`
`JOURNAL OF CLINICAL ONCOLOGY
`
`O R I G I N A L R E P O R T
`
`Susan Halabi, Andrew J. Armstrong,
`
`Ellen Kaplan, Chen-Yen Lin, and Nicole
`
`C. Solomon, Duke University, Durham,
`
`NC; Oliver Sartor, Tulane University,
`
`New Orleans, LA; Johann de Bono,
`
`Royal Marsden Hospital, Sutton, United
`
`Kingdom; and Eric J. Small, University
`
`of California at San Francisco, San
`
`Francisco, CA.
`
`Published online ahead of print at
`
`www.jco.org on October 7, 2013.
`
`Supported in part by National Institutes
`
`of Health Grant No. CA 155296-1A1
`
`and by sanofi-aventis.
`
`Terms in blue are defined in the glos-
`
`sary, found at the end of this article
`
`and online at www.jco.org.
`
`Presented orally at the 48th Annual
`
`Meeting of the American Society of
`
`Clinical Oncology, Chicago, IL, June
`
`1-5, 2012.
`
`The study sponsor did not have a role in
`
`the analysis, interpretation of data, writing
`
`of the manuscript, or decision to submit
`
`the manuscript for publication.
`
`Authors’ disclosures of potential con-
`
`flicts of interest and author contribu-
`
`tions are found at the end of this
`
`article.
`
`Corresponding author: Susan Halabi,
`
`PhD, Duke University Medical Center,
`
`2424 Erwin Rd, Suite 11088, Durham,
`
`NC 27710; e-mail: susan.halabi@
`
`duke.edu.
`
`© 2013 by American Society of Clinical
`
`Oncology
`
`0732-183X/13/3131w-3944w/$20.00
`
`DOI: 10.1200/JCO.2013.50.3201
`
`Prostate-Specific Antigen Changes As Surrogate for Overall
`Survival in Men With Metastatic Castration-Resistant
`Prostate Cancer Treated With Second-Line Chemotherapy
`Susan Halabi, Andrew J. Armstrong, Oliver Sartor, Johann de Bono, Ellen Kaplan, Chen-Yen Lin,
`Nicole C. Solomon, and Eric J. Small
`
`A
`
`B
`
`S
`
`T
`
`R
`
`A
`
`C
`
`T
`
`Purpose
`Prostate-specific antigen (PSA) kinetics, and more specifically a ⱖ 30% decline in PSA within 3
`months after initiation of first-line chemotherapy with docetaxel, are associated with improvement
`in overall survival (OS) in men with metastatic castration-resistant prostate cancer (mCRPC). The
`objective of this analysis was to evaluate post-treatment PSA kinetics as surrogates for OS in
`patients receiving second-line chemotherapy.
`
`Patients and Methods
`Data from a phase III trial of patients with mCRPC randomly assigned to cabazitaxel plus
`prednisone (C ⫹ P) or mitoxantrone plus prednisone were used. PSA decline (ⱖ 30% and ⱖ 50%),
`velocity, and rise within the first 3 months of treatment were evaluated as surrogates for OS. The
`Prentice criteria, proportion of treatment explained (PTE), and meta-analytic approaches were used
`as measures of surrogacy.
`
`Results
`The observed hazard ratio (HR) for death for patients treated with C ⫹ P was 0.66 (95% CI, 0.55
`to 0.79; P ⬍ .001). Furthermore, a ⱖ 30% decline in PSA was a statistically significant predictor
`of OS (HR for death, 0.52; 95% CI, 0.43 to 0.64; P ⬍ .001). Adjusting for treatment effect, the HR
`for a ⱖ 30% PSA decline was 0.50 (95% CI, 0.40 to 0.62; P ⬍ .001), but treatment remained
`statistically significant, thus failing the third Prentice criterion. The PTE for a ⱖ 30% decline in PSA
`was 0.34 (95% CI, 0.11 to 0.56), indicating a lack of surrogacy for OS. The values of R2 were ⬍ 1,
`suggesting that PSA decline was not surrogate for OS.
