Fax: (216) 445-3532
Predictors of clinical metastasis in prostate cancer patients receiving androgen deprivation therapy
Results from CaPSURE
Version of Record online: 27 JUL 2009
Copyright © 2009 American Cancer Society
Volume 115, Issue 19, pages 4470–4476, 1 October 2009
How to Cite
Abouassaly, R., Paciorek, A., Ryan, C. J., Carroll, P. R. and Klein, E. A. (2009), Predictors of clinical metastasis in prostate cancer patients receiving androgen deprivation therapy. Cancer, 115: 4470–4476. doi: 10.1002/cncr.24526
- Issue online: 17 SEP 2009
- Version of Record online: 27 JUL 2009
- Manuscript Accepted: 10 MAR 2009
- Manuscript Revised: 2 MAR 2009
- Manuscript Received: 18 DEC 2008
- prostatic neoplasms;
- neoplasm metastasis;
- antineoplastic agents;
Virtually all patients with prostate cancer who receive androgen deprivation therapy (ADT) will ultimately develop evidence of resistance to treatment. The prognosis for patients who develop metastatic castrate-resistant disease is reported to be poor, with overall survival historically estimated to be 24 to 36 months. The goal of the current study was to identify predictors of clinical disease progression in patients with prostate cancer who were receiving ADT.
Of the 13,740 men with biopsy–proven prostate cancer who were enrolled in the Cancer of the Prostate Strategic Urological Research Endeavor (CaPSURE) database from 1995 to 2007, 4003 men treated with ADT after diagnosis without evidence of metastases at treatment initiation were identified. The primary endpoint was the development of bone metastasis. Clinical and pathologic characteristics were compared between patients who developed metastasis and those who did not using chi-square tests in a Cox proportional hazards regression model.
The mean age of the men in the cohort was 70 years (range, 39-94 years). One hundred ninety-one men (4.8%) progressed to metastatic disease at a median of 18 months from the initiation of ADT (range, 1-139 months). On multivariate analyses, risk category (hazards ratio [HR], 2.58; P < .0001), percent of biopsies positive >33% (HR, 3.36; P = .003), age ≤65 years at diagnosis (HR, 2.11; P = .001, and prostate–specific antigen velocity on ADT (HR, 1.04; P < .001) were found to be significantly associated with the development of metastatic disease after ADT.
Younger men with high–risk disease appear to have worse prognosis than older men with similar disease. This, along with the other prognostic variables established in the current study, may help identify candidates for clinical trials evaluating secondary treatments for patients with castrate-resistant disease. Cancer 2009. © 2009 American Cancer Society.
Since the association between androgens and cancer of the prostate (CaP) growth was discovered, androgen deprivation therapy (ADT) has played an important role in the management of this disease.1 ADT is reported to be 1 of the most effective therapies against any solid tumor and is the cornerstone of treatment for patients with advanced CaP. Although the majority of patients will initially respond to this form of therapy,2 most patients with advanced CaP who receive ADT, if followed long enough, will develop evidence of resistance (ie, castrate-resistant prostate cancer). Historically, overall survival rates in patients presenting with symptomatic bone metastasis in whom ADT was initiated was poor, with death occurring in the majority of patients within 6 to 12 months.3 However, the current use of prostate-specific antigen (PSA) in population screening and diagnosis has led to a downward stage migration in prostate cancer. In addition, the use of PSA in the detection of disease recurrence and progression has led to a change in the population receiving ADT, such that longer survival times have been reported in more recent studies, likely reflecting a lead time effect.4, 5
The ability to predict which patients will ultimately progress while receiving ADT would be of great value to clinicians. This would help identify the population on ADT at highest risk of disease progression who may be candidates for clinical trials. Previous studies have found that PSA kinetics were predictive of survival in patients receiving ADT,4, 6-9 but to our knowledge few studies have assessed other clinical parameters. The Cancer of the Prostate Strategic Urological Research Endeavor (CaPSURE) database provided an opportunity to evaluate other potential clinical predictors of metastasis in the PSA era. This primarily community-based database contains information regarding CaP patients receiving all treatment modalities, including patient and disease characteristics.10 The objective of the current study was to identify both clinical and pathologic predictors of clinical disease progression (ie, development of bone metastasis) in patients with prostate cancer who were receiving ADT.
MATERIALS AND METHODS
CaPSURE is a longitudinal, observational registry of men with biopsy–proven CaP and has been previously described in depth.10, 11 Briefly, urologists recruit patients with newly diagnosed CaP and report complete clinic data and follow-up information concerning diagnostic tests and treatments. Patients are treated according to standard practice, and are followed until death or withdrawal. At the time of our analysis, the database contained information on 13,740 men from 1991 through 2007. At the time of our analysis, 29 urologic practices were contributing data to the registry, with an additional 11 sites active earlier in the project. A total of 34 of these practices were community based, 3 were university centers, and 3 were Veterans Affairs (VA) medical centers.
