For men receiving androgen-suppression therapy (AST) for a rising postoperative or postradiation prostate-specific antigen (PSA) recurrence, whether the time to an undetectable (u) PSA was significantly associated with prostate cancer-specific mortality (PCSM) was evaluated.
The study cohort comprised 585 men with a rising PSA and negative bone scan after surgery (n = 415) or radiation therapy (n = 170) that were treated with AST and achieved a uPSA. Gray's regression was used to evaluate whether the time to a uPSA after AST was significantly associated with the time to PCSM after the uPSA adjusting for known prognostic factors.
The median time (interquartile range) to achieve a uPSA was 4.6 (range, 2.8–7.8) months. There were 23 deaths, 4 of which were from prostate cancer. An increasing time to a uPSA (adjusted hazard ratio [HR]: 9.2, 95% confidence interval [CI]: 3.8, 22.1; P < .0001), a decreasing PSA doubling time (DT) (HR: 0.58, 95% CI: 0.43, 0.80; P = .0007), and Gleason score 8 to 10 cancers (HR: 8.6, 95% CI: 1.04, 77; P = .05) were significantly associated with a shorter time to PCSM.
Androgen suppression therapy (AST) is often administered after radical prostatectomy (RP) or radiation therapy (RT) for prostate-specific antigen (PSA) recurrence.1 In that setting prior studies2, 3 have shown that men who achieve an undetectable PSA (uPSA) within 8 months after AST have a significantly longer time to prostate cancer-specific mortality (PCSM) as compared with those whose PSA remains detectable. In addition, data from the Southwest Oncology Group (SWOG)4 disclosed a similar finding in men with newly diagnosed hormone-naive and metastatic (M1) prostate cancer. As a result, men achieving a uPSA level may not be considered for randomized studies that are investigating the use of AST plus immediate vs delayed chemotherapy. Yet despite the achievement of a uPSA after AST administration for a rising postoperative or post-RT PSA, some of these men will develop bone metastases and experience PCSM within 5 years.2, 3 Therefore, improved measures to identify men at risk for PCSM despite the achievement of a uPSA after AST administration are needed.
To address this issue, prior investigators5–8 have evaluated whether the time to PSA nadir is associated with the time to a detectable PSA in men undergoing AST as primary treatment for localized prostate cancer or treatment at the time of PSA recurrence after definitive local therapy or in men with hormone-naive M1 prostate cancer. Whereas all 4 prior retrospective studies showed a significant association between time to PSA nadir and the time to a subsequent PSA rise, the results were conflicting. Two studies5, 6 found that a shorter time to nadir was associated with a longer time to PSA recurrence, whereas 2 other studies7, 8 found the opposite result.
One explanation for this inconsistency is the choice of the endpoint of time to PSA recurrence as compared with the endpoint of PCSM. Attempts at estimating the time to a PSA recurrence when the sampling intervals of PSA while undergoing AST were not prospectively defined can lead to incorrect estimates of the time to PSA recurrence, subjecting the analyses to error. In addition, these prior studies did not specifically analyze the subgroup of men who achieved a uPSA while undergoing AST for post-RP or post-RT PSA recurrence where additional information regarding the association with time to PCSM is needed. To overcome these 2 potential issues in the current study, for men receiving AST for a rising postoperative or postradiation PSA we evaluated whether the time to the achievement of a uPSA (ie, PSA <0.2 ng/mL) was significantly associated with the time to PCSM after the achievement of a uPSA adjusting for known prognostic factors.
MATERIALS AND METHODS
Study Cohort, Staging, and Treatment
Between January 1, 1988, and January 1, 2002, 585 men who underwent AST for post-RP (n = 415) or post-70 Gray RT (n = 170) PSA recurrence and achieved a uPSA constituted the study cohort for this retrospective analysis. These men were patients who presented with clinically localized prostate cancer, derived from 44 institutions taken from 2 multiinstitutional databases, the Center for Prostate Disease Research (CPDR),9 and the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE).10 No man received neoadjuvant, concurrent, or adjuvant hormonal therapy that underwent RT. Each man provided written informed consent before study entry; the Institutional Review Board approved the study at each of the 44 participating institutions within CPDR and CaPSURE. The median age of the men at the time of the uPSA after AST administration was 69 years (interquartile range [IQR]: 64 to 74). Table 1 shows the clinical characteristics of all men in the study cohort. Before the initiation of AST, which occurred in most men (83%) before reaching a PSA level of 10 ng/mL, PSA failure was documented using the consensus definition from the PSA working group11 and then restaging with a bone scan and computerized tomography or magnetic resonance imaging was obtained. Patients with radiographic evidence of regional or distant metastatic disease were excluded, as were men who received postoperative adjuvant RT or AST or salvage RT or who had a persistent PSA postoperatively. Treatment with AST was continuous and consisted of a Luteinizing Hormone Releasing Hormone (LHRH) agonist in 357 (61%), orchiectomy in 23 (4%), antiandrogen monotherapy in 18 (3%), and combined hormonal blockade in 187 (32%) men. The selection of the type of AST was at the discretion of the treating physician.
