Ability of linear length of positive margin in radical prostatectomy specimens to predict biochemical recurrence.