Clinical predictors of survival in men with castration-resistant prostate cancer

Evidence That Gleason Score 6 Cancer Can Evolve to Lethal Disease


  • Informed consent was obtained from the subjects.



This study sought to characterize Modern patients with castration-resistant prostate cancer (CRPC) and identify pretreatment clinical predictors of survival.


A cohort of men with CRPC with and without metastases (M) treated with secondary hormonal therapy (2eHT) and/or chemotherapy (CT) was identified from the authors' institutional database. Associations of patient and disease characteristics at diagnosis, at androgen-deprivation therapy (ADT) initiation, at CRPC index date, and survival were evaluated. CRPC index date was defined as the start date of either 2eHT or CT, whichever came first.


In the cohort of 622 men, 434 men (70%) had M-positive disease; 552 men (89%) received 2eHT and 70 men (11%) received CT as their initial CRPC treatment. There were 410 deaths (66%) at the time of analysis. Median overall survival (OS) was 35 months (quartile 1, quartile 3: 21 months, 61 months). In multivariate analyses, higher biopsy Gleason score, the presence of M at ADT initiation, shorter time from ADT start to CRPC, higher prostate-specific antigen and poorer Eastern Cooperative Oncology Group performance status at CRPC and M at CRPC were predictive of shorter OS. Interestingly, whereas some men with biopsy Gleason scores of 6 died of their disease (N = 42), they had a longer OS after CRPC compared with those with a Gleason score ≥ 7.


This large retrospective study of patients with CRPC in a tertiary cancer center shows that biopsy Gleason score of 6 is associated with a less aggressive CRPC course, and the impact that M at ADT initiation and CRPC have on outcome is quantified. Cancer 2013;119:2990—2998. © 2013 American Cancer Society.


The number of treatment options demonstrating a survival benefit for men with castration-resistant prostate cancer (CRPC) has greatly increased since 2004, when docetaxel was shown to provide a survival benefit (eg, abiraterone, sipuleucel-T).[1-4]

When patients develop CRPC, the ability to evaluate prognosis is critical in order to counsel patients effectively about long-term outlook and for identification of optimal treatment. In the CRPC population, pretreatment nomograms predicting survival have been developed and validated and are widely used.[5, 6] In addition, a series of prognostic models have been published in relation to clinical trials evaluating response to certain drugs; however, the majority of those studies were conducted solely in patients with metastatic CRPC.[7]

In our practice, 70% of patients have metastases when they develop CRPC and 30% do not.[8] In general, the diagnosis of CRPC is followed by treatment with sequential secondary hormonal therapies (2eHT) before initiating chemotherapy. However, a subset of patients have rapidly progressive disease and receive chemotherapy rather than 2eHT when they develop CRPC. A better understanding of the clinical course of patients placed on 2eHT versus chemotherapy at CRPC and identifying clinical predictors of outcomes can help physicians negotiate complex treatment decisions.

The objective of this study was to characterize a modern CRPC patient population and to identify clinical predictors of survival in men with CRPC before treatment using our institutional clinical database of patients treated with a series of CRPC treatments at a large tertiary cancer center.



This cohort was generated from the Prostate Clinical Research Information System (CRIS) database at Dana-Farber Cancer Institute (DFCI).[9] The DFCI Institutional Review Board approved this study. All patients provided written informed consent allowing analysis of clinical data for research purposes. Data are stored in an Oracle database (Oracle, Redwood Shores, Calif). In general, all patients seen at DFCI with a diagnosis of prostate cancer are approached to participate in the database. The current version of the Prostate CRIS database became available for prospective data entry in November 2001. The consent rate is 86%. All clinical data in CRIS is quality controlled (QC) on a regular basis; the most recent QC demonstrated a ≤ 10% error rate.


