Time to PSA nadir independently predicts overall survival in metastatic hormone sensitive prostate cancer patients treated with androgen deprivation therapy

Abstract

Introduction

The purpose of this study was to evaluate the relationship between the kinetics of PSA decline after androgen deprivation therapy (ADT) initiation and overall survival (OS) in men with metastatic hormone-sensitive prostate cancer (HSPC).

Methods

We identified a cohort of metastatic HSPC patients treated with androgen deprivation therapy (ADT) using our institutional database. Patients were included if they had at least 2 serum PSA determinations before nadir PSA and at least one serum PSA value available within 1 month of ADT initiation. Patient characteristics, PSA at ADT initiation, nadir PSA, time to PSA nadir (TTN) and PSA decline (PSAD) in relation to OS were analyzed.

Results

179 patients were identified, with a median follow-up after ADT initiation of 4.0 years. Median OS after ADT initiation was 7.0 years. Median PSA at ADT initiation and PSA nadir were 47 and 0.28 ng/mL, respectively. On univariate analysis: TTN <6 months, a PSAD >52 ng/mL/year, PSA nadir ≥ 0.2 ng/mL, a PSA≥47.2 ng/mL at ADT initiation and Gleason score >7, were associated with a shorter OS. On multivariate analysis, TTN<6 months, Gleason score >7 and a PSA nadir ≥ 0.2 ng/mL independently predicted a shorter OS.

Discussion

To our knowledge, this is the first report to show that a faster time to reach a PSA nadir post-ADT initiation is associated with shorter survival duration in men with metastatic HSPC. These results need confirmation, but may indicate that a rapid initial response to ADT indicates more aggressive disease.

Introduction

Androgen deprivation therapy (ADT) is the most effective systemic therapy for patients with hormone sensitive prostate cancer (HSPC). Most patients will experience a substantial decline in PSA, and PSA may remain low or undetectable for years. Nevertheless, the emergence of castration-resistant prostate cancer (CRPC) is typical.¹ Once CRPC develops, median survival is approximately 24-36 months.

PSA kinetics have been established as useful prognostic indicators for survival in different clinical settings. For instance, a high pre-therapy PSA velocity (PSAV) is associated with a greater risk of death from prostate cancer despite radical prostatectomy ² or external beam radiation therapy (EBRT). ³ Similarly, PSA doubling time (PSADT) with biochemical recurrence after definitive local therapy is able to identify patients at high risk of prostate cancer specific mortality. ⁴ However, the influence of PSA kinetics post-ADT initiation on the outcome of patients with metastatic HSPC disease population is poorly characterized. A recent large study suggests that a PSA of 4 ng/mL or less after 7 months of ADT is a strong predictor of survival in this population. ⁵ We queried a longitudinal database of patients treated at Dana-Farber Cancer Institute (DFCI) to describe the outcome of ADT use in men with HSPC with metastases at the time of ADT initiation. Our interest was to determine which parameters of PSA kinetics predicted overall survival (OS).

Methods

Database

This cohort was generated from the Prostate Clinical Research Information System (CRIS) at Dana-Farber Cancer Institute (DFCI). The CRIS system consists of data entry software, a central data repository, collection of patient data including comprehensive follow-up of all patients, and tightly integrated security measures. This system has recently been described. ⁶ Data are stored in an Oracle relational database (Oracle, Inc., Redwood Shores, CA). The current version of Prostate CRIS became available for prospective data entry in November 2001. All patients seen at DFCI and Brigham and Women’s Hospital (BWH) with a diagnosis of prostate cancer are approached to participate. The consent rate is >85%. Patients are in general followed every 3 months with PSA tests and radiological exams if needed.

Patient selection

We identified and included patients with prostate cancer who had been treated with ADT [orchiectomy or luteinizing hormone releasing hormone (LHRH) agonists, with or without an anti-androgen] for metastatic HSPC. Patients had been consented to provide information on the CRIS protocol, and had blood collected for research purposes. Patients were eligible if they had metastastic disease at ADT start; there were at least one PSA measure within 1 month prior to (84.4% of all patients) or post-ADT initiation (15.6% of all patients) and at least two PSA values before nadir PSA (including nadir PSA). Patients were excluded if there were only two PSA values and the second one was > one year from the first measure, or PSA increased after ADT initiation.

Statistical considerations

Patient and disease characteristics were summarized as number (%) of patients or median and range of values. Time to PSA nadir (TTN) was defined as the duration of time from ADT initiation to the date the lowest PSA value was first observed after ADT. PSA decline (PSAD) was calculated from the slope of the linear regression of the raw PSA values over time, using PSA values from 1 month prior to ADT until the nadir PSA.

