The oncological outcome and validation of Japan Cancer of the Prostate Risk Assessment score among men treated with primary androgen-deprivation therapy

Abstract

Purpose

Although androgen-deprivation therapy (ADT) for prostate cancer is initially effective, most tumors eventually recur even during ADT. To predict their prognosis, the Japan Cancer of the Prostate Risk Assessment (J-CAPRA) score was developed. However, there is no validation of this model using data from a single institution. Therefore, in this study, we clarified the oncological outcome of primary ADT and its prognostic factors, as well as validated the J-CAPRA score model in our institution.

Methods

This study included 248 Japanese patients with hormone-naïve prostate cancer who were treated with primary ADT from 1996 through 2012. The oncological outcome and prognostic significance of several clinicopathological factors were analyzed. Also, J-CAPRA risk stratification model was validated in this cohort.

Results

During a median follow-up period of 42.2 months, the median progression-free survival (PFS) and overall survival (OS) were 89.3 and 103.3 months, respectively. Multivariate analysis identified clinical T-stage and M-stage for PFS and cancer-specific survival (CSS) and clinical M-stage for OS as significant predictors. The accuracy of J-CAPRA score model for predicting PFS, CSS, and OS was validated by high c-indices.

Conclusions

This study demonstrated the use of the J-CAPRA score system for predicting PFS, CSS, and OS among Japanese men treated with primary ADT in a single institution.

Electronic supplementary material

The online version of this article (doi:10.1007/s00432-014-1828-7) contains supplementary material, which is available to authorized users.

Introduction

Prostate cancer is one of the most common cancers among men of developed countries. Although the incidence of advanced prostate cancer has been decreasing in Western countries, contributing to a decreasing mortality rate, probably due to widespread prostate-specific antigen (PSA) screening, 5 % of men with prostate cancer in the USA are still diagnosed at advanced stages with nodal or distant metastatic disease (Siegel et al. 2012). However, Asian countries including Japan have experienced consistent mortality rates owing to various reasons, including wide-spreading Western-style diet and lifestyle and an increase in the aged-population as well as inadequate PSA screening (Center et al. 2012). In Japan, 10–20 % of men with prostate cancer have metastatic disease at the time of diagnosis, probably owing to inadequate PSA screening (Cancer Registration Committee of the Japanese Urological Association 2005; Fujimoto et al. 2011).

Androgen-deprivation therapy (ADT), which reduces androgen production and inhibits androgen action in prostate cancer cells, is a mainstay for the treatment of recurrent or advanced prostate cancer since its discovery by Huggins and Hodges (1941). Additionally, ADT is often used to treat localized prostate cancer without established evidence that it is effective. In Japan, the Japan Study Group of Prostate Cancer (J-CaP), a large, multicenter, population-based database of patients undergoing primary ADT to investigate their outcomes, was established and has been maintained since 2001. Based on the data from approximately 20,000 cases in the J-CaP study, a risk stratification tool predicting prognosis in men receiving primary ADT, the Japan Cancer of the Prostate Risk Assessment (J-CAPRA) score, was developed (Cooperberg et al. 2009). The accuracy of J-CAPRA score was externally validated for predicting cancer-specific survival (CSS) using data from a community-based cancer registry in the USA called Cancer of the Prostate Strategic Urologic Research Endeavor, CaPSURE (Cooperberg et al. 2009). However, there remains the possibility that ethnic differences between Japan and the USA, also lower reliability due to community-based registration, may affect the accuracy of validation result. Recently, the J-CAPRA score for predicting progression-free survival (PFS), cancer-specific survival (CSS), and overall survival (OS) has been validated among Japanese patients who were treated with combined androgen blockade (CAB) using bicalutamide as an antiandrogen agent in multiple institutions (Kitagawa et al. 2013). However, it has been found that the outcome of primary ADT differs between CAB and surgical or medical castration mono-therapy by meta-analysis in several large trials (Prostate Cancer Trialists’ Collaborative Group 2000) and in a double-blinded, randomized study in Japan (Akaza et al. 2009), although the survival benefit is not large. Therefore, the validation of J-CAPRA score may be influenced by the different mode of primary ADT. Additionally, because of the multi-institutional nature of the studies, diagnostic and therapeutic inconsistencies between institutions may also have affected the results.