Table 1. Clinical Characteristics of the 585 Men Who Received Androgen Suppression Therapy Following Postoperative or Postradiation Prostate-Specific Antigen Recurrence and Achieved an Undetectable PSA Level
Percentages may not sum to 100% due to rounding. uPSA indicates undetectable prostate-specific antigen; AST, androgen suppression therapy; DT, doubling time; RP, radical prostatectomy; RT, radiation therapy; IQR, interquartile range.
Time to uPSA, mo
Median (IQR): 4.6 [3.6, 7.8]
Pre-AST PSA DT, mo
Median (IQR): 5.6 [2.8, 9.3]
3 to 5.99
Median (IQR): 6 [6,7]
8 to 10
PSA level at initiation of AST, ng/mL
Median (IQR): 1.9 [0.6, 7.4]
>4 to 10
>10 to 20
Interval to post-RP or post-RT PSA failure, y
Median (IQR): 1.7 [0.8, 3.1]
Age at time of uPSA, y
Median (IQR): 69 [64, 74]
60 to 69
70 to 79
External beam radiation therapy
Follow-up and Determination of the Cause of Death
Before the documentation of PSA failure,11 men were followed generally every 3 months for 2 years, then every 6 months until 5 years posttreatment, then annually thereafter. The median follow-up (IQR) after achieving a uPSA was 2.3 (0.8–4.2) years and follow-up started on the day of the uPSA; no patient was lost to follow-up. After the administration of AST, PSA levels were monitored at a median (IQR) of 3.9 (2.7–6.0) months with a minimum and maximum of 2.0 and 6.9 months, respectively. Determination of the cause of death was made from death certificates in all cases. To record a death as being due to prostate cancer, there had to be documented hormone-refractory metastatic prostate cancer and evidence that the PSA level was rising at the time of the last follow-up visit before death. Overall, there were 23 deaths, 4 of which were from prostate cancer. Of the 19 remaining deaths, 14 were attributable to vascular events (eg, myocardial infarction), 4 to second cancers, and 1 to Alzheimer disease.
A proportional hazards model that accounted for the subdistributions of competing causes of mortality (Gray's formulation)12 was used to evaluate whether the time to uPSA (<0.2 ng/mL) after AST for postoperative or postradiation PSA recurrence was associated with the time to PCSM after the uPSA after adjusting for known prognostic factors. Examination of the Schoenfeld residuals and prostate cancer-specific mortality failure times for each covariate were performed in order to check the subdistribution proportional hazards assumption in Gray's regression for modeling the prostate cancer-specific mortality failure times. No evidence was found against Gray's regression model. The covariates examined included the continuous variables of time to the achievement of a uPSA, PSA doubling time (DT) before the initiation of AST, PSA level at the initiation of AST, age at the time of the achievement of the uPSA, interval to post-RP or post-RT PSA failure, and the categorical variables of biopsy or prostatectomy Gleason score,13 clinical or prostatectomy American Joint Commission on Cancer (AJCC) tumor category,14 and initial local therapy. Time zero was defined as the day of the achievement of a uPSA. The PSA level at the time of initiation of AST was log-transformed so that the values followed a normal distribution.
For all categorical variables the cutpoints selected were made before examining the data based on established strata.15 These cutpoints were Gleason 8 to 10 vs 7 or less (baseline), and RT vs RP (baseline). Adjusted and unadjusted hazard ratios (HRs)12 for PCSM with associated 95% confidence intervals (CI) and P-values were calculated for each covariate examined. For the purpose of illustration, estimates of time to PCSM and all cause mortality (ACM) after the achievement of a uPSA stratified about the median time to a uPSA were calculated and graphically displayed. A cumulative incidence16 and 1 − Kaplan Meier (KM)17 methodology were used to calculate these estimates and the K-sample18 and the log-rank test19 were used to compare the distribution of these estimates over time, respectively. A 2-sided P < .05 was considered statistically significant. R v. 2.1.1 (R Foundation for Statistical Computing, Vienna, Austria) was used for all calculations pertaining to the cumulative incidence probability estimates of time to PCSM. SAS v. 9.1.3 (SAS Institute, Cary, NC) was used for all remaining statistical analyses.