`
`Conclusion
`Surrogacy for any PSA-based end point could not be demonstrated in this analysis. Thus, the
`benefits of cabazitaxel in mediating a survival benefit are not fully captured by early PSA changes.
`
`J Clin Oncol 31:3944-3950. © 2013 by American Society of Clinical Oncology
`
`INTRODUCTION
`
`Investigators have long been challenged by the lack
`of surrogate end points for clinical trials in men
`with metastatic castration-resistant prostate cancer
`(mCRPC).1-5 True surrogacy requires meeting sev-
`eral rigorous statistical criteria defined by Prentice
`(Prentice criteria).6 The degree of surrogacy can also
`be measured by the proportion of treatment effect
`explained (PTE).7 Reductions in serum prostate-
`specific antigen (PSA) with systemic therapy may
`reflect reductions in tumor burden, which may be
`linked to improved long-term outcomes; this has
`been a natural intermediate end point to assess sur-
`rogacy. Kelly et al1 first proposed the use of post-
`therapy changes in PSA from baseline as an
`intermediate marker of response in patients with
`
`mCRPC. Numerous subsequent reports confirmed
`that patients with mCRPC who had experienced
`ⱖ 50% decline in PSA from baseline had improved
`survival, compared with those patients who did not
`achieve ⱖ 50% reduction in PSA.2-4 In retrospective
`studies, several investigators have reported that PSA
`decline ⱖ 50% correlated with improved survival.
`Not all investigators have correlated PSA decline
`from baseline with improved survival.8,9 However,
`Petrylak et al10,11 demonstrated that both ⱖ 30%
`and ⱖ 50% decline in PSA satisfied the Prentice
`criteria in patients with mCRPC treated with first-
`line chemotherapy, whereas ⱖ 50% decline in PSA
`failed to meet the surrogacy criteria as measured by
`PTE. By contrast, Armstrong et al12 found that al-
`though ⱖ 30% decline in PSA after docetaxel treat-
`ment in the phase III TAX327 trial fulfilled the
`
`3944
`
`© 2013 by American Society of Clinical Oncology
`
`Amerigen Exhibit 1118
`Amerigen v. Janssen IPR2016-00286
`
`
`
`PSA Changes As Surrogate for OS
`
`Prentice criteria, the degree of surrogacy as measured by PTE in this
`decline was modest.13 Evidence to support PSA decline as a surro-
`gate for overall survival (OS) across multiple agent classes and
`mechanisms of action is lacking.
`The primary objective of this analysis was to evaluate whether
`ⱖ 30% decline in PSA within 3 months of treatment initiation was a
`surrogate end point of OS in patients with mCRPC receiving second-
`line chemotherapy (cabazitaxel or mitoxantrone) after progression
`with docetaxel. A secondary objective was to assess whether ⱖ 50%
`decline in PSA was a surrogate end point for OS. In addition, we
`performed exploratory analysis of other PSA kinetics as surrogate end
`points for OS.
`
`PATIENTS AND METHODS
`
`Patients
`This analysis used data from the TROPIC trial, a phase III trial of 755
`men with mCPRC previously treated with a docetaxel-containing regimen.14
`Participants were randomly assigned to receive either 12 mg/m2 mitoxantrone
`intravenously over 15 to 30 minutes plus oral prednisone 10 mg daily (M ⫹ P)
`or cabazitaxel 25 mg/m2 administered over 1 hour every 3 weeks in combina-
`tion with prednisone (C ⫹ P). Eligible patients had progressive mCRPC after
`treatment with a docetaxel-based regimen, Eastern Cooperative Oncology
`Group performance status of 0 to 2, and adequate hematologic, hepatic, renal,
`and cardiac function. Those who received mitoxantrone, radiotherapy, or
`other cancer therapies within 4 weeks before enrollment were excluded. De-
`tails of eligibility have been previously reported.14
`
`End Points
`The primary end point of the clinical trial was OS, which was defined as
`the time from random assignment to date of death resulting from any cause.