To date, there has been no study reporting the validation of J-CAPRA score among men treated with primary ADT by castration and/or antiandrogen agent in a single institution. The prediction tool is valuable for both physicians and patients to predict outcome and to indicate the need for more aggressive treatment. Therefore, in this study, we investigated the outcome and prognostic factors in men with prostate cancer treated by primary ADT and validated the J-CAPRA score using the cohort in a single institution.

Patients and methods

This study enrolled patients with prostate cancer treated with primary ADT at Kyushu University Hospital (Fukuoka, Japan) from 1996 to 2012. This study was approved by the institutional review board. All patients were histopathologically diagnosed with adenocarcinoma of the prostate. The cases registered to J-CaP database were excluded. In total, 202 men were biopsied at Kyushu University Hospital, while the remaining 46 men were biopsied at another institution and the biopsy specimens from 13 cases were reviewed at our institution. Patients who received local treatments with ADT, or received other treatments before disease progression, were excluded. After 2006, at Kyushu University Hospital, Gleason score was evaluated according to the modification at 2005 International Society of Urological Pathology (ISUP) Consensus Conference (Epstein et al. 2005). Clinical staging was determined in accordance with the unified TNM criteria based on the results of a digital rectal examination, transrectal ultrasound, computed tomography, magnetic resonance imaging, and bone scan (International Union Against Cancer 1997). All patients were primarily treated by ADT with surgical castration or medical castration using a luteinizing hormone-releasing hormone agonist (goserelin acetate or leuprorelin acetate) and/or an antiandrogen agent (bicalutamide, flutamide, or chlormadinone acetate). Of these patients, 186, 57, and five men were treated with CAB, castration alone, and antiandrogen agent alone, respectively. Progressive disease was defined as an increase in serum PSA of >2 ng/mL and a 25 % increase over the nadir, or the appearance of a new lesion or the progression of one or more known lesions classified according to the Response Evaluation Criteria in Solid Tumours (RECIST) (Scher et al. 2008). Radiographic progression was defined as the progression of measurable disease or bone scan progression. Continuous or intermittent ADT was employed based on discussion between physician and patients because the oncological outcome of both treatment modes has been shown to equivalent (Niraula et al. 2013).

All statistical analyses were performed using JMP9 software (SAS Institute, Cary, NC, USA) and STATA 13 (StataCorp LP, College Station, TX, USA). PFS, CSS, and OS were determined using the Kaplan–Meier method and the log-rank statistic was used to compare survival duration across risk groups. Univariate and multivariate analyses were performed using the Cox proportional hazards regression model. The validation of J-CAPRA score was assessed by the c-index (Harrell et al. 1996); 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 (Harrell et al. 1996). The c-indices were calculated and compared using STATA 13 (StataCorp LP). P values <0.05 were considered significant.

Results

This study enrolled a total of 248 Japanese patients, whose clinical and pathological characteristics are shown in Table 1. The median age of the patients was 74 years (range 46–90 years), and the median PSA was 48.0 ng/mL (range 3.2–8,546 ng/mL) at diagnosis. The Gleason scores of biopsy specimens from 80 patients (32.3 %) and 129 patients (52.0 %) were 7 and >7, respectively. Among them, 133 (53.6 %) had lymph node and/or distant metastases. When compared with the patient characteristics in the J-CaP (Cooperberg et al. 2009), the proportion of patients with high Gleason scores (>7), higher PSA values (over 100 ng/mL), and higher clinical T-stage (T3 and T4) was larger in this study (Table 1), indicating appropriate indication for primary ADT to treat advanced prostate cancer in this cohort. During the median follow-up period of 42.2 months, disease progression, cancer-specific death, and all-cause death occurred in 87 cases (35.1 %), 54 cases (21.8 %), and 81 cases (32.7 %), respectively. Median PFS (Fig. 1a), CSS (Fig. 1b), and OS (Fig. 1c) were 89.3 months, not reached, and 103.3 months, respectively. The 5-year PFS rate (Fig. 1a) was 57.3 %, while the 5-year CSS (Fig. 1b) and OS rates (Fig. 1c) were 74.4 and 66.2 %, respectively.

