Original Articles Clinical Investigation
Japan Cancer of the Prostate Risk Assessment for combined androgen blockade including bicalutamide: Clinical application and validation
Correspondence: Yasuhide Kitagawa M.D., Ph.D., Department of Integrative Cancer Therapy and Urology, Graduate School of Medical Science, Kanazawa University, 13-1 Takaramachi, Kanazawa, Ishikawa 920-8640, Japan. Email: firstname.lastname@example.org
The Japan Cancer of the Prostate Risk Assessment was developed as a risk stratification instrument for patients undergoing primary androgen deprivation therapy. However, there have been no studies to validate the accuracy of the Japan Cancer of the Prostate Risk Assessment in predicting clinical outcomes. We examined whether the clinical outcomes of patients treated with combined androgen blockade could be stratified using the Japan Cancer of the Prostate Risk Assessment.
A total of 319 patients with prostate cancer treated with luteinizing hormone-releasing hormone agonist plus bicalutamide were included in this analysis. Progression-free survival, cause-specific survival and overall survival were compared among patients divided according to the prostate-specific antigen level at diagnosis, Gleason score on biopsy specimens, tumor–nodes–metastasis classification and Japan Cancer of the Prostate Risk Assessment score.
The median age of the patients was 75 years, and the median prostate-specific antigen at diagnosis was 25.4 ng/mL. A total of 102 patients (32.0%) had lymph node and/or distant metastases. On univariate analysis, the factors adopted in the Japan Cancer of the Prostate Risk Assessment points were significant predictors of progression-free survival. On multivariate analysis, clinical T stage and M stage were significant predictors of progression-free survival. The probabilities of progression-free survival and cause-specific survival were significantly different among the groups categorized according to the Japan Cancer of the Prostate Risk Assessment risk strata. The probability of overall survival in the low-risk group was higher than in the other groups.
The progression-free survival, cause-specific survival and overall survival of prostate cancer patients treated by combined androgen blockade with bicalutamide were stratified by the Japan Cancer of the Prostate Risk Assessment. The Japan Cancer of the Prostate Risk Assessment score is clinically useful as a predictor of the prognosis of prostate cancer treated with combined androgen blockade.
Abbreviations & Acronyms
combined androgen blockade
Japan Study Group of Prostate Cancer
Japan Cancer of the Prostate Risk Assessment
Cancer of the Prostate Strategic Urologic Research Endeavor
International Society of Urological Pathology
primary androgen deprivation therapy
randomized controlled trials
Prostate cancer is the most common form of cancer among men in Western countries, and the mortality rate as a result of this disease was high in the 1990s. However, the mortality rate has continuously decreased with the increased use of PSA-based screening in several countries. In the USA, downward stage migration over time has been well documented; among contemporary patients, just 5% are found to have nodal or distant metastatic disease at the time of diagnosis. In Japan, however, 21% of patients have distant metastases, and other measures of risk likewise tend to be higher than in the USA and western Europe. Because of this higher incidence of advanced disease, PADT is a mainstay of treatment for prostate cancer in Japan.
The Japanese Urological Association authorized the establishment of the J-CaP registry in 2001 as a large, multicenter, population-based database of patients undergoing PADT to show outcomes of PADT, which is frequently used internationally, but has been subjected to relatively little analysis compared with other major modalities. Approximately 20 000 cases were collected in the J-CaP database until 2003; based on these data, Cooperberg et al. established a risk stratification instrument developed for patients receiving PADT using the J-CAPRA database score.
The accuracy of the J-CAPRA score was externally validated for predicting CSS among patients in the USA community-based Cancer of the CaPSURE registry; however, there have been no studies to validate the accuracy of the J-CAPRA score in predicting PFS along with analysis of clinical situation. In the present study, we showed the accuracy of the J-CAPRA score in predicting PFS, CSS and OS among patients treated with CAB including bicalutamide[9-11] (Casodex; Astra-Zeneca Pharmaceuticals, Wilmington, DE, USA). This is the first study to validate the J-CAPRA score and risk strata for the probabilities of PFS, CSS and OS. Absolute outcomes are presented as a reference for clinicians with patients receiving CAB.