Eligible patients included those with CRPC treated with 2eHT and/or chemotherapy and who received primary androgen-deprivation therapy (ADT) after 1996. Patients were excluded from the analysis if systemic data problems were identified during QC or if they received abiraterone, because this agent was not approved at the time of the current study's data cutoff date.

Quality Control

For this study, a physician (M.N.) reviewed the medical records of approximately 10% of the original cohort (N = 822). Overall, 378 data points were reviewed, and the database's data were 86% concordant with the data generated from this review.

Statistical Analysis

Patient and disease characteristics were summarized as numbers and percentages for categorical variables; continuous variables were summarized with either a mean and standard deviation or a median and interquartile range. The CRPC index date was defined as start date of first treatment for CRPC.

Duration of primary ADT was defined as time between date of first injection of luteinizing hormone-releasing hormone agonist/antagonist or date of bilateral orchiectomy and CRPC index date. Neoadjuvant and adjuvant ADT were not included in the duration of primary ADT. Duration of subsequent therapy for CRPC was calculated as the time between the start date of treatment and the start date of subsequent treatment. The primary outcome variables were overall survival (OS) and maximal percent prostate-specific antigen (%PSA) change from baseline during first CRPC treatment. Maximal %PSA change was calculated as percent difference between the smallest PSA value during the first CRPC treatment and the PSA value prior to initiation of first CRPC treatment.

When patients are still alive or lost to follow-up, OS was censored either at date of last follow-up, last known PSA level, or imaging date, whichever comes latest. The OS distribution (median and 95% confidence intervals [CIs]) was estimated using the Kaplan-Meier method. Univariate and multivariate Cox proportional hazards modeling was used to assess the association of patient and disease characteristics prior to CRPC treatment initiation with OS from CRPC. Univariate and Cox proportional hazards modeling was conducted to evaluate the association between OS from second-line CRPC treatment and maximal %PSA decline from baseline during first-line CRPC treatment. Continuous characteristics were dichotomized at the median value within the cohort. An unknown category was created for any missing data in order to include all patients in the analysis.

The statistical analysis was performed using SAS, version 9.3 (SAS Institute, Cary, NC), and P < .05 (2-sided) was considered to be statistically significant.


Study Cohort

A total of 822 patients treated with 2eHT and/or chemotherapy for CRPC were identified in our database[9]; 200 patients were excluded from further analysis for lack of documentation of 2eHT or chemotherapy (CT) for CRPC (N = 56), absence of follow-up data (N = 41), ADT start date in 1995 or before (N = 75), or database error (N = 28). The final cohort consisted of 622 patients. Subjects were initiated with primary ADT between January 1996 and June 2010. Median follow-up after diagnosis was 7.3 years (quartile 1, quartile 3: 3.8 years, 11.4 years).

Patient and Disease Characteristics

Overall, 552 men (89%) received 2eHT as first CRPC treatment (2eHT-first) whereas 70 men (11%) received CT as first CRPC treatment (CT-first) (Table 1). Notably, 268 (43%) patients had biopsy Gleason score of 8 or greater disease, and 235 (38%) had clinically localized disease at diagnosis and were considered high-risk according to D'Amico criteria (T2c or PSA > 20 or Gleason score ≥ 8); only 32 men (5%) had low-risk disease (≤ T2a and PSA ≤ 10 ng/mL and biopsy Gleason score ≤ 6), 26 men (5%) in the 2eHT-first group and 6 (9%) men in the CT-first group. Ninety-five men (15%) had metastatic disease at diagnosis.

Table 1. Patient and Disease Characteristics of CRPC Patientsa
 CRPC Patients2nd HT as First CRPC TreatmentChemotherapy as First CRPC Treatment 
 N = 622N = 552N = 70P Value
  1. a

    Values given as N (%) or median (Q1, Q3).

  2. Abbreviations: ADT, androgen-deprivation therapy; CRPC, castration-resistant prostate cancer; HT, hormonal therapy; PSA, prostate-specific antigen; RP, radical prostatectomy; RT, radiation therapy; SD, standard deviation.