Table 1.

Clinical characteristics of the study population

Characteristics	Patients no. (%)	J-CAPRA points	J-CaP data no. (%)
Age (years)
≤75	138 (55.6 %)	–	9,934 (51.6 %)
>75	110 (44.4 %)	–	9,332 (48.4 %)
Gleason score
≤6	39 (16.7 %)	0	5,884 (35.1 %)
7	80 (32.3 %)	1	4,821 (28.7 %)
8–10	129 (52.0 %)	2	6,060 (36.2 %)
PSA at diagnosis (ng/mL)
0–10	42 (16.9 %)	0	4,727 (24.6 %)
>10–20	41 (16.5 %)	0	3,713 (19.3 %)
>20–100	75 (30.2 %)	1	5,865 (30.5 %)
>100-500	63 (25.4 %)	2	2,929 (15.3 %)
>500	27 (10.9 %)	3	1,972 (10.3 %)
cT stage
T1c	44 (17.7 %)	0	T1, 4,001 (20.8 %)
T2a	40 (16.1 %)	0	T2, 6,274 (32.6 %)
T2b	11 (4.3 %)	1
T3a	70 (28.2 %)	1	T3, 7,048 (36.6 %)
T3b	37 (14.9 %)	2
T4	46 (18.5 %)	3	T4, 1,943 (10.1 %)
cN stage
N0	161 (64.9 %)	0
N1	87 (35.1 %)	1
cM stage
M0	128 (51.6 %)	0
M1	120 (48.4 %)	3

Fig. 1.

PFS, CSS, and OS in patients with prostate cancer treated with primary ADT. PFS (a), CSS (b), and OS (c) in 248 patients with prostate cancer treated with primary ADT are shown

Using this cohort, we performed univariate and multivariate analyses for PFS based on pretreatment factors (age, Gleason score, PSA at diagnosis, and clinical TNM stage). On univariate analysis, age and the factors employed in J-CAPRA scores, including Gleason score, PSA value at diagnosis, and clinical TNM stage were significant predictors of PFS, and the hazard ratio (HR) increased according to each J-CAPRA score, although there was no difference in PSA value between >100–500 and >500 ng/mL (Table 2). On multivariate analysis, only clinical T-stage and M-stage were significant predictors of PFS (Table 2). Notably, HRs with Gleason score and clinical T-stage were larger compared with those in the J-CaP study, while the HR with PSA level was smaller compared with the J-CaP study (Cooperberg et al. 2009); these results are similar to the cohort in the study by Kitagawa et al. (2013). Table 3 shows the univariate analysis for PFS according to J-CAPRA score. The proportions of patients in our study with high J-CAPRA score were larger than those in the J-CaP study (Cooperberg et al. 2009). The HR roughly increased according to the increase in J-CAPRA score. The categorized scores for risk stratification according to the J-CAPRA score (Cooperberg et al. 2009) were also significant predictors of PFS. The probabilities of PFS at 4 years were 98.3, 68.8, and 27.5 % for those in the low (score, 0–2), intermediate (score, 3–7), and high-risk (score, 8–12) groups, respectively, which were better than in the J-CaP study (Cooperberg et al. 2009) and comparable to the cohort in the Kitagawa et al. (2013) study even when compared among patients treated with CAB (Supplementary Fig. 1A). Figure 2a shows PFS with J-CAPRA risk stratification. The probabilities of median PFS duration in the low-, intermediate-, and high-risk groups were not reached, 81.7 and 23.2 months, respectively.

Table 2.