Patients and procedure
After receiving institutional review board approval, we retrospectively analyzed the charts of patients diagnosed with prostate cancer pathologically between 2001 and 2008 in three departments of urology (Kanazawa University, Ishikawa Prefectural Central Hospital and Kanazawa Medical Center). We selected 319 patients treated with CAB including bicalutamide as primary treatment continually until disease progression or until censored at the final clinic follow up at which the status of the patient was known. The patients received local treatments with CAB, and those receiving other treatments before disease progression were excluded. Clinical staging was determined in accordance with the unified TNM criteria based on the results of digital rectal examination, transrectal ultrasound, computed tomography, magnetic resonance imaging and bone scan. All patients underwent transrectal biopsy before initiation of treatment and were diagnosed pathologically as prostate adenocarcinoma. In the present study, pathological findings were evaluated based on the 2005 ISUP Gleason grading by pathologists in each institution. All patients were treated with bicalutamide[9-11] (Casodex, 80 mg orally once a day) plus luteinizing hormone-releasing hormone agonist – either goserelin acetate (Zoladex; Astra-Zeneca Pharmaceuticals) or leuprorelin acetate (Luprin; Takeda Chemical Industries, Osaka, Japan). Goserelin acetate was given at a dose of 3.6 mg or 10.8 mg by subcutaneous injection every 4 or 12 weeks, respectively. Leuprorelin acetate was given at 3.75 mg or 11.25 mg by subcutaneous injection every 4 or 12 weeks, respectively. The patients were followed up every 4 or 12 weeks, and the serum PSA level was assessed. Disease progression was determined as increasing PSA, and the patients received other drugs and/or treatments after PSA failure.
The survival time was calculated from the date of starting CAB, and the status of each patient was assessed by his visits to the outpatient clinic.
PFS was analyzed by univariate and multivariate Cox proportional hazards regression models based on the pretreatment factors adopted in the J-CAPRA score. Predictors included patient's age at diagnosis, Gleason score of biopsy specimens, serum PSA level at diagnosis and 1997 TNM clinical stage. In terms of the sums of J-CAPRA score and risk stratification as defined and validated previously; that is, J-CAPRA low risk (0–2 points), intermediate risk (3–7 points) and high risk (8–12 points), PFS, CSS and OS were analyzed by univariate Cox proportional hazards regression models and Kaplan–Meier analysis. The significance of differences between scores and risk strata in J-CAPRA was analyzed by the log–rank test. In analyses of probabilities of survival, patients without evidence of progression and/or death were censored at the time of last follow up. All statistical analyses were carried out using commercially available software; that is, spss Statistics (IBM, Armonk, NY, USA) and Prism (GraphPad Software, San Diego, CA, USA). In all analyses, P < 0.05 was taken to show statistical significance. The validation cohort of J-CAPRA was assessed by the c-index. Its interpretation is similar to that of the area under the receiver operating curve for a diagnostic test: a c-index of 0.5 indicates no improvement over random guess, whereas a c-index of 1.0 indicates perfect predictive accuracy. The c-indexes were calculated using stata 10 (StataCorp LP, College Station, TX, USA).
The clinical and pathological characteristics of the patients are shown in Table 1. The median age of the patients was 75 years (range 46–91 years) and the median PSA was 25.4 ng/mL (range 0.57–13 435 ng/mL) at diagnosis. The Gleason scores of biopsy specimens from 152 patients (47.6%) were ≥8. Of all 319 patients, 102 (32.0%) had lymph node and/or distant metastases. In comparison with the patient characteristics in the J-CAPRA derivation paper by Cooperberg et al., the proportion of patients with Gleason scores ≤6 was low, and that of patients with clinical stage T1 was high in the present study population. Other characteristics including age and serum PSA level in the present study were similar to those in the J-CAPRA derivation paper.
