|Home | About | Journals | Submit | Contact Us | Français|
Prostate cancer epidemiology has been marked overall by a downward risk migration over time. However, in some populations, both in the United States and abroad, many men are still diagnosed with high-risk and/or advanced disease. Primary androgen deprivation therapy (PADT) is frequently offered to these patients, and disease risk prediction is not well-established in this context. We compared risk features between large disease registries from the United States and Japan, and aimed to build and validate a risk prediction model applicable to PADT patients.
Data were analyzed from 13,740 men in the United States community-based Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) registry and 19,265 men in the Japan Study Group of Prostate Cancer (J-CaP) database, a national Japanese registry of men receiving androgen deprivation therapy. Risk distribution was compared between the two datasets using three well-described multivariable instruments. A novel instrument (Japan Cancer of the Prostate Risk Assessment [J-CAPRA]) was designed and validated to be specifically applicable to PADT patients, and more relevant to high-risk patients than existing instruments.
J-CaP patients are more likely than CaPSURE patients to be diagnosed with high-risk features; 43% of J-CaP versus 5% of CaPSURE patients had locally advanced or metastatic disease that could not be stratified with the standard risk assessment tools. J-CAPRA—scored 0 to 12 based on Gleason score, prostate-specific antigen level, and clinical stage—predicts progression-free survival among PADT patients in J-CaP with a c-index of 0.71, and cancer-specific survival among PADT patients in CaPSURE with a c-index of 0.84.
The novel J-CAPRA is the first risk instrument developed and validated for patients undergoing PADT. It is applicable to those with both localized and advanced disease, and performs well in diverse populations.
Prostate cancer incidence varies tremendously across the world, depending on ethnic and population genetic factors, dietary and other environmental influences, and local screening and diagnostic practices. Incidence is highest in the United States, where prostate-specific antigen (PSA)–based screening is most intense: the world age-standardized rate (ASR) for diagnosis in the US in 2002 was 113.7 per 100,000, corresponding to a 15.3% lifetime risk of diagnosis. However, the ASR even within the country varies seven-fold with region and ethnic group, from a low of 30.9 among Koreans in Los Angeles county to a peak of 216.0 among African Americans in the Detroit area. Rates in Asia are lower overall, ranging from 1.4 in the Jiashan region of China to 50.2 in Israel; by comparison among Asians in the United States, the ASR is 58.0. ASRs in Japan range from 11.3 to 22, for a lifetime risk of diagnosis of 1.2 to 2.5%.1
Disease stage and risk also follow disparate patterns across regions. In the US, downward stage migration over time has been well-documented2; among contemporary patients only 5% are found to have nodal or distant metastatic disease at time of diagnosis.3 In Japan, by contrast, 21% of patients present with distant metastases and 19% with locally advanced disease; other measures of risk likewise tend to be higher than elsewhere in the developed world, although as elsewhere there exists evidence of downward risk migration.4 With this higher incidence of advanced disease, primary androgen deprivation therapy monotherapy (PADT) is a mainstay of treatment for prostate cancer in Japan. PADT is commonly used in Japan for lower stage disease as well, accounting for 46% overall of primary treatment among men with clinically localized disease.4 Growing use of PADT for localized disease is not unique to Japan; large US studies have demonstrated frequent PADT use across risk groups as well.5,6 While medical androgen deprivation therapy in all contexts has fallen in the US in recent years since the implementation of changes to reimbursement under Medicare,7 the specific impact on PADT for localized disease has not yet been ascertained.
Given the large experience with PADT in Japan, the Japanese Urological Association authorized the establishment of the J-CaP registry in 2001 as a large, multicenter, population-based database of men undergoing PADT,8 in order to shed light on outcomes for a treatment modality which is frequently used internationally but has been subjected to relatively little analysis to date compared with other major modalities.9 Risk assessment in this population entails special challenges related to the high proportion of advanced disease and the absence of risk instruments developed or validated for patients receiving PADT. We developed a collaborative study between Japan Study Group of Prostate Cancer (J-CaP) and the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) database, a large, community-based registry of prostate cancer patients in the US,10 in order to compare patterns of risk and outcomes.
