Background: Microvessel counting has proven to be of prognostic value in breast cancer, as shown in different retrospective studies. However, methodology has not been studied widely, and this must be done before the method can become clinically applicable. The aim of this study was to determine the degree of heterogeneity and reproducibility of microvessel counts (MC) in breast cancer.
Experimental design: In 10 cases of breast cancer, the available blocks (2-4 blocks) containing invasive tumor parts were selected, and four sections (4 microns) were cut with interdistances of 100 microns. In each section, two or three invasive areas (0.5 x 0.5 cm) were demarcated. Microvessels, visualized by immunohistochemistry (CD31 Ab), were counted by one observer in 10 systematically selected fields of vision (400 x magnification). Furthermore, microvessels were counted in four fields with the highest microvessel density ("hot spots"). Coefficients of variation (CV) were calculated for the different sampling levels.
Results: Repeated MC yielded high intra- and interobserver reproducibility (correlation coefficients > 0.92). For the systematic counting method, CV between MC from different areas within one section was on average (17.1% (0.7-52.1). When comparing MC from corresponding areas in different sections from the same black, CV was on average 14.7% (0.5-41.9), and for MC from different blocks of the same tumor, CV was on average 25.8% (9.9-44.6). Nested ANOVA showed an approximately equal contribution to the total variance of the different sampling levels, except for the variation between sections (not significant). For the hot spot MC, CV for different sections from the same block was on average 11.1% (0.7-29.5) and for different blocks from the same tumor, 24.2% (5.7-54.9). Nested ANOVA showed that variation between different blocks from the same tumor contributed most to the total variance.
Conclusions: There is a noteworthy heterogeneity in MC between different areas from the same section, between corresponding areas in different sections from the same block, and between different blocks from the same tumor. Consequently, one must carefully scan all the available tumor material in each case for the best hot spot. The hot spot approach is efficient and reproducible, but only a comparative prognostic evaluation can show whether it is clinically more useful than systematic counts.