At Diagnosis
Age, y, mean ± SD (median) (n = 586)61.5 ± 8.4 (61)61.7 ± 8.4 (62)59.9 ± 8.4 (60).1107
White587 (94%)525 (95%)62 (89%).0738
Black20 (3%)15 (3%)5 (7%)
Other15 (2%)12 (2%)3 (4%)
PSA, ng/mL, median (n = 513)
 (7.0 - 53.8)(7.0 - 49.0)(5.9 - 111.4) 
Clinical T stage
T1256 (41%)223 (40%)33 (47%).7158
T2145 (23%)129 (23%)16 (23%)
T334 (5%)32 (6%)2 (3%)
T42 (0%)2 (0%)0 (0%)
Tx185 (30%)166 (30%)19 (27%)
Clinical N stage
N0196 (32%)175 (32%)21 (30%).2043
N147 (8%)38 (7%)9 (13%)
Nx379 (61%)339 (61%)40 (57%)
Clinical M stage
M0211 (34%)189 (34%)22 (31%).0341
M195 (15%)77 (14%)18 (26%)
Mx316 (51%)286 (52%)30 (43%)
Biopsy Gleason score
≤681 (13%)72 (13%)9 (13%).4530
7185 (30%)169 (31%)16 (23%)
≥8268 (43%)232 (42%)36 (51%)
Unknown88 (14%)79 (14%)9 (13%)
Primary local therapy
RP195 (31%)174 (32%)21 (30%).4954
RT193 (31%)175 (32%)18 (26%)
RP + RT15 (2%)14 (3%)1 (1%)
None219 (35%)189 (34%)30 (43%)
Year of diagnosis
Before 199598 (16%)95 (17%)3 (4%).0004
1995–1999176 (28%)164 (30%)12 (17%)
2000–2004190 (31%)163 (30%)27 (39%)
2005–2011122 (20%)98 (18%)24 (34%)
Unknown36 (6%)32 (6%)4 (6%)
D'Amico risk categories
Low-risk32 (5%)26 (5%)6 (9%).1719
Intermediate-risk275 (44%)251 (45%)24 (34%)
High-risk235 (38%)203 (37%)32 (46%)
Unknown80 (13%)72 (13%)8 (11%)
At ADT Initiation For Recurrent Disease or Metastatic Disease
Age, mean ± SD (median)65.3 ± 9.3 (65)65.6 ± 9.3 (66)62.6 ± 8.7 (62).0096
Duration of primary ADT [1], months, median192112<.0001
 (10 - 37)(11 - 40)(6 - 23) 
Type of ADT, n (%)
LHRH monotherapy or orchiectomy281 (45%)265 (48%)16 (23%)<.0001
Combined ADT341 (55%)287 (52%)54 (77%)
PSA, ng/mL, median (n = 388)18.417.829.1.1946
 (6.3 - 80.9)(6.0 - 76.0)(10.3 - 105.8) 
At CRPC Index Date
Time from ADT start, months, median19.02112<.0001
 (10 - 37)(11 - 40)(6 - 23) 
Age, mean ± SD (median)67.5 ± 9.6 (68)68.0 ± 9.6 (69)64.0 ± 9.1 (64).0012
Time from prostate cancer diagnosis, years, median (n = 586)<.0001
 (1.7 - 8.3)(1.8 - 8.5)(0.9 - 4.0) 
Status of metastases
Metastases434 (70%)368 (67%)66 (94%)<.0001
No metastases188 (30%)184 (33%)4 (6%)
PSA, ng/mL, median (n = 521)<.0001
 (2.2 - 24.7)(2.1 - 18.4)(4.1 - 140.0) 
Year of CRPC index date
2000 or before42 (7%)42 (8%)0 (0%).0690
2001–2002103 (17%)94 (17%)9 (13%)
2003–2004117 (19%)103 (19%)14 (20%)
2005–2006122 (20%)111 (20%)11 (16%)
2007–2008127 (20%)108 (20%)19 (27%)
2009–2011111 (18%)94 (17%)17 (24%)
At Chemotherapy StartN = 416N = 346N = 70 
Age, mean ± SD (median)67.4 ± 9.7 (68)68.1 ± 9.7 (69)64.0 ± 9.1 (64).0013
PSA, ng/mL, median (n = 371)102.4125.031.3<.0001
 (28.8 - 305.0)(41.8 - 310.2)(4.1 - 140.0) 
Status of metastases
Metastases407 (98%)341 (99%)66 (94%).0252
No metastases9 (2%)5 (1%)4 (6%)
Time from diagnosis, months, median (n = 394)616929<.0001
 (29.0 - 105.0)(35 - 112)(11 - 48) 
Time from CRPC index date, months, median10130<.0001
 (3.0 - 23.0)(6 - 26)(0 - 0) 

