PURPOSE We developed an algorithm that prospectively defines when to excise the neurovascular bundles during radical retropubic prostatectomy with the goal of maximizing the performance of nerve sparing procedures while minimizing positive surgical margins. MATERIALS AND METHODS From January 1 to December 31, 2000 a single surgeon performed 272 radical retropubic prostatectomies and 263 were performed from January 1 to December 31, 2001. A single pathologist analyzed all specimens with positive margins. There were no prospectively defined criteria to guide decisions regarding excision of the neurovascular bundles in the 2000 study cohort. Gleason score, percent tumor volume and perineural invasion were independently analyzed in the biopsy specimens according to the site of origin (right versus left side) for the 2001 group only. The ipsilateral neurovascular bundle was excised for Gleason 6 or less tumors when there were 50% or greater tumor volume in the biopsy specimen and perineural invasion, for Gleason 7 tumors when there was 30% or greater tumor volume, or perineural invasion and for Gleason 8 to 10 tumors when there was 10% or greater tumor volume, or perineural invasion. RESULTS There were no statistically significant differences between the 2000 and 2001 groups in regard to preoperative prostate specific antigen, clinical and pathological stage, biopsy Gleason score and percent tumor volume in the surgical specimen. There was a statistically significant decrease in the incidence of positive margins between the 2000 and 2001 groups (14% versus 8%, p = 0.027). The lower positive margin rate was not achieved because of a tendency to excise more neurovascular bundles since a significantly greater percent of neurovascular bundles was preserved in the 2001 group. The sensitivity, specificity, positive and negative predictive values, and accuracy of our algorithm were 18%, 93%, 28%, 89% and 84%, respectively. In sides of the prostate with extraprostatic extension ipsilateral wide excision of the neurovascular bundle was associated with positive margins in 33% of cases compared with 22% when the neurovascular bundle was preserved (p = 0.42). CONCLUSIONS The New York University nerve sparing algorithm prospectively defines when to excise the neurovascular bundle based on Gleason score, perineural invasion and tumor volume in the biopsy specimen. Use of this algorithm decreases positive surgical margin rates, while significantly increasing the preservation of neurovascular bundles.