Of the patients included in the registry, we identified 5201 men who were treated with ADT at any point. Seven hundred seventy-five patients had evidence of metastatic disease when ADT was initiated and therefore were excluded from the study. To allow for adequate follow-up, patients diagnosed in the 2 years before our study (n = 423) were also excluded, resulting in a cohort of 4003 patients (Table 1). There were 197,401 person–months of follow-up for analysis. Year of diagnosis was 1991 through 1994 in 683 (17.1%) patients, 1995 through 1998 in 743 (18.6%) patients, 1999 through 2002 in 1694 (42.3%) patients, and 2003 through 2005 in 883 (22.1%) patients. The primary treatment modality was radiotherapy in 1340 men (33%), radical prostatectomy in 733 men (18%), and watchful waiting in 445 men (11%). The remaining 1485 patients (37%) did not receive local therapy for their disease because of limited life expectancy, significant medical comorbidity, and/or rapid disease progression.
|Mean patient age at diagnosis (range), y||70 (39-94)|
|African American||527 (13%)|
|Mean BMI (n=2897)||27 (13-54)|
|Median no. of medical comorbidities (n=3069)||2 (0-11)|
|PSA at diagnosis, ng/mL|
|Clinical T classification at diagnosis|
|T1, T2a||2399 (60%)|
|T2b, T2c||1278 (32%)|
|T3, T4||188 (4.7%)|
|Biopsy Gleason score|
|Prognostic risk group|
|Mean percent of prostate biopsies involved (n=3559)||45 (4-100)|
|LHRH agonist||2027 (51%)|
|LHRH antagonist||16 (0.4%)|
For the purposes of the current study, we included patients receiving any type of ADT (Table 1). The majority were treated with a luteinizing hormone-releasing hormone (LHRH) agonist (n = 2027; 51%), or an LHRH agonist in combination with another modality (n = 1477; 37%). Calculated time on ADT was defined as the earliest start date, to the latest end date or the end of the study, with the assumption made that therapy was continuous.
Patients were grouped into risk categories based on preoperative disease characteristics (ie, PSA, clinical stage, and biopsy Gleason score) using a modification of the stratification system described by D'Amico et al,12 with the difference being that clinical stage T2c was included in the moderate–risk category in CaPSURE. Clinical staging was performed using the 1997 American Joint Committee on Cancer-International Union Against Cancer (AJCC–UICC) system. Low-risk patients were defined as having a PSA level ≤10 ng/mL, stage T1 or T2a disease, and a Gleason sum ≤6. High–risk patients had a PSA level >20 ng/mL, stage ≥T3 disease, a Gleason sum ≥8, or a primary Gleason grade ≥4. The moderate–risk category included all other patients. Concomitant medical comorbidity was estimated and analyzed by simply adding up the number of medical conditions recorded in the database for each individual patient. For the purposes of our analysis, tumor volume was macroscopically estimated by dividing the number of core needle biopsies positive for CaP by the total number of core needle biopsies taken (expressed as a percentage). The median percentage of core needle biopsies involved in the CaPSURE database (ie, 33% or one-third of core needle biopsies involved) was used to dichotomize patients, with 1432 (40%) men considered to have low–volume disease (<33%) and 2127 (60%) men considered to have high–volume disease (≥33%). PSA nadir after the initiation of ADT was defined as the first lowest value that preceded 3 consecutive PSA measurements without a decrease. Because there had to be 4 PSA measures after the start of ADT to designate a nadir, this was only able to be assessed in a subset of the original analysis dataset (n = 1780). PSA velocity was calculated using the least squares regression on at least 2 measures during ADT, and was able to be assessed in 1224 patients.
The primary endpoint of the study was the development of bone metastasis as indicated by positive bone scan (n = 84), irradiation to bone (n = 10), or progression to bone indicated on either the Clinic Visit Form (n = 84) or the Clinical TNM Staging Form (n = 13). We performed a survival analysis using time ADT was initiated to metastasis or censoring to estimate the hazards ratio of the development of metastasis for various clinical or demographic explanatory variables. Other treatments combined with ADT (radical prostatectomy vs radiotherapy vs no treatment), age (≥65 years vs <65 years), clinical risk category at diagnosis (low vs moderate vs high), PSA velocity on ADT, percentage of biopsies positive (<33% vs ≥33%), body mass index (BMI) (<25 vs ≥25), race (Caucasian vs African American vs other), and number of comorbidities were compared between patients who developed metastases and those who did not using chi-square tests in a Cox proportional hazards regression model. We used backward stepwise selection to build the best parsimonious model, with 5% significance required for inclusion. We tested key assumptions of the proportional hazards model by measuring the effects of time dependent covariates. Metastasis-free survival was plotted using Kaplan-Meier curves for patients without and with risk factors found to be independent predictors of metastasis. The log-rank statistic was used to determine the differences between the curves. All analyses were performed using SAS statistical software (version 9.1; SAS Institute Inc, Cary, NC).