J-CaP includes men diagnosed in Japan with any stage of prostate cancer between 2001 and 2003 and treated with androgen deprivation therapy, as monotherapy or in combination. Three hundred eighty-four institutions contributed patients, comprising approximately 50% of all men diagnosed in Japan during the accrual period, and nearly 95% of those treated with PADT. Eighteen point five percent of the patients were treated in academic medical centers, and the remainder in the community. 1997 TNM stage was reported directly by participating clinicians; detailed biopsy data were not included. Urologists at participating institutions report follow-up every 3 months on an ongoing basis, including information on additional treatments, progression, and all-cause mortality. Clinical progression is determined by the clinicians, and may reflect rising PSA, progression of local symptoms, and/or development of metastases. Twenty-six thousand two hundred seventy-two men were enrolled in J-CaP; of these 19,265 were treated with PADT and were included for analysis in this study. Additional information regarding J-CaP has been published previously.8,11
CaPSURE accrued patients from 40 clinical practice sites across the US, with 11.4% of the patients treated at academic centers; 13,740 men were accessioned between 1995 and 2007, fewer than 1% of whom were of Asian ancestry. Patients are treated according to the participating physicians' usual practice; 2,077 (15.1%) were treated with PADT. Rather than reporting clinical stage or progression directly as in J-CaP, CaPSURE clinicians submit data on digital rectal exam and ultrasound findings, initial and subsequent PSA values and imaging results, and secondary treatments. Additional clinical data are acquired through semiannual patient surveys, and hospitalization data are verified by medical record audit. Additional details regarding CaPSURE have been published previously.10 Analyses were performed both on the CaPSURE cohort as a whole and, to allow better comparability with J-CaP, on the subset of those treated with PADT since 2000. Data in both CaPSURE and J-CaP are collected and managed under local and central institutional review board supervision.
For this study, patient risk was assessed with the three-level risk stratification system published by D'Amico et al12 and adopted in the American Urological Association's clinical practice guideline for localized prostate cancer.13 Because J-CaP reports clinical stage under the 1997 TNM system which does not include a T2c designation, the risk classification was modified for both cohorts to stratify patients to high risk based on clinical stage T3a. The original Kattan preoperative nomogram score was calculated for all patients with known PSA, Gleason score, and clinical stage.14 A Kattan-type nomogram has not been published for patients receiving PADT; the updated preoperative instrument15 was not used because it requires detailed biopsy data not included in J-CaP.
Finally, the University of California, San Francisco Cancer of the Prostate Risk Assessment (CAPRA) score was calculated for both cohorts.16 Because detailed biopsy data are not included in J-CaP, a previously published modification of the CAPRA score was used, which omits the percent of positive cores variable and calculates the score on a 0 to 9 rather than 0 to 10 scale.17 For patients missing exactly one input variable needed to calculate the score, the score was imputed using least squares regression; the distribution of imputed CAPRA scores was similar to the distribution of nonimputed scores. These instruments are all intended to stratify patients with localized disease; in this study those with “advanced” disease (clinical stage > T3aN0M0) were categorized separately as advanced. Univariate comparisons were made between categorized age and other ordinal risk factors in J-CaP versus CaPSURE using Mantel-Haenszel χ2 analysis.
J-CaP data were built into a new Cox proportional hazards regression model to predict likelihood of clinical progression as defined above. Parameter estimates for variables in the model which were statistically significant independent predictors of progression were translated to points, with each increase in parameter estimate (β) up to .25 corresponding to one point. The ability of the new instrument, designated the Japan Cancer of the Prostate Risk Assessment (J-CAPRA) score, to predict clinical progression-free survival (PFS) was assessed with Kaplan-Meier analysis, Cox proportional hazards regression, and calculation of Harrell's c-index. The c-index assesses the accuracy of a predictive instrument. Its interpretation is similar to that of the area under a receiver operating curve for a diagnostic test: a c-index of 0.5 would indicate no improvement over random guessing, whereas a c-index of 1.0 would indicate perfect predictive accuracy.18
The assumption of proportionality was tested via construction of log-minus-log and smooth Schoenfeld residual plots, both of which demonstrated essentially parallel curves; a LOWESS smooth drawn through the latter plot was horizontal. CIs for the Cox model were calculated with bias corrected and accelerated bootstrap correction. The J-CAPRA score was also categorized to identify three groups at low (J-CAPRA 0 to 2), intermediate (J-CAPRA 3 to 7), and high (J-CAPRA 8 to 12) risk of recurrence.
