Can sparsely and heterogeneously expressed proteins be detected using tissue microarrays? A simulation study of the...

Tissue microarrays (TMAs) are often used to evaluate the expression of biological markers across large patient populations. We investigated the number of core biopsies required to accurately classify soft tissue sarcomas with respect to their expression of carbonic anhydrase IX (CA IX), a potential prognostic marker with a sparse and heterogeneous expression pattern. Paraffin-embedded tissue sections from 47 high-grade soft tissue sarcomas had previously been immunostained for CA IX and quantified using image analysis. Sampling of core biopsies was computer simulated by analyzing 2 mm-diameter disc-shaped areas from these images. Core areas were scored as CA IX-positive if the number of CA IX-positive pixels exceeded a threshold value. Overall, 94% of the tumors were correctly classified as CA IX-positive (true-positive rate) if four biopsies at the quadrant centers of each of two sections per tumor were assessed for their CA IX expression with a sensitivity threshold of 0.2%. Similarly, 91% of the tumors were correctly classified as CA IX-positive when only three of the four biopsies per section were assessed. The corresponding false-positive rates were 13% and 11%, respectively. In contrast, sampling of four biopsies at random positions on each section with otherwise similar sampling criteria resulted in only 81% of the tumors being scored as CA IX-positive with a probability >0.80. In conclusion, soft tissue sarcomas can be accurately classified with respect to their expression of CA IX, a protein with sparse and heterogeneous expression, by systematic sampling of at least three 2mm-diameter core biopsy areas from distant locations within a section and from two different areas per tumor. This suggests that TMAs can be used to detect even heterogeneous and sparse markers, but the sampling strategy should be validated for the marker under investigation.