Since the widespread use of prostate-specific antigen (PSA) for the detection of prostate cancer in Japan, the diagnosis of clinically localized prostate cancer is not as rare as before.1 This phenomenon led Japanese urologists to commonly use radical prostatectomy to treat otherwise healthy patients with localized prostate cancer.2 To date, radical prostatectomy is considered the preferred treatment for patients with localized disease and at least a 10-year life expectancy. Japanese men have the longest life expectancy in the world;3,4 however, the therapeutic benefit for radical prostatectomy remains uncertain in Japanese men. In order to understand the natural history after radical prostatectomy in Japanese men, we compared long-term survival following radical prostatectomy with the life expectancy of the general Japanese male population of the same age during the same period to speculate on the impact of radical prostatectomy in Japanese patients. We also evaluated the most important factors that influenced survival rate, in order to confirm risk factors of progression after radical prostatectomy.
Patients and methods
One hundred and twenty-three consecutive patients with previously untreated and histologically confirmed clinical T1b, T1c, T2a, and T2b tumors were included in this retrospective analysis (Table 1). These 123 patients underwent radical prostatectomy from February 1982 through February 2000 without receiving androgen deprivation before surgery.
|n||Pathological stage||P-value||Final tumor grade||P-value|
|Initial PSA (ng/mL)|
|Final tumor grade|
Mean follow up was 51 months (range 3–186 months), 13 patients were followed for 10 years or more. The ages (age at surgery) of the patients ranged from 55 to 85 years with both the mean and median being 68 years. All 123 patients underwent radical retropubic prostatectomy (RRP) preceded by a bilateral pelvic lymph node dissection. One of the authors (Kiyoki Okada) performed RRP as either an operator or a supervising assistant on all patients.
Clinical staging before radical retropubic prostatectomy was determined in accordance with the unified tumor node metastasis (TNM),5 by digital rectal examination (DRE), transrectal ultrasonography (TRUS), endorectal magnetic resonance imaging (MRI), computer tomography (CT) and bone scan. Preoperative pathological diagnosis and tumor grading were made by transrectal biopsy before the initiation of any treatment in all patients. Finger-guided prostate biopsy was used until 1992. Subsequently, ultrasound-guided systematic biopsies were performed with a Brüel & Kjaer (Gentofte, Denmark) model 1846 scanner or a Mochida (Tokyo, Japan) Model MeU-1581 Sonovista EX. Biopsies were graded as well differentiated, moderately or poorly differentiated. Where a Gleason score was assigned, a score of 2–4 was grouped with the well-differentiated tumors, 5–7 was grouped with the moderately differentiated tumors and 8–10 with the poorly differentiated tumors. Whole mount paraffin sections were routinely used for pathological analysis. Each specimen was cut into slices 3–5 mm in thickness. Tumors contained within the prostatic capsule without penetration of the capsule were considered pT2N0. Specimens with prostatic carcinoma through the capsule and into the periprostatic fat, with or without seminal vesicle involvement, were designated as pT3N0 for statistical purposes. Positive regional lymph node metastasis with any pT was defined as pN1. Patients were seen every month during the first year postoperatively and every 3 months thereafter. Before the introduction and widespread use of serum PSA determination, patients were evaluated by periodic DRE, measurement of prostatic acid phosphatase and/or bone scan. Subsequently, determination of serum PSA levels, DRE and TRUS were the methods used in all patients. Serum PSA (Tosoh AIA-PACK PA) levels were determined every 3 months postoperatively.
Adjuvant hormonal therapy, consisting of surgical castration or an LHRH agonist was started within 1 month after surgery in 24 patients with pT3N0 and 15 patients with pN1. Postoperative external beam adjuvant therapy was not performed. Thirty-nine of these 123 received immediate adjuvant hormonal therapy, which probably rendered PSA follow-up data unreliable. Thirty-five patients, before the PSA era, did not have complete data available for statistical analysis of PSA failure-free survival.
Thus, the analysis of PSA failure-free survival as a surrogate end-point was carried out in the 64 patients who did not receive adjuvant therapy and received complete PSA follow up since initial diagnosis. PSA failure was defined as two consecutive detectable PSA levels (≥0.4 ng/mL). The patients in whom PSA failure was confirmed received salvage hormone therapy consisting of surgical castration or an LHRH agonist.
Clinical progression was seen as local tumor recurrence or distant metastasis by DRE, TRUS biopsy, or bone scan. A bone scan was performed every 6–12 months after PSA failure. We analyzed long-term clinical progression-free survival, prostate cancer-specific survival, and overall survival using the Kaplan–Meier method. In addition, we also analyzed PSA failure-free survival of patients that could be evaluated. Comparisons of survival curves were performed using the log-rank test. Overall survival was compared with expected survival of age-matched Japanese men using Japanese life tables.6
Univariate and multivariate analyses were performed using Cox proportional hazard analysis for the assessment of the prognostic significance of age, clinical stage, initial PSA, nerve-sparing status, pathological stage, tumor grade and adjuvant therapy, as well as for calculating the relative risk. The covariates, age (≥70 vs <70), nerve-sparing status (done vs not done), adjuvant therapy (done vs not done) were used as categorical variables. The covariate, initial PSA level, was used as the continuous numeric variable. The covariates, clinical stage (first, T1b and T1c; second, T2a; third, T2b), pathological stage (first, pT2N0; second, pT3N0; third, pN1) and tumor grade (first, well; second, moderately; third, poorly) were used as the ordinal numeric variables. A commercially available statistical package (SPSS Version 11) was used to conduct statistical analyses. Values of P < 0.05 were considered statistically significant.