Men treated with chemotherapy as initial CRPC therapy were significantly more likely to have metastatic disease at diagnosis and to exhibit a shorter time from diagnosis to CRPC index date as well as shorter time from ADT start to CRPC index date.

At CRPC index date, patients who received CT-first were 5 years younger (P = .0012), had significantly higher PSA (median PSA = 31.3 ng/mL [quartile 1, quartile 3: 4.1 ng/mL, 140.0 ng/mL] versus 7.0 ng/mL [quartile 1, quartile 3: 2.1 ng/mL, 18.4 ng/mL] in the 2eHT-first group, P < .0001) and were more likely to have metastatic disease (94% versus 67%; P < .0001) than the 2eHT-first group. Chemotherapy was given first in 4 men with metastasis (M)-negative disease because their PSA doubling time on primary ADT was less than 3 months.

Treatment Patterns and Response to First CRPC Treatment

Patients treated with 2eHT-first received a median of two 2eHT regimens. Median duration of 2eHT was 23 months (range, 9-49 months). Among 416 patients who received CT at any time, 59% (N = 245) received at least 2 CT regimens and 27% received at least 3 CT regimens. Overall median duration of CT was 17 months (range, 10-27 months). Docetaxel was the most commonly used first-line CT agent, followed by mitoxantrone.

Median maximal %PSA change on any type of first CRPC treatment was −25% (quartile 1, quartile 3: increase of 32%, decline of 70%). For 2eHT-first patients, median maximal PSA decline was 24% (quartile 1, quartile 3: increase of 33%, decrease of 69%) and for CT-first patients, it was 30% (quartile 1, quartile 3: increase of 12%, decrease of 75%). Median time to maximal PSA decline during the any-first CRPC treatment was 2.0 months (quartile 1, quartile 3: 1 months, 4 months) (N = 496). In univariate analysis, a maximal %PSA decline < 25% (from baseline during any-first CRPC treatment regardless of the type of treatment) was significantly associated with a shorter OS from second-line CRPC treatment (hazard ratio [HR] = 0.64, P < .0001).

Overall Survival

There were 410 deaths (66%) at the time of analysis. Median OS from CRPC index date was 35 months (quartile 1, quartile 3: 21 months, 61 months) for the entire cohort, 38 months (quartile 1, quartile 3: 24 months, 65 months) in the 2eHT-first group and 15 months (quartile 1, quartile 3: 9 months, 34 months) in the CT-first group. Median OS from CT start was 18 months (quartile 1, quartile 3: 10 months, 27 months) and 15 months (quartile 1, quartile 3: 9 months, 34 months) in the 2eHT-first group and CT-first group, respectively.

Univariate Analysis: OS From CRPC Index Date

Statistically significant predictors of longer OS from CRPC index date included biopsy Gleason score ≤ 6, PSA level at ADT start (< 18.4 ng/mL) and at CRPC index date (< 7.8 ng/mL), absence of M disease, time from ADT start to CRPC index date (duration of primary ADT; > 19 months) and Eastern Cooperative Oncology Group (ECOG) performance status 0 at CRPC index date (Table 2). PSA at diagnosis was not associated with OS from CRPC index date.

