Assessment of Immunohistochemical Protein Expression in Tissue Microarrays Using Virtual Microscopic Image Analysis
TA Summers, GT Clifton, JD Gates, L Dobson, I Horkayne-Szakely, JC Shaw, A Nissan, GD Sandberg, GE Peoples, A Stojadinovic. Walter Reed Army Medical Center, Washington, DC; Brooke Army Medical Center, San Antonio, TX; SlidePath, Dublin, Ireland; Armed Forces Institute of Pathology, Washington, DC; Hadassah University Hospital Mount Scopus, Jerusalem, Israel
Background: Interpretations of immunohistochemical (IHC) staining parameters are subjective and prone to inter- and intra-observer variability. This is highlighted by studies demonstrating that approximately 20% of HER2 analyses performed at primary treatment sites were incorrect when re-evaluated retrospectively. Virtual microscopy digitizes glass slides, therefore creating digital image environments for subsequent computer assisted interpretation of pathologic information. We provide additional evidence that Image Analysis (IA) can provide an objective analysis of IHC stain expression.
Design: Tissue microarrays (TMAs) were assembled from 160 Stage IV colon cancers metastatic to the liver (160 primary specimens, 160 metastatic liver specimens), 65 matched metastatic lymph nodes specimens, 75 normal colon specimens, and 42 transitional area specimens. TMAs were stained with MSH2 and p53. Stain expression analysis was performed by two pathologists independently and a separate uncalibrated IA was conducted using nuclear stain detection algorithms (SlidePath; Dublin, Ireland).
Results: 885 total specimens were analyzed (449-MSH2 and 436-p53). Agreement of stain expression amongst pathologists and IA was achieved for 82.3% of the specimens (729-Total; 367-MSH2, 362-p53). The individual pathologists' results were incongruent with IA in 11.6% (κ=0.767, p<0.0005) and 14.9% (κ=0.702, p<0.0005) of samples. Stain expression results between the pathologists and IA, when the pathologists agreed (807-Total), were incongruent in 9.7% of the specimens (78-Total; 37-MSH2, 41-p53; κ=0.81, p=0.054). The pathologists disagreed over stain expression for 8.8% of the specimens (78-Total; 44-MSH2, 34-p53; κ=0.83, p<0.0005). In the subset of specimens where pathologists disagreed, IA detected no stain expression for 80.7% of specimens (63-Total; 42-MSH2, 21-p53) suggesting over interpretation of minimal staining by the pathologists.
Conclusions: This uncalibrated study of protein expression by IA further highlights the potential of IA to increase objectivity in IHC stain interpretation and provides additional evidence that IA may fill the need for a standardized and objective assessment of clinically relevant IHC stains.
Monday, March 22, 2010 1:00 PM
Poster Session II # 168, Monday Afternoon