Abstract Purpose: Subtle changes in nuclear shape, size and texture precede the histological recognition of prostate cancer (PCa) and thus might provide a useful biomarker for high-risk benign tissue. The objective of this study was to develop and validate a multifeature nuclear score based on direct DNA staining and test its association with field effects and the subsequent detection of PCa in benign biopsies. Methods: Whole tissue sections from 39 radical prostatectomy subjects were Feulgen-stained and scanned (400x) on a digital microscope, allowing maps of DNA content per nuclear pixel to be generated in Matlab@. Samples of PCa and benign epithelial nuclei were randomly selected for measurement of 51 morphometric features. Logistic regression models for discriminating benign from PCa nuclei were built and cross-validated by AUC analysis. Subject-level models, for discriminating benign from malignant nuclear populations taken from individual subjects, were also built and similarly validated. Both backwards elimination and all-subsets approaches were used to select covariates. Nuclear populations were randomly collected < 1mm or > 5mm from cancer foci, and from cancer-free prostates, HGPIN, and PCa Gleason grade 3-5. In addition, nuclear images were collected from negative biopsy cases and controls matched on age and date of biopsy (20 pairs); cases had PCa detected at least 2 years later, controls remained cancer-free and had at least 2 additional negative biopsies. For the case-control study, nuclei were selected using an automatic process that correlated well with results from manual selection. Frequency distributions for multifeature nuclear score from various tissue compartments were compared using the Kolmogorov D statistic. Scores for cases vs. controls were compared using a t test for paired data. Results: In prostatectomy samples, both nuclear- and populationlevel models revealed cancer-like features in benign nuclei adjacent to PCa, compared to nuclei that were more distant or were obtained from PCa-free glands. The score distribution for distant nuclei was also significantly shifted compared to nuclei from cancer-free glands. Interestingly, the best model by leave-one-out cross-validation (AUC = 0.91, 95% CI: 0.81-1.00) included only 5 variance features, reflecting the importance of nuclear pleomorphism. In negative biopsies, a validated model with 5 variance features (including features related to DNA content, size and texture) yielded significantly higher scores in cases than controls (P=0.026). This model had an AUC = 0.71 (95% CI: 0.54-0.87) for discriminating cases from controls. Conclusions: A multifeature nuclear morphometric score, obtained by automated digital analysis, was validated for discrimination of benign from cancer nuclei. This score demonstrated field effects in benign epithelial nuclei at varying distance from PCa lesions, and was associated with subsequent PCa detection in negative biopsies. This biomarker shows definite promise as an intermediate endpoint in chemoprevention trials; however, its utility as an independent predictor of PCa in the clinical setting requires much further study. Citation Information: Cancer Prev Res 2010;3(12 Suppl):B38.