Table 2. Univariate Analyses of Predictors of OS From CRPC Index Date
FactorPatients (N)Events (N)Median (range)P Valuea
  1. a

    P values are from a log-rank test of equality across strata.

  2. Abbreviations: ADT, androgen-deprivation therapy; CRPC, castration-resistant prostate cancer; ECOG, Eastern Cooperative Oncology Group; OS, overall survival; PSA, prostate-specific antigen.

At diagnosis
Biopsy Gleason score    
≤ 6814846 (26–70).0232
718512638 (24–70)
≥ 826817330 (18–57)
PSA level at diagnosis    
Below median (<15.0 ng/mL)25915740 (24–61).2595
Above median (>15.0 ng/mL)25417234 (18–66)
Presence of metastases at diagnosis   
Yes805728 (14–45).0077
No54235337 (22–65)
At ADT start
 PSA level at ADT
Below median (<18.4 ng/mL)19411139 (24–73).0092
Above median (>18.4 ng/mL)19413534 (17–56)
Presence of metastases at ADT    
Yes32022230 (18–55)<.0001
No30218840 (25–68)
At CRPC index date
Below median (<19 months)31423328 (16–47)<.0001
Above median (>19 months)30817745 (27–70)
PSA level at CRPC    
Below median (<7.8 ng/mL)26214744 (27–74)<.0001
Above median (>7.8 ng/mL)25918928 (16–47)
ECOG performance status at CRPC   
033819939 (24–67)<.0001
>0866818 (11–32)
Presence of metastases at CRPC   
Yes43430330 (18–51)<.0001
No18810750 (31–78)

Multivariate Analysis

In multivariable analysis, statistically significant factors of longer OS were biopsy Gleason score ≤ 6, absence of metastatic disease at ADT initiation, longer time from ADT start to CRPC index date (duration of primary ADT), lower PSA level, ECOG performance status 0, and absence of metastatic disease at CRPC index date (Table 3). The strongest predictor was ECOG performance status of 0.

Table 3. Multivariate Analysis: Overall Survival From CRPC Index Date (N = 622)
 Hazard Ratio95% CIP Value
  1. Abbreviations: ADT, androgen-deprivation therapy; CRPC, castration-resistant prostate cancer; ECOG, Eastern Cooperative Oncology Group; OS, overall survival; PSA, prostate-specific antigen.

At diagnosis
 Biopsy Gleason score
≤ 60.71(0.51–0.99).0445
≥ 8Reference  
Presence of metastases0.96(0.69–1.33).8025
At ADT start
 PSA level (ng/mL)
Below median0.94(0.71–1.24).6752
Above medianReference  
Presence of metastases0.70(0.53–0.92).0097
At CRPC index date
 Time from ADT start (duration of primary ADT)
Below medianReference  
Above median0.57(0.47–0.70)<.0001
PSA level (ng/mL)   
Below median0.58(0.46–0.73)<.0001
Above medianReference  
 ECOG performance status
Presence of metastases2.15(1.59–2.91)<.0001

