Prostate cancer staging is frequently inaccurate. By combining transrectal ultrasonography (TRUS) with a retrospectively derived grade-stratified prostate-specific antigen (PSA-GS) scale, we demonstrated 77 percent staging accuracy in 155 men with clinically localized prostate cancer undergoing radical prostatectomy. When used as the first step in a staging algorithm, PSA-GS (Score > or = 7: PSA > or = 4.0 ng/mL, uncontained; Score = 5 or 6: PSA > or = 8.0 ng/mL, uncontained; Score < or = 4: PSA > or = 16.0 ng/mL, uncontained) had a sensitivity of 75 percent and a specificity of 72 percent. The addition of TRUS to the staging algorithm, necessary only in patients with negative predictions by PSA-GS (46%), increased the sensitivity to 86 percent and the negative predictive value to 79 percent, while positive predictive value was unchanged at 77 percent. The combination of TRUS with PSA < 4 ng/mL or > or = 16 ng/mL identified subsets of patients with 85 percent and 88 percent likelihood of contained and uncontained disease, respectively. Our algorithm minimizes operator dependency by requiring TRUS in less than half of the patients. It produced improved staging, but the overall results were inaccurate in 23 percent of patients. Further refinements in prostate cancer staging are still necessary.