Table 1. Clinical characteristics of the study population
|Age (years)|| || || |
|≤75||161 (50.5)|| ||9934 (51.6)|
|>75||158 (49.5)|| ||9332 (48.4)|
|Gleason score|| || || |
|6||75 (23.5)||0||5884 (35.1)|
|7||92 (28.8)||1||4821 (28.7)|
|9||77 (19.7)||2||6060 (36.2)|
|10||4 (1.3)||2|| |
|PSA at diagnosis (ng/mL)|| || || |
|0–10||85 (26.7)||0||4727 (24.6)|
|>10–20||63 (19.7)||0||3713 (19.3)|
|>20–100||81 (25.4)||1||5865 (30.5)|
|>100–500||53 (16.6)||2||2929 (15.3)|
|>500||37 (11.6)||3||1972 (10.3)|
|cT stage|| || || |
|T1c||117 (36.7)||0||T1 4001 (20.8)|
|T2a||42 (13.2)||0||T2 6274 (32.6)|
|T2b||48 (15.0)||1|| |
|T3a||47 (14.7)||1||T3 7048 (36.6)|
|T3b||32 (10.0)||2|| |
|T4||33 (10.2)||3||T4 1943 (10.1)|
|cN stage|| || || |
|N0||263 (82.4)||0|| |
|N1||56 (17.6)||1|| |
|cM stage|| || || |
|M0||233 (73.0)||0|| |
|M1||86 (27.0)||3|| |
Table 2 shows the results of univariate and multivariate Cox proportional hazards regression analysis for time to PFS based on the pretreatment factors. The mean follow-up period of all patients was 3.67 years, and disease progression occurred in 79 cases. In all of these patients, disease progression was determined with PSA failure, and there were no patients with progression of local symptoms and metastases before PSA failure. For statistical analyses, the Gleason scores of biopsy specimens, serum PSA levels at diagnosis and 1997 TNM clinical stage were classified according to categories of J-CAPRA variables. Patients were divided into two groups according to age (≤75 years and >75 years). On univariate analysis, the factors adopted in J-CAPRA scores were significant predictors of PFS, and the hazard ratio increased according to each J-CAPRA score. On multivariate analysis, clinical T stage and M stage were significant predictors of PFS.
Table 2. Univariate and multivariate Cox proportional hazards regression analysis for time to PFS
|Age (years)|| || || || || |
|≤75||161||1 (reference)|| ||1 (reference)|| |
|>75||158||0.71 (0.45–1.11)||0.129||0.69 (0.43–1.10)||0.116|
|Gleason score|| || || || || |
|6 (0)||75||1 (reference)|| ||1 (reference)|| |
|7 (1)||92||5.84 (1.31–26.14)||0.021||1.43 (0.28–7.42)||0.671|
|8–10 (2)||152||22.78 (5.56–93.38)||<0.001||2.55 (0.51–12.82)||0.255|
|PSA (ng/mL)|| || || || || |
|0–20 (0)||148||1 (reference)|| ||1 (reference)|| |
|>20–100 (1)||81||7.83 (3.15–19.42)||<0.001||1.94 (0.68–5.52)||0.214|
|>100–500 (2)||53||18.91 (7.80–45.84)||<0.001||2.25 (0.76–6.61)||0.142|
|>500 (3)||37||26.81 (10.89–66.03)||<0.001||1.74 (0.54–5.62)||0.353|
|cT stage|| || || || || |
|T1c, T2a (0)||159||1 (reference)|| ||1 (reference)|| |
|T2b, T3a (1)||95||14.26 (5.55–36.67)||<0.001||4.89 (1.63–17.73)||0.005|
|T3b (2)||32||26.35 (9.67–71.82)||<0.001||6.33 (1.89–21.21)||0.003|
|T4 (3)||33||41.58 (15.76–109.69)||<0.001||9.09 (2.79–29.57)||<0.001|
|cN stage|| || || || || |
|N0 (0)||263||1 (reference)|| ||1 (reference)|| |
|N1 (1)||56||6.74 (4.29–10.58)||<0.001||1.54 (0.85–2.78)||0.152|
|cM stage|| || || || || |
|M0 (0)||233||1 (reference)|| ||1 (reference)|| |
|M1 (3)||86||7.07 (4.44–11.26)||<0.001||2.02 (1.19–3.44)||0.009|
Table 3 shows the results of univariate Cox proportional hazards regression analysis for time to PFS based on the J-CAPRA score. The proportions of patients with each J-CAPRA score were similar to those in the J-CAPRA derivation paper. A J-CAPRA score ≥4 was a significant predictor of PFS compared with the reference, and the hazard ratio roughly increased according to the increase in J-CAPRA score. The categorized scores for risk stratification were significant predictors of PFS. Kaplan–Meier estimates of the probabilities of PFS at 4 years, which was set to compare the probabilities in the J-CAPRA derivation paper, were 100% and 16.7% for those with J-CAPRA scores of 0 and 12, respectively. The c-indexes for this validation cohort in the continuous and categorized scores were 0.855 and 0.833, respectively.