Gleason ≤ 6 Cancer Can Progress to CRPC

A total of 81 (13%) men in the cohort who developed CRPC had Gleason ≤ 6 disease at diagnosis. Given the significant association of Gleason ≤ 6 with OS in this cohort, we further characterized the differences in clinical characteristics between men with biopsy Gleason ≤ 6 and Gleason ≥ 7 disease (Table 4). At diagnosis, men with biopsy Gleason ≤ 6 disease were slightly older than those with higher Gleason scores (62.9 versus 61.2 years, P = .040). None of the patients with biopsy Gleason scores ≤ 6 had N1 or M1 disease. Thirty-two men (40%) were classified as having low-risk disease, and 93% (75) received definitive local therapy, as compared to 61% (276) men with Gleason 7 or above at diagnosis (P < .0001). Men with biopsy Gleason score ≤ 6 disease experienced a more indolent course than men with higher Gleason scores: duration of primary ADT was 10 months longer (28 versus 18 months) (P = .0016), median time from diagnosis to CRPC index date was 5.6 years longer (9 versus 3.4 years) (P < .0001), and median time from diagnosis to CT initiation was approximately 6 years longer (10 versus 4 years) in patients with Gleason ≤ 6 compared to those with Gleason score of 7 to 10 (P < .0001). At the time of analysis, there were 48 deaths (42 prostate cancer deaths). There were 4 non–prostate cancer deaths (colon cancer, pancreatic cancer, vascular disorder of intestine, and sepsis) and 2 deaths from unknown causes. Median OS from CRPC index date was 12 months longer in patients with biopsy Gleason ≤ 6 disease (P = .0493) (Fig. 1).

Figure 1.

Kaplan-Meier curves are shown for overall survival from castration-resistant prostate cancer index date stratified by biopsy Gleason score.

Table 4. Comparison of Patient and Disease Characteristics of CRPC Patients With a Gleason Score ≤6 Versus Those With a Gleason Score 7 or Highera
 Gleason Score ≤6 N = 81 (15%)Gleason Score ≥7 N = 453 (85%)P Value
  1. a

    Values given as N (%) or median (Q1, Q3).

  2. Abbreviations: ADT, androgen-deprivation therapy; CRPC, castration-resistant prostate cancer; HT, hormonal therapy; PSA, prostate-specific antigen; RP, radical prostatectomy; RT, radiation therapy; SD, standard deviation.

At Diagnosis
Median follow-up time, years (n = 552)12.46.5<.0001
 (9.1 - 16.0)(3.4 - 10.2) 
Age, year, mean ± SD (median) (n = 552)62.9 ± 6.5 (64)61.2 ± 8.7 (61).0430
White79 (98%)425 (94%).4057
Black1 (1%)16 (4%)
Other1 (1%)12 (3%)
PSA level, ng/mL (n = 490)9.016.7<.0001
 (6.0 - 16.4)(7.3 - 71.4) 
Clinical T stage
T141 (51%)207 (46%).3239
T225 (31%)115 (25%)
T32 (2%)31 (7%)
T40 (0%)2 (0%)
Tx13 (16%)98 (22%)
Clinical N stage
N028 (35%)164 (36%).0059
N10 (0%)47 (10%)
Nx53 (65%)242 (53%)
Clinical M stage
M034 (42%)173 (38%)<.0001
M10 (0%)95 (21%)
Mx47 (58%)185 (41%)
Biopsy Gleason score
≤681 (100%)0 (0%) 
70 (0%)185 (41%)
≥80 (0%)268 (59%)
Primary local therapy
RP41 (51%)122 (27%)<.0001
RT31 (38%)143 (32%)
RP + RT3 (4%)11 (2%)
None6 (7%)177 (39%)
Year of initial diagnosis
Before 199537 (46%)44 (10%)<.0001
1995–199926 (32%)134 (30%)
2000–200415 (19%)161 (36%)
2005–20112 (2%)110 (24%)
Unknown1 (1%)4 (1%)
D'Amico risk categories
Low-risk32 (40%)0 (0%) 
Intermediate-risk24 (30%)246 (54%)
High-risk12 (15%)207 (46%)
Unknown13 (16%)0 (0%)
At ADT Initiation For Recurrent Disease or Metastatic Disease
Age, mean ± SD (median)70.0 ± 7.3 (70)64.1 ± 9.1 (64)<.0001
Type of ADT
LHRH monotherapy or orchiectomy35 (43%)207 (46%).6790
Combined ADT46 (57%)246 (54%)
PSA level, ng/mL (n = 346)16.418.4.4960
 (7.8 - 54.9)(6.3 - 92.3) 
At CRPC Index Date
Time from ADT start, months2818.0016
 (16 - 46)(10 - 34) 
Age, mean ± SD (median)72.6 ± 7.7 (73)66.3 ± 9.5 (66)<.0001
Time from diagnosis, years (n = 552)9.03.4<.0001
 (5.7 - 11.5)(1.4 - 7.0) 
Status of metastases, n (%)
Metastases46 (57%)327 (72%).0054
No metastases35 (43%)126 (28%)
PSA level, ng/mL (n = 472)
 (1.6 - 15.9)(2.1 - 27.1) 
At Chemotherapy StartN = 52N = 308 
Age, mean ± SD (median)73.0 ± 7.7 (74)66.2 ± 9.6 (66)<.0001
PSA level, ng/mL (n = 321)98.0113.5.9772
 (31.3 - 232.0)(32.7 - 319.2) 
Status of metastases
Metastases50 (96%)303 (98%).2830
No metastases2 (4%)5 (2%)
Time from diagnosis, months (n = 356)12150<.0001
 (77 - 153)(26 - 86) 
Time from CRPC index date, months179.0512
 (6 - 33)(3 - 23) 