As a validation analysis, the accuracy of the J-CAPRA score was tested in predicting cancer-specific survival (CSS) among 1,718 patients in CaPSURE receiving PADT with sufficient data to calculate the score. Cox analysis was performed, treating the score both as a continuous and a categorized variable, and the c-index was again calculated for the validation cohort.
The mean ± standard deviation (SD) for age among all CaPSURE patients was 66.2 ± 8.7, among the subset treated with PADT since 2000 (n = 1,024) 72.3 ± 8.9, and among J-CaP patients 75.0 ± 7.6. Fifteen point nine percent of the CaPSURE PADT patients were African American, compared to 10.8% of all CaPSURE patients. Table 1 summarizes the clinical and risk characteristics for J-CaP, CaPSURE, and the CaPSURE PADT subset. All comparisons between J-CaP and either the full or subset CaPSURE cohorts were significant at P < .001. In general, patients in the CaPSURE subset undergoing PADT since 2000 were older and had higher risk features by any metric than the overall CaPSURE cohort, but younger and with lower risk features than the J-CaP patients.
The mean ± SD Kattan and CAPRA scores were, respectively, 77.2 ± 19.1 and 2.7 ± 1.8 among the overall CaPSURE cohort, 66.1 ± 24.8 and 3.9 ± 2.2 among the CaPSURE PADT subset, and 53.8 ± 31.7 and 4.6 ± 2.4 among the J-CaP cohort. However, these figures underestimate the risk differences among the cohorts, as 4.1% of the CaPSURE patients overall had advanced disease at diagnosis (> T3aN0M0) for which the D'Amico, Kattan, and CAPRA calculations do not apply, compared to 9.7% of the CaPSURE PADT subset and 42.8% of the J-CaP cohort.
Figure 1 illustrates patterns of risk at diagnosis according to patient age in the different cohorts. In CaPSURE, older patients tend to present with higher risk disease, with rates of high risk or advanced disease increasing from 10.5% among men younger than 55 years to 65.4% of those diagnosed who were older than 85 years. Conversely in J-CaP, 89.2% of men younger than 55 have high risk or advanced disease; with rising age risk falls until after the age of 85, at which point it rises again. The association between risk and age among CaPSURE PADT patients appears similar to that seen among J-CaP patients.
Among the J-CaP cohort, 5,683 (29.5%) progressed at a median of 16.8 months; median follow-up among those not progressing was 23.7 months. The statistical derivation and calculation of the J-CAPRA score are presented in Table 2. Patients may be assigned 1 point for Gleason score 7 or 2 points for Gleason score 8 to 10; 1 point for PSA 20 to 100 ng/mL, 2 points for PSA 100 to 500 ng/mL, or 3 points for PSA higher than 500 ng/mL; 1 point for stage T2b or T3a, 2 points for T3b, or 3 points for T4; 1 point for N1 disease; and 3 points for M1 disease. Points for each variable are summed to yield a total score with a range of 0 to 12. Due to small numbers of patients at extremely high risk, scores over 10 were combined to 11 to 12.
The distribution of J-CAPRA scores among the J-CaP cohort are listed in Table 3, along with results of the Cox analysis and actuarial 2- and 4-year PFS estimates. The HR (95% CI) for progession with each point increase in J-CAPRA score was 1.25 (1.24 to 1.26) corresponding to roughly a 50% increase in risk of progression with every two-point increase in score (1.252 = 1.56). The accuracy as measured by the c-index was 0.72. For the categorized J-CAPRA score, relative to low-risk patients (J-CAPRA 0 to 2), the HR for progression was 2.38 (2.21 to 2.56) for intermediate-risk patients (J-CAPRA 3 to 7), and 7.06 (6.55 to 7.61) for high-risk patients (J-CAPRA 8 to 12). Kaplan-Meier plots for both the continuous and categorized scores are presented in Figure 2.
Among 1,718 CaPSURE patients treated with PADT, 178 (10.1%) died of prostate cancer, at a median of 45.6 months. Median follow-up among those not dying of prostate cancer was 40.3 months. The J-CAPRA score distribution, along with results of the Cox analysis and actuarial 2- and 4-year CSS estimates are given in Table 4. The HR (95% CI) for cancer-specific mortality (CSM) with each point increase in J-CAPRA score was 1.46 (1.39 to 1.53), corresponding to roughly a doubling of risk of death from prostate cancer with every two-point increase in score. Relative to J-CAPRA 0 to 2 patients, J-CAPRA 3 to 7 patients had an HR for cancer mortality of 7.34 (4.92 to 10.93), and J-CAPRA 8 to 12 patients had an HR of 30.96 (18.49 to 51.83). Kaplan-Meier curves for CSS among CaPSURE PADT patients are presented in Appendix Figure A1 (online only). The c-index for this validation cohort was 0.84.
