Materials and Methods: Preoperative records and biopsy findings from 404 patients who underwent radical prostatoseminalvesi- culectomy between 9/91 and 9/93 were reviewed. Patients with prior RT or androgen suppression were excluded. Whole-mount radical prostatectomy and seminal vesicle (SV) specimens were examined, and SVI was measured and characterized in detail in 37 patients with SVI. The following characteristics were determined in each SV in which there was SVI: (1) maximum distance of SVI from the prostate edge, (2) seminal vesicle cancer volume (SVCaV) and SV percent cancer involvement, (3) seminal vesicle dimensions, (4) percent distance of SVI, (5) contiguous vs. noncontiguous extension of SVI from the prostate, (6) maximum length of the largest focus of SVI, (7) presence of extension of SVI outside of the SV, (8) presence of cancer at the SV and soft tissue resection margin, and (9) ratio of SVCaV to intraprostatic cancer volume. Kaplan-Meier survival analysis with a logrank test was used to identify significant predictors of biochemical, local, and/or systemic progression. Results: The clinical characteristics of the cohort include: age (range 47-75; mean 65 yrs), pathologic Gleason score (7;76%, 8;16%, 9;8%), preoperative serum PSA (mean; 23, median; 10.5, range 2-122), and 1997 AJCC clinical stage (T1c; 3%, T2a; 22%, T2b; 59%, T3a; 11%, T3b; 5%). Fifty-one percent of SVI specimens demonstrated contiguous SVI from the prostate, 35% had bilateral SVI and 57% showed cancer in adipose tissue adjacent to the SV. Twenty-seven percent of SV specimens with SVI demonstrated SVI at the margin. The maximum distance of SVI from the prostate edge ranged from 0.25 cm to 4.7 cm with a mean of 1.3 cm. The mean difference between the SVI distance and SV length was 0.9 cm with a range of 0 to 2.8 cm. Forty percent of patients demonstrated SVI within 0.5 cm of the distal end of the SV. The mean SVCaV was 0.85 cc and the maximum was 7.2 cc. The mean percent length of SV having SVI was 58% with a range from 9.6% to 100%. The maximum ratio of SVCaV to intraprostatic cancer volume was 27% and the mean was 3.7%. Logrank testing identified seminal vesicle length less than the median length (p=0.042) and the presence of perineural invasion (p = 0.037) to be associated with biochemical, local, and/or systemic progression. Conclusion: These data suggest that the entire SV should be considered as the target volume in RT management of patients considered at risk for SVI. Effective treatment of patients at risk for SVI by prostate brachytherapy alone is not likely to cover all cancer when seeds are placed solely in intraprostatic locations or within the bases of the SVs. Attention to margin status at SVI locations may provide information useful for consideration of adjuvant RT. 251 Correlation Between Dose-Volume Constraints and Late Rectum Bleeding in Patients Treated for Prostate Cancer at Dose Between 70 and 76 Gy C. Fiorino 1 , C. Cozzarini 2 , G. Sanguineti 3 , V. Vavassori 4 , C. Bianchi 5 , F. Foppiano 6 , A. Magli 4 , A. Piazzolla 5 , G.M. Cattaneo 1 1 Medical Physics, IRCCS H. S. Raffaele, Milano, Italy, 2 Radiotherapy, IRCCS H. S. Raffaele, Milano, Italy, 3 Radiotherapy, National Cancer Institute, Genova, Italy, 4 Radiotherapy, Varese Hospital, Varese, Italy, 5 Medical Physics, Varese Hospital, Varese, Italy, 6 Medical Physics, National Cancer Institute, Genova, Italy Purpose: To search for possible correlation between late rectal bleeding and dose-volume information taken from 3D treatment planning. Materials and Methods: Data from three institutions were included in current analysis; 3D treatment planning (including DVH) and clinical data of 364 patients treated for prostate cancer were collected. Patients with large air/fecal rectum content in the planning CT were excluded from the analysis(cut-off equal to 100 cc, n=61). 204/303 patients received an ICRU dose larger than or equal to 70 Gy and current analysis is referred to these patients (median follow up: 2.5 years; range: 1-7 years, median ICRU dose: 72 Gy). Rectum was contoured from the anal verge up to the sigmoid flessure by one observer for each institute. Inter-observer variations in rectum contouring was previously investigated and found to be quite modest. 179/204 3D treatment plans were recovered on the Cadplan 3D TPS; the remaining 25/204 were recovered on the Plato 3D TPS. Dose statistics-DVH calculation consistency between the two systems was investigated and found to be acceptable.We considered as bleeders the patients who experienced grade 2-3 late bleeding (RTOG/EORTC).Actuarial late bleeding incidences were calculated by the Kaplan-Meier method. Median and quartile values of all parameters were considered as cut-off values and log-rank tests were performed. Results: The bleeders were 15/204 (2.5 years actuarial incidence: 7.8 %). In Table I the results concerning the most predictive cut-off values for each parameter are shown (3rd quartiles for all parameters except the median value for “volume”). A fraction of rectal volume receiving at least 50 Gy (V50) greater than 70 % was found to be highly predictive of late rectum bleeding(p 0.001). Cut-off values for V50, V55, V60, V65,volume, mean, maximum and median dose were also found to be predictive(p  0.05). No other correlation was found with clinical variables (diabetes, previous surgery and age). Conclusion: The results indicate that dose-volume constraints taken from the DVH of the rectum (including filling) are predictive for late bleeding for patients without large air/fecal content during the planning CT scan. PARAMETER CUT-OFF 2.5 y % ACTUARIAL INCIDENCE ( CUT-OFF) 2.5 y % ACTUARIAL INCIDENCE ( CUT-OFF) p (LOG-RANK) Mean dose 57 Gy 5.8 13.0 0.044 Max. dose 74 Gy 5.0 18.0 0.017 Median dose 64 Gy 5.0 18.0 0.028 V50 70 % 4.8 17.4  0.001 V55 64 % 5.3 14.5 0.010 V60 55 % 4.9 14.3 0.007 V65 45 % 4.3 13.2 0.029 V70 35% 6.7 14.5 0.115 volume 55 cc 13.2 2.3 0.012 141 Proceedings of the 43rd Annual ASTRO Meeting