Table 3. Univariate Cox proportional hazards regression analysis for time to PFS by the J-CAPRA score and risk strata
|J-CAPRA|| || || || |
|0||61 (19.1)||1 (reference)|| ||100|
|1||50 (15.7)||1.35 (0.09–21.64)||0.831||100|
|2||40 (12.5)||6.10 (0.63–58.81)||0.118||89.1|
|3||25 (7.8)||2.30 (0.14–36.86)||0.557||100|
|4||27 (8.5)||22.68 (2.78–185.09)||0.004||74.4|
|5||15 (4.7)||27.27 (3.04–245.01)||0.003||60.2|
|6||15 (4.7)||37.24 (4.57–303.54)||0.001||58.3|
|7||19 (6.0)||39.55 (4.99–313.37)||<0.001||43.0|
|8||11 (3.4)||36.83 (4.14–327.85)||0.001||54.5|
|9||20 (6.3)||99.16 (12.92–761.42)||<0.001||29.2|
|10||15 (4.7)||105.94 (13.69–819.73)||<0.001||19.4|
|11||15 (4.7)||76.28 (9.80–593.60)||<0.001||24.0|
|12||6 (1.9)||150.91 (17.27–1318.76)||<0.001||16.7|
|Risk strata|| || || || |
|0–2 (low)||151 (47.3)||1 (reference)|| ||97.3|
|3–7 (intermediate)||101 (31.7)||9.59 (3.70–24.87)||<0.001||70.5|
|8–12 (high)||67 (21.0)||36.73 (14.52–92.88)||<0.001||28.4|
Figure 1 shows Kaplan–Meier estimates of survival among the J-CAPRA scores and risk stratification. Figure 1a,b show PFS among the J-CAPRA scores and risk stratification, respectively. The probabilities of PFS with 95% confidence interval at 2 years in the low-, intermediate- and high-risk groups were 100% (100%), 90.4% (84.4–96.4%) and 38.0% (26.1–50.0%), respectively. Those at 4 years were 97.3% (94.3–100%), 70.5% (59.5–81.5%) and 28.4% (16.7–40.0%).
Figure 1c shows CSS and Figure 1d shows OS among the J-CAPRA risk stratification. A total of 23 patients died of prostate cancer and 19 died as a result of other causes during the follow-up period. Each survival probability decreased according to the J-CAPRA risk stratification. The probabilities of PFS and CSS showed significant differences among the groups categorized by risk strata on log–rank test. The probability of OS in the low-risk group was significantly higher than those in the other groups. The probabilities of CSS with 95% confidence interval at 4 years in the low-, intermediate- and high-risk groups were 100% (100%), 94.2% (89.3–99.2%), and 71.8% (58.5–85.0%), respectively. The probabilities of OS at 4 years were 93.8% (89.5–98.0%), 87.0% (80.0–93.9%) and 69.6% (56.1–83.1%), respectively. The c-indexes of CSS and OS in the categorized scores were 0.833 and 0.665, respectively.