In the US, incidence and mortality of prostate cancer have been falling steadily since 2001 and 1993, respectively; 5-year CSS from diagnosis now approximates 100% for locoregional disease, and is 32% even for metastatic disease.3 In contrast, data from the Osaka Cancer Registry demonstrate a 50% increase in prostate cancer incidence in Japan during the 1990s, during which time 5-year survival rose sharply from 48% to 64%.19 Other investigators have likewise documented downward stage and grade migration in Japan as PSA screening has become more widespread.20 Nonetheless, men in Japan on average, as reflected in the J-CaP cohort, are diagnosed with markedly higher-risk, later-stage disease than those in most other developed countries.
As demonstrated in Figure 1, in CaPSURE, older patients tend to present with higher-risk disease, likely because they have been less intensively screened and therefore present later in the disease course.21 Conversely, in J-CaP, with rising age men present with consistently lower risk until after the age of 85, at which point risk increases again. Of note, the age-risk association among men in CaPSURE receiving PADT mirrors that seen among J-CaP patients—even the age at which the trend reverses—although CaPSURE patients undergoing PADT still have generally lower risk features than J-CaP patients (Tables 1, ,3,3, and and44).
Although PADT is commonly administered as monotherapy for localized prostate cancer in the US,5,6 outcomes data from US cohorts regarding this approach are sparse.9 Observational studies often include patients undergoing PADT with those undergoing conservative management,22 which may bias such studies either for or against conservative management. A recent instrumental variable analysis suggested that men with lower-risk tumors had better survival when followed expectantly than when treated with PADT. The study, however, was restricted in its evaluation of risk by the limitations of administrative databases; data on PSA and appropriate Gleason scoring, for example, were not available.23 A number of recent articles have raised awareness of the potential adverse metabolic, musculoskeletal, and cardiovascular effects of androgen deprivation24–26; given these concerns and the lack of prospective studies, the 2007 American Urological Association practice guideline for prostate cancer did not endorse PADT as a recommended option for clinically localized prostate cancer.13
A 1-year registration study of Japanese men diagnosed with prostate cancer was conducted in 2000 before the inauguration of J-CaP. Among 4,529 men, 70% had stage M0 at diagnosis, 81 had clinical T stage T3a or lower, and 30% had a PSA level lower than 10 ng/mL. More than 50% of men across clinical stages received PADT.4 Analysis of Japanese cohorts have previously produced unique findings regarding the efficacy of PADT27,28; J-CaP was initiated to yield better insights into PADT outcomes at the large-scale, population-based level.8,11
Risk stratification in the J-CaP population presents unusual challenges. Among 111 entries in a recently updated catalog of published risk instruments for prostate cancer, none had been validated among Japanese patients, nor among those presenting across such a broad range of stages; most instruments intended for use at time of diagnosis explicitly exclude those with locally advanced or metastatic disease.29 A Japanese validation study of the Partin tables has recently been published, together with a novel risk instrument for Japanese men undergoing prostatectomy.30 Other instruments have been published for advanced disease,31,32 but are intended for patients with castrate resistant disease. No previously published instrument has been intended or validated to predict outcomes for any population of previously untreated patients initiating PADT, nor for cohorts including patients with both localized and advanced disease.
The novel J-CAPRA score, a straightforward instrument requiring neither paper tables nor computer software to calculate, appears to perform well in predicting progression both among the J-CaP development cohort and the CaPSURE PADT external validation cohort, which is both demographically and clinically distinct from J-CaP. These findings highlight the importance of risk stratification even for patients who present with locally advanced or metastatic disease. Patients with multiple adverse risk features at diagnosis may be expected to progress rapidly on PADT, and might benefit from early enrollment in clinical trials of chemotherapeutics, novel targeted agents, or combination therapy.