`Secondary end points were ⱖ 50% decline in PSA using the Prostate Cancer
`Working Group 2 criteria.15 Serum PSA was measured at baseline and then
`every 3 weeks until progression, with a PSA response per protocol defined as
`ⱖ 50% decline from baseline PSA, if baseline PSA was ⬎ 0.2 ng/mL and was
`maintained for at least 3 weeks.
`For the purposes of this analysis, the surrogate end point to be examined
`was ⱖ 30% decline in PSA. The rationale for using this end point as a binary
`was to confirm the findings reported in patients with mCRPC after docetaxel
`treatment.11,12 Similar to previous studies, ⱖ 30% was defined as a decline
`ⱖ 30% from the baseline PSA measurement at any time within the first 90 days
`of treatment.11,12 In a case in which confirmation was required, there had to be
`a second consecutive decline at least 21 days after the first decline. Ranges of
`PSA decline/rise and velocity were explored as markers of OS. PSA velocity was
`calculated as the slope of log PSA (log 2 scale) by time based on the least squares
`method using at least two postbaseline PSA measurements. PSA rise was
`computed as percent increase from the baseline PSA measurement. An indi-
`cator variable was created if the percent value was ⱖ the percent specified in
`the analysis.
`
`Data Analysis
`As part of a research federal grant, this analysis was approved by the Duke
`University Institutional Review Board. We used the logistic regression model
`to test whether treatment arm predicted ⱖ 30% and ⱖ 50% decline in PSA and
`employed three different approaches to evaluate whether PSA decline or PSA
`rise was a surrogate end point for OS. These approaches were: one, Prentice
`criteria; two, PTE; and three, meta-analysis. The Prentice criteria define a
`surrogate as a “response variable for which a test of the null hypothesis of no
`relationship to the treatment groups under comparison is also a valid test of the
`corresponding null hypothesis based on the true end point.”6(p432) To test the
`Prentice criteria, we fit a series of three proportional hazards models of OS with
`the following covariates: model one, included treatment arm; model two,
`included PSA decline (or rise) as a surrogate marker; and model three, in-
`cluded both treatment arm and PSA decline (or rise). To fulfill the Prentice
`
`criteria, a marker is considered a surrogate end point if it is statistically signif-
`icantly associated (P ⬍ .05) with OS in both univariate models. However, in
`the multivariable model, the marker but not treatment arm needs to be
`statistically significant. The Schoenfeld test was used to check for the propor-
`tional hazards assumption, and there was no evidence that this was violated in
`these three models.16
`PTE is obtained from two proportional hazards models and is computed
`as 1 minus the ratio of the estimated regression coefficient for treatment effect
`in model three (adjusted) over estimated regression coefficient for treatment
`arm in model one (unadjusted).7 A value of 1 for the PTE indicates a perfect
`surrogate end point, whereas a value of 0 represents no surrogacy. The 95% CI
`for PTE was computed using a nonparametric bootstrapped procedure to
`estimate the variance-covariance matrix of the estimated regression coeffi-
`cients for unadjusted and adjusted treatment effects (models one and three).7
`Following Burzykowski et al17 and Busye et al,18 we considered a meta-
`analytic approach to assess the surrogacy of PSA decline for OS. The meta-
`analysis procedure allows one to evaluate the surrogacy from individual and
`trial levels. The trial level assessed the overall prediction power of the surrogate
`end point on the true end point, whereas the individual level evaluated the
`strength of the dependency between surrogate end point and true end point
`after adjusting for the treatment effect. A surrogate end point is considered
`valid if it presents a high degree (closer to one) of association at both the trial
`and individual levels.
