Microvessel area using automated image analysis is reproducible and is associated with prognosis in breast cancer

Abstract

Microvessel density may be one measure of tumor associated angiogenesis but is methodologically difficult to standardize and reproduce. We used our automated quantitative image analysis system, AQUA, to more objectively assess microvessel area. Cytokeratin and CD31 were used to create tumor and vessel compartments respectively with AQUA. Microvessel area was defined as CD31 compartment area normalized to the tissue spot area (CD31 area/area of entire tissue spot). Consecutive breast cancer whole sections were stained with CD31 to compare pathologist-based microvessel density with AQUA microvessel area. Microvessel areas of 3-fold redundant tissue microarrays of 652 primary breast cancers were also assessed. CD34 and factor VIII-related antigen were also tested. There was nearly linear correlation between pathologist's microvessel density and AQUA microvessel area with regression coefficient R = 0.846. On the redundant arrays, of the 67% evaluable cases, 52% were microvessel area high and 48% low with good reproducibility of scores (Spearman rho 0.551). AQUA microvessel area was associated with larger tumors, node positivity, and estrogen receptor negativity, with 20 year survival at the univariate and multivariate levels (P < .0001 and P = .0121, respectively). CD34 or factor VIII-related antigen were more heterogenous, had poor association with CD31, and did not correlate with outcome. AQUA-based microvessel area was significantly correlated with both standard breast cancer prognostic parameters as well as with clinical outcome. In the future, it may also allow the use of the AQUA-based algorithms to quantify the expression of angiogenic biomarkers to either tumor or microvessel area-specific compartments.