Several limitations of this analysis should be noted. J-CaP includes only patients receiving androgen deprivation therapy; while these represent roughly half of all prostate cancer patients in Japan, they cannot be assumed to represent the whole Japanese population. Similarly, CaPSURE, while a large community-based US database, is not a random sample of the US prostate cancer population. Although there is moderate representation of African Americans in the CaPSURE PADT cohort, other ethnicities are underrepresented; therefore a comparison of outcomes among patients of Asian descent between the two cohorts is not possible at this time.
Progression in J-CaP is reported directly by clinicians, who may use varied definitions of progression as noted in the Methods section earlier; this heterogeneity could be a source of random error. This variation may explain the substantially higher c-index noted for the validation cohort in CaPSURE, which was tested with the more consistently defined CSM end point. CSM has not yet been ascertained in J-CaP, but as the database reaches further maturity, additional evaluation of J-CAPRA at mortality end points in the J-CaP cohort will be important. Finally, the present analysis does not allow any conclusions to be drawn regarding the merits of PADT relative to other therapeutic modalities.
J-CaP comprises a remarkable group of prostate cancer patients, many of whom presented with far higher risk and more advanced disease than those in other contemporary cohorts reported to date. Despite limited prospective studies reported to date, PADT is used commonly on both sides of the Pacific; with longer follow-up, the J-CaP registry will represent an invaluable source of outcomes data for patients treated with this modality. Care must be taken in applying risk instruments across patient populations and settings for which they have not been validated. The J-CAPRA score is a novel, validated score for predicting outcomes among patients undergoing PADT across the full spectrum of risk and stage, including advanced disease. We hope and expect that future international collaborative studies across disease registries will continue to yield insights into prostate cancer behavior and outcomes independent of local patterns of practice.
We acknowledge the contribution of the following additional Japan Study Group of Prostate Cancer (J-CaP) investigators: Susumu Kagawa, President, Tokushima University Hospital, Tokushima; Tadaichi Kitamura, President, Asoka Hospital, Tokyo; Yoshihiko Hirao, Professor and Chairman, Department of Urology, Nara Medical University, Kashihara-city, Nara; Masaru Murai, President, International Goodwill Hospital, Yokohama; Naoto Miyanaga, Assistant Professor, Department of Urology, Institute of Clinical Medicine, University of Tsukuba, Tsukuba; Seiji Naito, Professor and Chairman, Department of Urology, Graduate School of Medical Sciences, Kyushu University, Kyushu.
The Cancer of the Prostate Strategic Urologic Research Endeavor was supported until 2007 by TAP Pharmaceutical Products Inc (Lake Forest, IL), and currently is funded internally by the University of California, San Francisco Department of Urology. The Japan Study Group of Prostate Cancer (J-CaP) project is supported in part by Takeda Pharmaceutical Company Limited (Tokyo, Japan). This work was also supported by National Institutes of Health/National Cancer Institute University of California-San Francisco Specialized Program of Research Excellence p50 c89520.
Presented in part at the Annual Meeting of the American Urological Association April 25-30, 2009, Chicago, IL.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: None Consultant or Advisory Role: None Stock Ownership: None Honoraria: Matthew R. Cooperberg, Abbott Laboratories, Takeda Pharmaceuticals; Shiro Hinotsu, Takeda Pharmaceuticals; Peter R. Carroll, Astra-Zeneca, TAP Pharmaceutical Products, Inc, Takeda Pharmaceuticals; Hideyuki Akaza, Takeda Pharmaceuticals Research Funding: None Expert Testimony: None Other Remuneration: None
Conception and design: Matthew R. Cooperberg, Shiro Hinotsu, Peter R. Carroll, Hideyuki Akaza
Financial support: Peter R. Carroll, Hideyuki Akaza
Administrative support: Jeanette Broering, Peter R. Carroll, Hideyuki Akaza
Provision of study materials or patients: Mikio Namiki, Kazuto Ito, Peter R. Carroll, Hideyuki Akaza
Collection and assembly of data: Shiro Hinotsu, Mikio Namiki, Kazuto Ito, Jeanette Broering
Data analysis and interpretation: Matthew R. Cooperberg
Manuscript writing: Matthew R. Cooperberg, Shiro Hinotsu
Final approval of manuscript: Matthew R. Cooperberg, Shiro Hinotsu, Mikio Namiki, Kazuto Ito, Jeanette Broering, Peter R. Carroll, Hideyuki Akaza