The primary outcome variable was overall survival (OS), defined as the time from ADT initiation to the date of death, censored at the date of last known alive. The distribution of OS was estimated using the Kaplan-Meier product-limit method; median and OS along with 95% confidence intervals (CIs) were summarized. OS was compared according to patients’ characteristics using log-rank tests. Continuous variables were dichotomized at the median value within the cohort, with the exception of PSA nadir, which was dichotomized at 0.2 ng/mL because of its known correlation with prostate cancer-specific survival. As PSA nadir, TTN and PSA decline were measures that developed over time after ADT initiation, we also did secondary analyses of OS from the landmarks of PSA nadir on ADT (post-nadir OS) and 1 year after ADT initiation. Multivariate modeling was undertaken to arrive at a parsimonious model to predict OS; hazard ratios (HRs), 95% CIs, and Wald chi-square test p-values were reported from the models. The statistical analysis was undertaken using SAS version 9 (SAS Institute Inc., Cary, NC) and p<0.05 (two-sided) was considered as statistically significant.

Results

Patient Characteristics and PSA Kinetics

One-hundred and seventy nine patients met the eligibility criteria for this analysis (Table 1). Median age of the cohort at ADT initiation was 64 years (42-84), respectively. A Gleason score greater than 7 was present in 47% of patients. Thirty-two percent of patients had metastatic (M1) disease at diagnosis.

Table 1.

Patient and disease characteristics of the metastatic HSPC cohort at ADT initiation (N=179)

	N	%	Median	Interquartile range (IQR)
At diagnosis
Clinical stage ≥ T3	10	5.6
Clinical M1 stage	58	32.4
Clinical N1 stage	25	14.0
Biopsy Gleason >7	84	46.9
Definitive local therapy (RP or RT)	85	47.5
Hormonal therapy for local disease	21	11.7
At ADT initiation
Age, years			64	57-70
Years from diagnosis to ADT			0.3	0.1-4.9
PSA at ADT initiation, ng/mL			47.2	12.3-167
Orchiectomy +/− LHRH analog	11	6.1
Intermittent LHRH	5	2.9
LHRH+Anti-androgen	144	80.4
During ADT
Nadir PSA during ADT, ng/mL			0.28	0.01-1.66
Time to PSA nadir, months			6.0	3.1-9.8
PSAD after ADT, ng/mL/year			52	15, 271

ADT: androgen deprivation therapy

LHRH: luteinizing hormone-releasing hormone

PSAD: PSA decline

Median time from diagnosis to initiation of ADT was 0.3 year (interquartile range (IQR): 0.1-4.9 years). Eighty percent of patients received an anti-androgen as part of ADT. Median PSA at ADT initiation and PSA nadir were 47 and 0.28 ng/mL, respectively; 45% of patients reached PSA nadir <0.2 ng/mL. The median TTN was 6 months and the median PSAD post ADT initiation was 52 ng/mL/year. PSA at the start of ADT was similar between the groups with short (< 6 months) and prolonged (≥ 6 months) TTN (median 31 versus 55.6 ng/mL, p=0.21). However, the group with rapid PSA decline (>52ng/mL/year) had much higher baseline PSA compared to those with slow decline (median 146 versus 12.5 ng/mL, p<0.0001). Patients who had rapid PSAD or shorter TTN were also less likely to achieve a nadir of PSA lower than 0.2 ng/mL (P<0.05).

Univariate and multivariate analyses of predictors for OS

With a median follow-up of 4 years (range 0.3~13.6 years), 63 (35%) patients have died. Median OS after ADT initiation was 7.0 years (95% CI: 4.9 ~ 9.4 years; range 0.3 ⁺ ~13.6 ⁺ years). Table 2 summarizes survival estimates at one-year intervals after ADT initiation.

Table 2.

Kaplan-Meier estimates of overall survival at yearly intervals after ADT initiation.

Years after ADT	Overall survival rate	95% CI
1	96%	92%-98%
2	90%	84%-94%
3	84%	77%-89%
4	71%	62%-78%
5	59%	49%-68%

ADT: Androgen Deprivation Therapy

Several factors were found to predict OS after ADT initiation on univariate analysis (Table 3). A Gleason score of greater than 7, higher PSA at ADT initiation, higher PSA nadir, shorter TTN, and a more rapid PSAD were all associated with a shorter OS. Figures 1a and 1b demonstrated overall survival curves by TTN (<6 months vs. ≥ 6 months, log rank p-value=0.01) and PSAD (< 52 ng/mL/year vs. ≥ 52 ng/mL/year, logrank p-value=0.0012), respectively. Secondary landmark analyses comparing post-nadir survival among patient groups defined by these variables, or comparing survival for those who had been alive for at least 1 year after ADT initiation, were consistent with our initial results. Noted that prior locally therapy (RT or RP) was associated with OS on univariate (p=0.009) but not on multivariate analysis (p=0.98)

Table 3.