Univariate and multivariate analyses for PFS

Varible (J-CAPRA)	Patients no.	Univariate		Multivariate
		HR (95 % CI)	P value	HR (95 % CI)	P value
Age (years)
≤75	138	1		1
>75	110	0.46 (0.28–073)	0.0008*	0.89 (0.52–1.47)	0.65
Gleason score
≤6 (0)	39	1		1
7 (1)	80	11.93 (2.48–214.01)	0.0004*	2.58 (0.45–49.20)	0.33
8–10 (2)	129	38.15 (8.36–675.38)	<0.0001*	4.08 (0.70–78.75)	0.13
PSA at diagnosis (ng/mL)
0–20 (0)	83	1		1
>20–100 (1)	75	5.44 (2.38–14.68)	<0.0001*	0.92 (0.35–2.82)	0.88
>100–500 (2)	63	14.70 (6.68–38.81)	<0.0001*	1.45 (0.55–4.50)	0.47
>500 (3)	27	13.89 (5.74–38.66)	<0.0001*	1.20 (0.42–3.96)	0.74
cT stage
T1c, T2a (0)	84	1		1
T2b, T3a (1)	81	13.18 (4.68–55.05)	<0.0001*	3.98 (1.16–19.24)	0.026*
T3b (2)	37	22.30 (7.62–94.82)	<0.0001*	4.84 (1.30–24.99)	0.017*
T4 (3)	46	44.71 (15.97–186.38)	<0.0001*	7.24 (1.93–37.49)	0.0021*
cN stage
N0 (0)	161	1		1
N1 (1)	87	4.80 (3.11–7.56)	<0.0001*	1.34 (0.80–2.26)	0.27
cM stage
M0 (0)	128	1		1
M1 (3)	120	10.01 (5.70–19.03)	<0.0001*	2.60 (1.32–5.54)	0.0050*

* Statistically significant

Table 3.

PFS by J-CAPRA score and risk strata

Varible	Patients no. (%)	J-CaP data no. (%)	HR (95 % CI)	P value	4-Year PFS (%)
J-CAPRA
0	30 (12.1 %)	2,858 (17.1 %)	0.00 (0.00–1.72)	0.10	100
1	35 (14.1 %)	2,332 (14.0 %)	1		96.2
2	16 (6.5 %)	2,253 (13.5 %)	1.21 (0.06–12.66)	0.88	100
3	11 (4.4 %)	1,970 (11.8 %)	4.88 (0.58–40.75)	0.13	83.3
4	16 (6.5 %)	1,489 (8.9 %)	0.00 (0.00–4.43)	0.26	100
5	18 (7.3 %)	1,022 (6.1 %)	5.04 (0.98–36.44)	0.053	68.6
6	9 (3.6 %)	854 (5.1 %)	8.14 (1.35–61.91)	0.024*	71.4
7	14 (5.6 %)	798 (4.8 %)	18.74 (4.78–123.60)	<0.0001*	31.3
8	25 (10.1 %)	849 (5.1 %)	16.96 (4.71–108.25)	<0.0001*	46.6
9	24 (9.7 %)	820 (4.9 %)	15.55 (4.26–99.83)	<0.0001*	39.2
10	18 (7.3 %)	752 (4.5 %)	24.34 (6.69–155.99)	<0.0001*	22.1
11	26 (10.5 %)	11–12, 719 (4.3 %)	45.46 (13.02–287.46)	<0.0001*	7.1
12	6 (2.4 %)		28.22 (6.01–198.50)	<0.0001*	16.7
Risk strata
0–2 (low)	81 (32.7 %)	7,443 (44.6 %)	1		98.3
3–7 (intermediate)	68 (27.4 %)	6,133 (36.7 %)	9.88 (3.34–42.26)	<0.0001*	68.8
8–12 (high)	99 (39.9 %)	3,140 (18.8 %)	34.58 (12.75–142.02)	<0.0001*	27.5

* Statistically significant

Fig. 2.

Prognostic stratification of patients with prostate cancer treated with primary ADT according to J-CAPRA score. PFS (a), CSS (b), and OS (c) in patients stratified by categorized scores are shown