Gleason Upgrading in Men With Biopsy Gleason Score ≤ 6

Among 44 patients (54%) who had biopsy Gleason score ≤ 6 and radical prostatectomy (RP), RP pathology was available in 42 men (52%) (2 cases with RP Gleason grading unknown). Gleason score was upgraded to Gleason score ≥ 7 in 32 patients (76%) after RP. Notably, in only 15 cases were both biopsy and RP Gleason scores reviewed at the same institution. The remaining 27 cases were reviewed at outside hospitals. There were 18 deaths in 32 patients whose Gleason score was upgraded and 7 deaths in 10 patients whose Gleason score was not upgraded.


This study represents the largest cohort to date reporting the natural history of CRPC treated at a large tertiary cancer center. In the current cohort, clinical predictors of OS include biopsy Gleason score, presence of metastatic disease at ADT initiation and at CRPC (not at diagnosis), duration of primary hormonal therapy, PSA at CRPC, as well as ECOG performance status at CRPC.

In this cohort, 10% of all CRPC patients received CT first for CRPC. We found that patients who receive CT first for CRPC differ substantially from those who first receive 2eHT; the CT-first group were slightly younger at CRPC, demonstrated a shorter duration of response to primary ADT, had higher PSA levels, and were more likely to have metastatic disease when they developed CRPC. In our current practice, medical oncologists are concerned that this type of disease is less likely to be responsive to hormonal therapy and are more inclined to start CT as initial treatment for CRPC. It is interesting that physician practice, by initiating CT instead of starting 2eHT for first-line CRPC treatment, turned out to be prognostic.

We also examined the association of response to first-line CRPC treatment and OS. Maximal PSA decline < 25% from baseline during the first-line CRPC treatment (regardless of type of treatment) was significantly associated with shorter survival. Notably, our findings are similar to previously published results of the TAX327 and SWOG 99-06 studies in which PSA decline of 30% within 3 months of treatment initiation provided surrogacy for OS.[10, 11] The presented 30% cutoff had a higher degree of statistical surrogacy than traditional 50% cutoff by using the most stringent surrogacy criterion, the proportion of treatment effect explained (PTE) in SWOG 99-06.[12]

In this study, we found that 13% of men with CRPC had biopsy Gleason ≤ 6 disease at diagnosis. The median follow-up of this patient population was 12.4 years: median time from diagnosis to CRPC in these men was 9 years (compared with 3.4 years in men with higher Gleason scores). Even after developing CRPC, biopsy Gleason ≤ 6 cancer continued to behave less aggressively than Gleason ≥ 7 disease. These data are informative given the recent interest in active surveillance for men with low-risk prostate cancer.

