We previously reported that reactive stroma grading in prostate cancer (PCa) is predictive of biochemical recurrence in prostatectomies and biopsies. In this study, we tested whether quantifying the percentage of reactive stromal grade 3 (RSG 3; stromogenic carcinoma pattern) in the entire tumor is predictive of PCa-specific death. Whole-mount prostatectomies operated by a single surgeon obtained between 1983 and 1998 were reviewed. Reactive stroma was evaluated as described previously, and areas of RSG 3 in the entire tumor were registered as percentages of total tumor. Statistical analysis was performed using Spearman, Kaplan-Meier, and Cox analyses. In all, 872 cases were Prostate cancer (PCa) remains the most common malignancy affecting men and is the second-leading cause of cancer-related death of men in the United States. 1,2 The most accepted and internationally used histological classification to predict the behavior of PCa is the Gleason scoring system. 3-7 Tools such as the Kattan nomogram or the Partin tables combine predictive information to optimize the prediction of time to biochemical recurrence (BCR). 8 -11 However, all of these histological parameters examine the epithelial component of the cancer, and not the interaction between the tumor and the host.The most common histological interaction between the tumor and the native host stroma in many tumors is desmoplasia, or reactive stroma (RS). Reactive stroma in PCa was described in our previous studies and characterized at the phenotypic, 12-15 morphological, 16,17 and genomic levels. 18 We immunophenotyped RS as cells coexpressing vimentin and smooth muscle ␣-actin, with loss of terminal muscle differentiation markers such as desmin and/or calponin. Other studies described RS as a mixture of fibroblasts, myofibroblasts, endothelial cells, and immune cells in breast cancer. 19 -21 We recently demonstrated that RS is promising as a new marker to predict a patient's time to progression in PCa, using both biopsies and tissue microarrays from radical prostatectomy specimens. [12][13][14] To determine whether RS carries predictive information, we charac-