Subsequently, prognostic factors for CSS and OS were examined. On univariate analysis, the factors employed in J-CAPRA scores, including Gleason score, PSA value at diagnosis, and clinical TNM stage were significant predictors of CSS and OS; however, on multivariate analysis, only clinical T-stage and M-stage were significant predictors of CSS (Supplementary Table 1). Moreover, only clinical M-stage was a significant predictor of OS (Supplementary Table 2). Figure  2b, c show CSS and OS with J-CAPRA risk stratification, respectively. Each survival probability decreased according to the increased risk determined by J-CAPRA stratification. The probabilities of CSS and OS showed significant differences among the groups categorized by risk strata. The probabilities of CSS at 4 years in the low-, intermediate-, and high-risk groups were 100, 85.6, and 63.3 %, respectively (Supplementary Table 3). The median duration of CSS in the low-, intermediate-, and high-risk groups were not reached, 115.1 and 59.2 months, respectively. The probabilities of OS at 4 years in the low-, intermediate-, and high-risk groups were 85.1, 75.2, and 61.2 %, respectively (Supplementary Table 4). The median duration of OS in the low-, intermediate-, and high-risk groups were not reached, 97.9 and 59.2 months, respectively. CSS and OS in this study were better than in the J-CaP study (Cooperberg et al. 2009) and comparable to the cohort in the Kitagawa et al. (2013) study even when compared among patients treated with CAB (Supplementary Fig. 1B and Supplementary Fig. 1C) although CSS and OS in intermediated-risk patients was slightly worse. The c-indices of PFS for the continuous and categorized scores were 0.890 [95 % confidence interval (CI) 0.835–0.945] and 0.856 (95 % CI 0.798–0.914), respectively (Fig. 3a). The c-indices of CSS for the continuous and categorized scores were 0.836 (95 % CI 0.773–0.899; Fig. 3b) and 0.786 (95 % CI 0.724–0.848; Fig. 3b), respectively. The c-indices of OS for the continuous and categorized scores were 0.700 (95 % CI 0.616–0.784; Fig. 3c) and 0.671 (95 % CI 0.591–0.751; Fig. 3c), respectively.

Fig. 3.

Validation of J-CAPRA score. Receiver operator characteristic curves of J-CAPRA score (left panel, continuous scores; right panel, categorized scores) for the prediction of PFS (a), CSS (b), and OS (c). The curves describe the association between sensitivity and specificity

Gleason score system was modified at 2005 ISUP Consensus Conference (Epstein et al. 2005), which may affect the accuracy of J-CAPRA score. In addition, period change may alter the patients’ background and compromise the accuracy of J-CAPRA score. Then, we compared the distribution of J-CAPRA score between 1996–2005 and 2006–2012 (Supplementary Table 5), which showed the tendency of increased J-CAPRA score in recently diagnosed patients. Also, the c-indices on PFS and CSS of J-CAPRA score was similar between before and after 2005 ISUP modification (Supplementary Fig. 2), suggesting little impact of period change as well as 2005 ISUP modification on the accuracy of J-CAPRA score.

Discussion

For the last 7 decades, primary ADT using CAB or castration mono-therapy has been the mainstay for the treatment of recurrent or advanced prostate cancer and as an option for the treatment of localized prostate cancer. Concerning localized prostate cancer, long-term outcome of primary ADT has been good, although the therapeutic benefit is marginal (Potosky et al. 2014) and there is the possibility of adverse effects by ADT, such as osteoporosis, decreased lean mass, dyslipidemia, diabetes mellitus, and cardiovascular disease (Ahmadi and Daneshmand 2013). Recent analyses of practice patterns showed that substantial patients with localized cancer are treated with primary ADT, rather than active surveillance, surgical treatment, or radiotherapy (Harlan et al. 2001; Akaza et al. 2006). Conversely, primary ADT for advanced prostate cancer has been less effective and eventually progresses to castration-resistant prostate cancer, which is defined as disease progression despite castrate levels of testosterone (Heidenreich et al. 2014). Thus, the response to primary ADT varies according to the disease nature. To date, various clinical and pathological parameters, such as age (Kimura et al. 2014) and PSA halftime and doubling time (Park et al. 2009) in addition to pathological grade, PSA level and disease extension have been reported to predict the oncological outcome of primary ADT. Nevertheless, there have been no reports of instruments to predict outcomes for any population of patients treated with primary ADT until 2009 (Cooperberg et al. 2009). J-CAPRA score was developed using a large database from a variety of Japanese patients undergoing primary ADT and is valuable to predict the prognosis in patients with not only localized disease, but also those with advanced cancer when treated primarily with ADT. In the present study, we showed that the pretreatment clinical characteristics included in J-CAPRA score were good predictors of PFS on univariate analysis and clinical T-stage and M-stage on multivariate analysis. As these factors are necessary and used for the diagnosis of prostate cancer in a real-world clinical setting, J-CAPRA scores are feasible 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 primarily with castration and/or antiandrogen agent. The accuracy of the J-CAPRA score for predicting PFS has been shown by the c-indices, which were superior to those in the J-CAPRA derivation paper (Cooperberg et al. 2009) and similar to those in another validation study composed of only CAB patients (Kitagawa et al. 2013). Concerning CSS and OS, the present study showed that risk stratification by J-CAPRA score predicted CSS and OS of the patients similar to those in the previous validation study in Japan (Kitagawa et al. 2013). The results obtained in this validation cohort verify that J-CAPRA score is a clinically powerful prediction tool for Japanese men treated with primary ADT.