With early detection of prostate cancer by PSA screening, the number of men diagnosed with low-risk disease has significantly increased. The recent publication of Prostate Cancer Intervention versus Observation Trial (PIVOT) study provided some insight into the natural history of low-risk disease in the post-PSA era.[13] In that study, a total of 731 men were randomized to RP or watchful waiting (WW) and followed for a median of 10 years. Although the study was underpowered, consisted of older patients (mean age 67), and 20% men on WW were treated with curative intent, a subgroup analysis of 148 men with low-risk disease showed no benefit to RP over WW in OS or prostate cancer–specific death.

However, with longer follow-up, the current study identified a subset of men with biopsy Gleason ≤ 6 cancer that progressed to a lethal disease after treatment. The magnitude of this subset is no doubt influenced by referral bias to our tertiary cancer center, as is the fact that only 40% of these men with biopsy Gleason ≤ 6 in our cohort had low-risk disease. However, the existence of these men identified with lethal disease after many years should be recognized by all practitioners. Observation for men with low-risk prostate cancer has gained increasing attention in recent years, and the identification of men with lethal biopsy Gleason ≤ 6 disease in our cohort underscores the need for improved prognostic tools such as biomarkers to distinguish lethal from indolent disease at diagnosis.

Retrospective database studies of prostate cancer are plagued by limitations regarding Gleason scoring, in particular the shift to higher Gleason scores over time and interobserver variability among pathologists.[14] This cohort consists of a heterogeneous population spanning more than 30 years (prostate cancer diagnosis ranged from 1977 to 2010). In our referral clinic, central pathology review of all outside biopsy and RP samples is impossible. Therefore, our conclusions regarding the natural history of Gleason score 6 disease reflect the heterogeneity of community grading as well as changing practices with regard to Gleason grading and to biopsy technique over these years. In addition, it has been reported consistently in the literature that approximately 30% of cancers are upgraded from biopsy Gleason score at RP (consistent with our unpublished data; Yoo S et al, March 2013). In this cohort, 76% of Gleason score 6 cancers were upgraded. However, given that our pathology was not centrally reviewed, it is difficult to assess the impact of this finding. Because of these limitations, it must be emphasized that conclusions regarding Gleason scoring in this article reflect the grading practices in the community over the past 3 decades.

The pathological features at RP for low-risk disease are not uniformly favorable; increasingly, the proportion of Gleason pattern 4 in RP specimen is being recognized as a critical factor in predicting biochemical recurrence and prostate cancer specific mortality.[15] According to recent studies, tertiary 4 pattern in Gleason ≤ 6 cancer could increase the risk of relapse.[16] Because of the retrospective nature of our study, evaluation of the volume of high grade cancer in initial biopsy specimens is not feasible; however, we could speculate that tertiary 4 patterns might have played a role in the development of lethal Gleason ≤ 6 cancers in our cohort. Characterization of pathological Gleason patterns using RP specimens to identify molecular markers that define the aggressiveness of low-risk disease is currently ongoing.

In this study, we present a natural history of CRPC patients treated at a large tertiary cancer center and identify clinical predictors of survival in contemporary patients with CRPC regardless of the status of metastatic disease. The present cohort demonstrates that despite mounting evidence that many men with low-risk prostate cancer are treated unnecessarily, biopsy Gleason ≤ 6 cancer can progress to CRPC after a median of 12.4 years; however, even in these men with lethal disease, initial Gleason score continues to be a useful tool to predict outcomes. Further studies are warranted to understand and predict the different biological behavior of individual cancers in order to select the most effective treatment for individual patients. The predictors identified in this study must be validated in a prospective study.


No specific funding was disclosed.


Mr. Lefebvre, Ms. Lafeuille, and Dr. Duh are employees of Analysis Group, Inc., a consulting company that has received research grants from Janssen Scientific Affairs, LLC.