproducibility of assessment.
Design: This study evaluated 147 cases of invasive breast cancer with HER2 IHC stain by automated image analysis (AIA) and compared the evaluation of HER2 status with manual review by a pathologist and gene amplification status by FISH, the gold standard method of evaluation. The AIA measured several membrane stain parameters and obtained HER2 scores of 0/1 (negative), 2 (equivocal), and 3 (positive). Manual review score was made according to the ASCO/CAP scoring system where scores of 0-3+ were assigned. FISH positivity was determined by HER2/Chr17 ratio (positive >2.2, equivocal 1.8 to 2.2, and negative < 1.8).
Results: Results show that AIA could accurately determine HER2 protein expression with 94% agreement (κ = 0.73) with the pathology review across 147 cases. Eight negative cases were in complete agreement. The discrepancy was observed in 9 cases with equivocal 2+ score in either AIA or the manual review. Of the 3+ cases assigned by AIA, 95% were also found to be FISH positive. Those that were determined to be 2+ or equivocal for HER2 expression were found to be positive for gene amplification in 30% cases. Overall, an excellent concordance rate of 96% was observed between gene amplification status and AIA.
Conclusions: These results suggest that image analysis could concomitantly eliminate subjectivity and increase precision when assessing HER2 protein status in IHC stained tissue. Based on these findings, image analysis has great potential as a diagnostic support tool for pathologists and biomedical scientists, and may significantly improve the standardization of HER2 testing by providing a quantitative reference method for interpretation.
Monday, March 22, 2010 1:00 PM
Poster Session II # 170, Monday Afternoon