Since Huggins and Hodges first reported the use of hormonal therapy for prostate cancer, PADT including CAB has been the mainstay for the management of advanced prostate cancer. In terms of clinically localized prostate cancer, Labrie et al. reported that approximately 80% of stage B prostate cancer could be controlled for long periods with PADT, and recent analyses of practice patterns showed that many patients with localized cancer treated conservatively receive PADT rather than local treatment or active surveillance/watchful waiting.[17, 18] Nevertheless, there have been no reports of instruments to predict outcomes for any population of previously untreated patients initiating PADT until the publication of the J-CAPRA derivation paper. J-CAPRA was developed by analysis of a large database of a variety of patients undergoing PADT in Japan, and should be useful for clinical application to predict PADT in patients with not only advanced disease, but also localized cancer.
Several predictors of PFS of prostate cancer patients treated with ADT; for example, advanced age and high Gleason score for localized and locally advanced prostate cancer and PSA half-time and doubling time for metastatic disease, were reported previously. In the present study, we showed that the pretreatment clinical characteristics included in J-CAPRA were good predictors of PFS on univariate analysis, and clinical T and M stage on multivariate analysis. As the factors are necessary for the diagnosis of prostate cancer and are not specific, J-CAPRA scores might be reasonable for clinical application.
In this validation cohort, we showed that the J-CAPRA score and risk stratification can be used as predictors of PFS for patients treated by CAB including bicalutamide using univariate Cox proportional hazards regression analysis. The accuracy of the J-CAPRA score for prediction of PFS was shown using the c-indexes, and the points of the c-indexes were similar or superior to those in the J-CAPRA derivation paper. In terms of CSS and OS, the validation for CSS of PADT patients in CaPSURE was shown, whereas that of OS was not reported in the J-CAPRA derivation paper. We showed that risk stratification of the J-CAPRA score predicted CSS and OS of the patients well. The results obtained in this validation cohort verified that J-CAPRA score is a clinically powerful prediction tool for patients treated with CAB. In the clinical setting, physicians may predict PFS, CSS and OS of patients treated with CAB including bicalutamide as shown by the present results.
The benefits of CAB compared with non-CAB hormonal therapy have been controversial. In the 2000s, it was reported that the survival advantage of maximum androgen blockade compared with surgical or medical castration alone in published RCT and meta-analyses was negligible. In contrast, recent evidence from RCT suggested greater benefits of CAB in certain subgroups of patients,[9, 10] and the 2006 update of the guidelines of the American Society for Clinical Oncology recommended that CAB should be considered as a therapeutic option for initial hormonal management of androgen-sensitive, metastatic, recurrent or progressive disease. Bias of patient characteristics might be an important factor for comparison of the efficacy of each hormonal treatment. In the J-CaP database, further analysis showed a better prognosis of patients treated with CAB than with non-CAB therapy in the groups with both intermediate- and high-risk J-CAPRA stratifications. In the present study, the probabilities of PFS at 4 years (100–16.7%) were better than those in the J-CAPRA derivation paper (74.9–8.2%) in the groups with each J-CAPRA score. Although we could not directly compare the present results with those of the J-CAPRA derivation paper, further examinations using J-CAPRA score are expected to reveal the efficacy of each hormonal treatment and the J-CAPRA score will be a useful tool to adjust patient characteristics in further clinical trials for PADT.
The present study had several limitations. The study design was retrospective, and the sample size was relatively small for multivariate analysis. In addition, imperfect discrimination was found between adjacent J-CAPRA score levels. In terms of pathological diagnosis, there might have been a bias for Gleason scoring because of the lack of a central pathologist. However, as J-CAPRA was derived from a multicenter, population-based database, J-CAPRA can be applied clinically, including possible biases. We showed that the J-CAPRA score is a useful tool for predicting the clinical outcomes of patients treated by CAB including bicalutamide.
We thank Dr Matthew R Cooperberg of University of California, San Francisco, USA, for valuable advice. We also thank Dr Kiyoshi Koshida and Dr Yuta Takezawa of Kanazawa Medical Center, and Dr Masayoshi Shimamura, Dr Takao Nakashima and Dr Kazuaki Machioka of Ishikawa Prefectural Central Hospital for providing clinical data of prostate cancer patients.
Conflict of interest