Univariate analysis of predictors for overall survival (OS), from time of ADT initiation, and from alternative landmark timepoints.

		OS after ADT (N=179)		OS in patients who were alive > 1 year after ADT (N=159)		Post-nadir OS (N=179)
	N	Median (years)	P-value	Median (years)	P-value	Median (years)	P-value
All patients	179	7.0
Biopsy Gleason
≤7		7.0	0.005
>7		5.2
Unknown		3.8
PSA at ADT initiation, ng/mL
<47.2		7.8	0.036
≥47.2		5.2
PSA nadir, ng/mL
< 0.2		10.7	<0.0001	10.7	<0.0001	9.6	<0.0001
≥ 0.2		4.7		4.8		4.0
TTN, Months
<6		4.5	0.010	4.6	0.020	4.3	0.075
≥6		7.8		7.8		7.3
PSAD, ng/mL/yr
> 52		4.9	0.001	4.9	0.001	4.3	0.002
≤ 52		7.8		7.8		7.9

Figure 1.

Overall Survival from time of ADT initiation by PSAD (figure 1a) and by TTN (figure 1b)

In multivariate analysis, statistically significant predictors for overall survival were PSA nadir, TTN and Gleason score. Higher PSA nadir (≥ 0.2 ng/mL), shorter TTN (<6 months) and higher Gleason (>7) were associated with shorter survival (Table 4: model 1). We further examined whether the association of TTN with OS was modified by the levels of PSA nadir (Table 4: model 2). The results showed that the relationship between prolonged TTN and an improved survival became more apparent in patients with a nadir ≥0.2 ng/mL (HR=0.31, 95% CI: 0.16-0.62). However, in the group of patients who had a nadir < 0.2 ng/mL, TTN had no influence on survival (HR=1.0, 95% CI: 0.42-2.40; interaction effect, p=0.04). Kaplan Meier plot of survival by PSA nadir and TTN (figure 2) illustrated that the two survival distributions (TTN≥6 versus TTN<6 months) crossed and there was a lot of early censoring among patients with nadir < 0.2 ng/mL, which suggested that the association between survival and TTN remained unclear in patients who achieved a very low nadir in this cohort. The results were consistent when variables were analyzed on a continuous fashion rather than dichotomized (data not shown).

Table 4.

Multivariate models predicting overall survival

	Hazard Ratio	95% CI	p-value
Model 1: main effects only
Biopsy Gleason (>7 vs. ≤7)	1.92	1.10~3.37	0.022
Biopsy Gleason (Unknown vs. ≤7)	2.72	1.19~6.21	0.018
PSA nadir, ng/mL (≥0.2 vs.<0.2)	3.85	2.12~6.99	<.0001
TTN, month (≥6 vs. < 6)	0.48	0.28~0.80	0.005
Model 2: investigating interaction
Biopsy Gleason (>7 vs. ≤7)	1.83	1.05~3.21	0.034
Biopsy Gleason (Unknown vs. ≤7)	2.42	1.06~ 5.54	0.036
Nadir <0.2 ng/mL
TTN < 6 mos	1.00(reference)
TTN ≥ 6 mos	1.00	0.42~2.40	0.041*
Nadir ≥0.2 ng/mL
TTN < 6 mos	1.00(reference)
TTN ≥ 6 mos	0.31	0.16~0.62

^*Wald chi-square test for the interaction between nadir PSA and TTN

Figure 2.

Overall Survival by nadir PSA and TTN

Discussion

In 2007, the American Society of Clinical Oncology (ASCO) published clinical practice guidelines on the initial hormonal management of androgen-sensitive, metastatic prostate cancer. ⁷ ADT remains the mainstay of treatment for metastatic HSPC. PSA kinetics prior to definitive curative-intent therapy and during biochemical recurrence when these therapies fail are established prognostic factors associated with outcome ^{2, 3, 8}

Petrylak et al showed that PSA level declines of at least 20%–40% in addition to a PSA velocity at 3 months after beginning treatment with docetaxel chemotherapy, are associated with overall survival. ⁹ An analysis from the TAX327 trial ¹⁰ demonstrated that a PSA decline of ≥ 30% within 3 months of chemotherapy initiation had the highest degree of surrogacy for overall survival, confirming data from the Southwest Oncology Group (SWOG) 9916 trial.