The median follow-up in our cohort was 3.6 years (range, 1 month-15 years), with 197,401 person-months of follow-up available for analysis. The median length of time receiving ADT was 123 days (4 months), with a range of 1 to 174 months. The median time from diagnosis to the initiation of ADT was 42 days (range, 0-167 months). There were 191 events for an incidence rate of 1 per 1073 person-months. The median time from primary treatment to the development of metastatic disease was 53 months (range, 0-172 months). Of those patients who developed metastasis, 83 (43%) were still receiving ADT; those who were taken off treatment developed metastases at a median of 14 months after discontinuing ADT.
In univariate analysis, race (P = .0415), risk category (P < .0001), percentage of biopsies positive (P < .0001), age at diagnosis (P = .0008), and PSA velocity (P < .0001) were found to be significantly associated with risk of bone metastasis (Table 2). Patients classified in the race category of “other” had a greater risk of bone metastasis compared with African American patients, who in turn had a greater risk than Caucasian patients. An increase in 1 risk category held a near 2-fold increase in risk of bone metastasis compared with patients in the lower risk group. Furthermore, men with ≥33% of the biopsy cores involved and those aged <65 years are more likely to progress to metastatic disease. In the subset analysis, months to PSA nadir after ADT was found to be significantly inversely associated with progression to metastasis (hazards ratio [HR], 0.97; 95% confidence interval [95% CI], 0.95-0.98 [P < .0001]).
|Variable||No.||Univariate HR (95% CI)||P|
|Other treatment RT||4003||0.87 (0.64-1.19)||.380|
|Other treatment RP||4003||1.31 (0.94-1.84)||.116|
|Risk category (high vs intermediate vs low)||3875||2.37 (1.86-3.01)||<.0001|
|Percent positive biopsy at diagnosis (≥33% vs <33%)||3559||6.77 (3.83-11.94)||<.0001|
|No. of comorbidities (range, 0-11)||3069||1.04 (0.94-1.15)||.435|
|BMI (≥25 vs <25)||2897||1.38 (0.95-2.00)||.090|
|Race (other vs African American vs Caucasian)||4003||1.31 (1.01-1.71)||.042|
|Age at diagnosis, y (<65 vs ≥65)||4003||1.68 (1.24-2.27)||.001|
|PSA velocity||1224||1.04 (1.03-1.06)||<.0001|
|Mo to PSA nadir on ADT||1780||0.97 (0.95-0.98)||<.0001|
In the multivariate analysis, the best parsimonious model indicated that clinical risk group (P < .0001), percentage of biopsies positive (P < .0001), age at diagnosis (P = .028), and PSA velocity (P < .0001) were independently associated with risk of bone metastasis after ADT (Table 3). Dropped out of the multivariate Cox proportional hazards regression using a .05 significance level (data not shown) were number of comorbidities, race, BMI, and other treatments combined with ADT. After controlling for risk, and percentage of biopsies positive, primary therapy (radical prostatectomy vs radiotherapy) in conjunction with ADT was found to have no effect on risk of bone metastasis.
|Variable||Multivariate HR (95% CI)||P|
|Risk category (high vs intermediate vs low)||2.57 (1.60-4.15)||<.0001|
|Percent positive biopsy at diagnosis (≥33% vs <33%)||3.36 (1.53-7.38)||.003|
|Age at diagnosis, y (<65 vs ≥65)||2.11 (1.36-3.28)||.001|
|PSA velocity||1.04 (1.02-1.06)||<.0001|
A Kaplan-Meier survival plot depicting 5-year metastasis-free survival in high-risk and low-risk patients is shown in Figure 1. The high–risk group was comprised of 284 participants with 3 of the characteristics found to be independently statistically significantly associated with progression to metastasis: age <65 years, high–risk category, and ≥33% of the biopsies positive. The remaining 3719 patients were considered low-risk patients. The 5-year metastasis-free survival rate for the high-risk group was 84% (95% CI, 75%-89%) and was 95% for the low-risk group (95% CI, 94%-96%). The log-rank test found a statistically significant difference between the curves (P < .0001).
The CaPSURE database provided a rare opportunity to identify clinical predictors of metastasis in patients diagnosed with CaP who were receiving ADT. This large database included community practices and academic institutions, as well as VA hospitals, leading to a widely varied study population and a greater external validity for our study results.
