Computer-assisted analysis of biopsy specimen microvessels predicts the outcome of esophageal cancers treated with chemoradiotherapy.

PURPOSE A computer-assisted microvessel analysis system was developed to evaluate correlations between the architecture of biopsy specimen microvessels and the outcome for patients with esophageal cancer treated with chemoradiotherapy. EXPERIMENTAL DESIGN Biopsy specimens from 51 patients with esophageal cancer (T(2-3), any N, M0) treated with chemoradiotherapy were immunostained with an anti-CD31 antibody and quantified using computerized image analysis. We evaluated the association of several microvessel factors with overall survival, including the ratio of total microvessel perimeter to total tumor area (TP/TA), the tumor hypoxic ratio, and the ratio of total microvessel number to total tumor area (TN/TA). Results from traditional manual microvessel density (MVD) hotspot count and computerized hotspot count were compared and the relation between hotspot MVD count and survival rate was evaluated. RESULTS The median follow-up time was 32 months. Both univariate and multivariate analyses revealed that computer-counted hotspot MVD and TN/TA and TP/TA ratios correlated significantly with the outcome of chemoradiotherapy. Kaplan-Meier survival curves showed that patients with high computer-counted hotspot MVDs and high TN/TA and TP/TA ratios had better overall survival rate than patients with low MVDs or ratios (P = 0.025, 0.008, and 0.031, respectively). Combining computer-counted MVD or TN/TA ratio with TP/TA ratio proved more predictive than any single factor. Two researcher-counted hotspot MVDs had no significant relation with outcome. CONCLUSION Computer-assisted tumor microvessel analysis is a powerful tool in predicting the outcome for patients with esophageal cancer treated with chemoradiotherapy because intraobserver and interobserver variability is minimized.