However, this study revealed several differences from the community-based cohort in the J-CaP study. First, our cohort included a larger proportion of prostate cancer patients at advanced stages compared with J-CaP study. Second, this study showed the significance of Gleason score and clinical T-stage for predicting prognosis, although serum PSA level is not significant, which may be derived from the consistent and accurate diagnosis in our institution. Nevertheless, J-CAPRA score retained its ability to predict accurate prognosis with acceptable levels. Recently, Kitagawa et al. (2013) showed the relevance of J-CAPRA score among Japanese men all treated with CAB at three institutions. Compared with this previous study, our study showed the significance of J-CAPRA score among men treated with various modes of ADT (CAB, castration only, or antiandrogen agent only) at a single institution, characteristics that increase the reliability and decrease noise by using the data from a single institution, and also reflects real-world clinical settings in which patients are treated with various ADT modalities.

It is well known that the prognosis is worse in high-risk prostate cancer patients. Therefore, improvement of oncological outcome in this high-risk group by understanding biological differences among risk groups is critical. It has been reported that inadequate suppression of androgen in serum may be a poor prognostic marker among patients treated with ADT (Morote et al. 2009), suggesting that ADT-resistant prostate cancer require a more complete form of ADT than conventional ADT by castration and antiandrogens. In addition to docetaxel chemotherapy, several novel agents targeting both androgen receptor (AR) signaling and non-AR signaling have recently been or are being developed (Shiota et al. 2013a). Therefore, for men at high risk, novel therapeutic strategies, such as docetaxel chemotherapy during hormone-naïve stages, and the use of novel agents combined with conventional ADT, may be promising (Shiota 2014). Further, the Chemohormonal Therapy versus Androgen Ablation Randomized Trial for Extensive Disease in Prostate Cancer, CHAARTED study has recently shown superior OS among patients with hormone-naïve metastatic prostate cancer when treated with docetaxel chemotherapy and primary ADT, compared with primary ADT alone (Sweeney et al. 2014), which is in line with preclinical experimental results (Shiota et al. 2013b).

In addition to the clinical and pathological parameters included in the J-CAPRA score, several biomarkers have been reported to be predictable. So far, gene expression, gene amplification, gene mutation, and gene variations in the AR and genes related to AR signaling were suggested to predict ADT outcome. Additionally, several physical data, such as body mass index and serum markers, including C-reactive protein, interleukin-6, and circulating tumor cells, were also shown to be predictive markers of ADT (Grivas et al. 2013). In future, the incorporation of these biomarkers into J-CAPRA score system may improve its accuracy.

The present study had several limitations. For instance, the study design was retrospective and the sample size was relatively small. Additionally, imperfect discrimination was found between adjacent J-CAPRA score levels. However, despite these limitations, the results of this study show the significance of J-CAPRA scoring among Japanese men treated with ADT using various modalities in a single institution.

Conclusion

This study showed the use of the J-CAPRA scoring system for predicting PFS, CSS, and OS among Japanese men treated with primary ADT in a single institution. In future, further improvement of outcome by additional treatment and improvement of prediction by incorporating other biomarkers are desired.

Electronic supplementary material

Below is the link to the electronic supplementary material.