Many previous studies have found PSA nadir4, 6, 7, 9 and PSA doubling time4, 8 to be predictors of PSA failure and survival in patients receiving ADT. One study examining the natural history of patients with a rising PSA despite ADT found that a PSA level >10 ng/mL and PSA velocity predicted shorter time to first bone metastasis.13 Similarly, the results of the current study indicated that PSA velocity and time to PSA nadir were associated with the development of bone metastasis. In addition, we found clinical parameters present at the time of original CaP diagnosis to be predictors of metastasis: age <65 years, high–risk category, and ≥33% of the biopsies positive. Although the association between risk category and tumor volume was intuitive, it was not as clear why younger patients would be more likely to progress than their older counterparts. Lengths of follow-up in the 2 groups were not significantly different, eliminating this as a potential explanation for this finding. One large epidemiologic study found that younger patients with CaP were more likely to have aggressive disease, with an advanced stage at diagnosis as well as poorly differentiated tumors.14 However, other studies have failed to demonstrate this link between younger age at diagnosis and worse outcome.15
We attempted to include PSA doubling as a variable in our analysis because this had been shown by others to be correlated with outcome in patients receiving ADT.4, 7-9 However, this value was only able to be obtained in a small subset of our original cohort (ie, 1441 patients). In addition, although we were unable to demonstrate a correlation between PSA doubling time and metastasis, the small numbers in the current study precluded us from drawing any strong conclusions based on this result. We were, however, able to include PSA velocity as a measure of PSA kinetics in the analysis, and, as expected, this was found to correlate strongly with prognosis in our patients.
The current study thus identifies clinical and pathologic characteristics present at the time of the initiation of ADT that may help stratify patients by risk. Knowledge of these clinical predictors also may help tailor follow-up for these men. Furthermore, the identification of patients at higher risk of prostate cancer progression may help identify potential candidates for clinical trials assessing bisphosphonates or chemotherapeutic agents. However, the findings of the current study need to be validated before the clinical usefulness of these predictors can be determined.
In contrast to early studies examining patients receiving ADT, the current study is more contemporary, allowing us to assess the clinical course of patients receiving this form of treatment in the PSA era. The use of PSA in the detection of early disease recurrence has led to a significant change in this patient population. Even in the group of patients with all 3 predictors of bone metastasis, only 12% of patients developed clinical disease recurrence at 5 years, emphasizing the indolent nature of most CaP and/or the effectiveness of ADT in delaying disease progression. Similarly, a recent systematic review of patients receiving ADT who were treated primarily with surgery, radiotherapy, or hormones alone found significantly higher 5-year disease-free and overall survival rates when compared with previously published rates.16
To have an adequate enough sample to conduct a meaningful analysis, we included all forms of ADT in the current study. The effect of surgical castration on the development of metastases is unlikely to be the same as the effect of antiandrogens or finasteride; however, the latter 2 comprised only a small fraction of the population (<6%). Nearly 90% of patients were treated with an LHRH agonist, either alone or in combination with another modality (ie, antiandrogen or 5 α-reductase inhibitor). Furthermore, due to the limitations of the database, we were unable to discriminate between those who received continuous as opposed to intermittent ADT.
Of interest, in our study sample, patients who underwent local treatment with surgery or radiotherapy had outcomes similar to those treated primarily with ADT with respect to the development of bone metastasis. The independent effect of treatment on the development of bone metastasis was not found to be significant. This likely reflects the poor prognostic features present in a large proportion of the patients in the current study. In fact, nearly half of the patients in the current study were classified as high risk by the modified criteria of D'Amico et al.12 It also suggests that clinical disease progression in patients with prostate cancer depends more on disease factors (ie, clinical risk category and volume of disease) than the primary treatment modality used. It remains to be seen whether the effect of primary treatment will be found to be significant with longer follow-up.
The limitations of the current study include those common to all retrospective analyses. A sampling bias may exist because the database only included information regarding patients who consented to the CaPSURE study. In addition, the subgroup of patients in which the PSA nadir could be calculated may have differed from the remaining patients. Finally, a limitation to the generalizeability of the current study findings is the low proportion of African Americans in the study sample (ie, 13%).
Age, volume of disease, risk category, and PSA velocity appear to be important predictors of clinical disease progression in patients with prostate cancer who are receiving ADT. This suggests that younger men with high–risk disease have a worse prognosis than older men with similar disease. The use of these prognostic variables may help identify candidates for clinical trials evaluating secondary treatments for patients with castrate-resistant disease.
Conflict of Interest Disclosures
The authors made no disclosures.
- 1Studies on prostate cancer: the effects of castration, of estrogen and of androgen injection on serum phosphatases in metastatic carcinoma of the prostate. Cancer Res. 1941; 1: 293-297., .