Using data from a randomized phase III trial, Hussain et al recently reported that a PSA value of 4 ng/mL or less after 7 months of ADT is a strong independent predictor of improved survival in new metastatic HSPC ⁵ These results may allow the identification of patients who are unlikely to benefit with standard ADT long before developing castration-resistant disease, thus providing an alternative to investigate newer therapies. Unfortunately, the influence of PSA kinetics such as PSAD, TTN and others as predictors of survival in patients with metastatic but HSPC remain largely unknown. Most studies evaluate “PSA responses” ^{5, 11, 12} as surrogate for survival. Overall, efficacy of ADT has generally correlated with more profound PSA declines that reflect subsequently improved survival. In our dataset, a lower PSA nadir, as expected, was independently associated with longer survival. We attempted to look additionally at other parameters of PSA kinetics to refine the predictive information in this context.

We looked at several parameters of PSA kinetics after ADT initiation in an attempt to evaluate their relationship with survival. Interestingly, a more rapid PSAD and a shorter TTN (a cut-off of 6 months was used for convenience since this was the median value) were associated with a shorter survival, and TTN was independently associated with survival even after adjusting for several other variables. One would assume that a rapid decrease in PSA and reaching a nadir “faster” would indicate more cancer cell death that would subsequently translate into a higher survival, similar to the data regarding docetaxel chemotherapy in hormone refractory disease. ⁹ One might hypothesize that perhaps this is an ADT-specific phenomenon only; the rapid fall in PSA may not be related to cell viability and may simply be a transcriptional effect from ADT on PSA production rather than cell death. Another explanation could be due to the fact that we may be selecting a subset of aggressive prostate cancer cells that quickly become resistant to ADT or are not as reliant on the androgen receptor for viability. In fact, significant clinical heterogeneity may exist in metastatic HSPC as supported by Hussain et al data that showed a striking variability in survival exists: patients had median survival differences ranging from 13 to 75 months depending upon the absolute PSA nadir achieved.

In this study, the association between TTN and survival was limited only to patients with a PSA nadir above 0.2 ng/mL. We also attempted to look at the length of time from ADT initiation to the first documentation of PSA ≤ 0.2 ng/mL, since some investigators equate a PSA ≤ 0.2 ng/mL to an undetectable PSA, especially with less sensitive PSA tests assays. By this alternative definition, the results were not different for patients who have not achieved a nadir ≤0.2 ng/mL. However, there was no association between TTN and survival for those who were able to achieve a nadir ≤0.2 ng/mL after ADT. The relationship between TTN and survival deserves further investigation in patients with a lower PSA nadir.

Rapid PSAD predicted a worse survival on the univariate analysis, but did not retain its importance in the multi-variable models when we included Gleason, PSA nadir and TTN. We also observed a strong association between PSAD and PSA at the start of ADT. Although the variability in PSA was relatively large (Range: 0~15090 ng/mL; IQR: 12.3-167 ng/mL) at ADT initiation, nearly 75% of patients in this cohort could achieve a PSA of 3 ng/mL or less within 6 months of ADT. Therefore, the rate of PSAD was primarily determined by the PSA levels at ADT initiation.

To our knowledge, our finding has not been reported previously in studies of ADT. This report is the first to use PSA kinetics (such as TTN) outcome measure to evaluate ADT efficacy in metastatic HSPC. This analysis has some potential pitfalls. First, it is retrospective. The measurements of PSA kinetics will depend on the number and the pattern of PSA testing post-ADT initiation. The quality of the data however should be high since we performed a quality control and found the numbers of available PSA values at 6 months of ADT were similar between the groups with shorter and prolonged TTN. Second, while all patients had metastatic HSPC, this analysis also draws on patients with heterogeneous clinical backgrounds not controlled for in the context of a prospective clinical trial. Some important factors such as lactate dehydrogenase, ECOG performance status and hemoglobin were not included in this analysis because they are not routinely measured for patients treated with ADT in our institution. Finally, we did not study patients never treated with ADT or those who were treated with anti-androgens alone, but we believe that these patients represent very rare case scenarios.

To summarize, we found that PSA kinetics at the initiation of ADT can predict overall survival in metastatic HSPC patients, with faster TTN and more rapid PSAD being associated with a shorter overall survival. We postulate that this may be a consequence of underlying biologic heterogeneity differences in patients’ tumors. Further prospective data and external validation in independent data sets are needed to confirm these finding.