Factors Associated With Prostate-Cancer Specific Mortality
The median time (IQR) to achieve a uPSA was 4.6 (2.8–7.8) months. As shown in Table 2, after adjusting for known prognostic factors an increasing time to a uPSA (adjusted HR: 9.2, 95% CI: 3.8, 22.1; P < .0001), a decreasing PSA DT (HR: 0.58, 95% CI: 0.43, 0.80; P = .0007), and Gleason score 8 to 10 cancers (HR: 8.6, 95% CI: 1.04, 77; P = .05) were significantly associated with a shorter time to PCSM, whereas age at the time of the uPSA (P = .21), the PSA level when AST was initiated (P = .54), the time interval to PSA failure after primary therapy (P = .53), and initial local therapy (P = .54) were not. The HR of 9.2 for the covariate of time to a uPSA in years is equivalent to an HR of 1.2 in months or a 20% increase in risk in PCSM for each additional month that a patient needs to reach a uPSA level after the initiation of AST.
Table 2. Summary of the Baseline Prognostic Factor Unadjusted and Adjusted Hazard Ratios for the Time to Prostate Cancer-Specific Mortality Following the Achievement of an Undetectable PSA
For the purpose of illustration, the cumulative incidence estimates of PCSM and 1 − KM estimates of overall survival stratified by the median time to the achievement of a uPSA were calculated and graphically displayed for men. Some men with a short PSA DT had Gleason score 7 cancers; as a result, these men were also included in the example. Specifically, for the 160 men (27% of the study cohort) with a PSA DT <6 months and Gleason 7 to 10 prostate cancer there were significantly higher cumulative incidence estimates of PCSM (P = .024) and 1 − KM estimates of survival (P = .047) or ACM if the time to a uPSA after AST administration was greater than the median (4.6 months) as compared with the median or less as shown in Figures 1 and 2, respectively. At 5 years after the achievement of a uPSA, estimates of PCSM (95% CI) were 12% (0.4, 24) and 0% (0, 0) for men whose time to a uPSA after AST administration was greater than the median as compared with the median or less, respectively. These respective estimates were 23% (10, 46) and 7% (2, 20) for the endpoint of ACM. In these 160 men after a median follow-up (IQR) of 3.3 (1.1–4.5) years, there were 11 deaths, 4 of which were from prostate cancer.
The main finding in the current study was that a longer time to the achievement of a uPSA in addition to a shorter pre-AST PSA DT and high-grade prostate cancer were significantly associated with a shorter time to PCSM after the achievement of a uPSA. Specifically, as shown in Figures 1 and 2, there were significantly higher estimates of both PCSM and ACM in men whose time to a uPSA exceeded the median (4.6 months) as compared with all others. This group of men comprised 27% (160/585) of all study patients and their PCSM and ACM estimates were 12% and 23% as compared with 0% and 7%, respectively, meaning that death from prostate cancer within 5 years of achieving the uPSA was estimated to comprise more than half of all deaths. Therefore, these men are also at high risk of PCSM and should be considered for enrollment into the immediate vs delayed chemotherapy trials despite achieving a uPSA after AST administration. Given the survival benefit observed from 2 randomized studies20, 21 when docetaxel was used in men with hormone-refractory and metastatic prostate cancer, docetaxel would be a leading choice for study in these men.
Several issues require further discussion. First, the PSA assay used in this study had a lower limit of detection of 0.2 ng/mL. Whether a shortened time to the achievement of a uPSA would remain significantly associated with a longer time to PCSM after a uPSA in men who achieve a uPSA level if an ultrasensitive PSA assay had been used remains to be studied. Second, the measurement of the PSA levels after AST were obtained at a median of 3.9 months with an IQR of 2.7 to 6.0 months. Whereas the sampling interval and IQR are relatively short and narrow, respectively, given that there was a distribution of sampling times and not a fixed sampling time, some uncertainty could have been introduced into the estimation of the time to a uPSA.. Third, of 23 total deaths, 4 men died of prostate cancer, which was sufficient, as noted in Figures 1 and 2, to show significant differences in PCSM and ACM when evaluating men with a short PSA DT and high-grade cancer who experienced times above or below the median to achieve a uPSA after AST administration. However, longer follow-up yielding more PCSM events will be needed to assess the strength of the association observed in the current study and to more tightly define the 95% confidence limits around the HR of 9.2 for the covariate of time to a uPSA. Finally, the type of AST varied in this study and consisted of an LHRH agonist monotherapy in 61% and combined hormonal blockade in 187 (32%) men. If a future randomized study reveals that the combination of an LHRH agonist and an antiandrogen are superior to LHRH agonist monotherapy with respect to PCSM in men with a post-RP or post-RT PSA recurrence, then this study would need to be reanalyzed adjusting for the type of AST administered as a covariate in the proportional hazards model.12
Despite these potential limitations, in men who achieve a uPSA after AST administration for post-RP or post-RT PSA recurrence, the risk of PCSM increases significantly as the time to the achievement of a uPSA lengthens especially in men with a short pre-AST PSA DT and high-grade prostate cancer. Therefore, these men should be considered for randomized studies evaluating immediate vs delayed chemotherapy after the achievement of the uPSA.