`The data in this report were from a single trial, and to implement the
`meta-analysis framework, we randomly partitioned the TROPIC data into five
`clusters and assumed that each cluster was obtained from an independent trial.
`The number five was chosen to ensure the number of patients in each treat-
`ment group was ⱖ 50 within each fold. Because the partitioning was per-
`formed randomly, the procedure was repeated 500 times. The global odds ratio
`(OR) and R2 were averaged over 500 replicates.
`The Kaplan-Meier estimator was used to estimate the OS distributions
`by patients who experienced and did not experience ⱖ 30% and ⱖ 50%
`decline in PSA. R software (R Foundation for Statistical Computing, Vienna,
`Austria) was used for the data analyses, and all statistical tests were two sided.
`
`RESULTS
`
`Baseline Characteristics
`Of the 755 patients enrolled onto the TROPIC trial, 17 patients
`did not have PSA data at baseline, and 85 patients had PSA ⬍ 0.20
`ng/mL and were excluded from the analysis. The current analysis
`was based on 653 patients (86%) who had sufficient PSA data
`post-treatment. Participants in this analysis had similar baseline
`characteristics compared with patients who did not have PSA
`decline data. Moreover, the survival distributions were not differ-
`ent between patients who were and were not included in the
`analysis (log-rank P ⫽ .852).
`Baseline clinical and laboratory characteristics of the 653 patients
`are summarized in Table 1. A majority were white, with a median age
`of 67 years; 91% had Eastern Cooperative Oncology Group perfor-
`mance status of 0 to 1; 54% had measurable disease. Median PSA was
`170 ng/mL (interquartile range [IQR], 68 to 465). There were no
`differences between the two arms with respect to baseline variables
`(Table 1).
`
`PSA Decline
`Median PSA decline in each arm was 31.1% (IQR, 0 to 61.4) and
`0% (IQR, 0 to 31.2) for C ⫹ P and M ⫹ P, respectively. Two hundred
`fifty men (38%) experienced ⱖ 30% decline in PSA from baseline
`(51% with C ⫹ P; 26% with M ⫹ P), whereas 25% of patients had
`ⱖ 50% decline in PSA (33% with C⫹ P; 26% with M ⫹ P). Treatment
`
`www.jco.org
`
`© 2013 by American Society of Clinical Oncology
`
`3945
`
`
`
`Halabi et al
`
`Table 1. Baseline Characteristics of Patients With PSA Decline Data by
`Treatment Arm
`
`Table 2. OS by Treatment Arm and Percent Decline in PSA Within 3 Months
`of Treatment
`
`Characteristic
`
`M ⫹ P
`(n ⫽ 325)
`
`C ⫹ P
`(n ⫽ 328)
`
`Total
`(N ⫽ 653)
`
`Age, years
`
`Median
`
`25th and 75th percentile
`
`Race, %
`
`White
`
`Asian
`
`67
`
`62-73
`
`81
`
`9
`
`68
`
`62.75-73
`
`86
`
`6
`
`67
`
`62-73
`
`84
`
`8
`
`M ⫹ P
`
`C ⫹ P
`
`Decline in
`PSA (%)
`
`No. of
`Patients
`
`Median
`OS
`(months)
`
`No. of
`Patients
`
`Median
`OS
`(months)
`
`ⱖ 0
`
`ⱖ 5
`
`ⱖ 10
`
`ⱖ 20
`
`154
`
`143
`
`129
`
`108
`
`15.1
`
`14.8
`
`14.8
`
`15.2
`
`242
`
`237
`
`221
`
`192
`
`16.3
`
`16.3
`
`16.5
`
`16.7
`
`Total
`
`Median
`OS
`(months)
`
`15.6
`
`15.5
`
`15.5
`
`15.9
`
`Black
`
`Other
`
`ECOG PS, %
`
`0
`
`1
`
`2
`
`6
`
`3
`
`32
`
`59
`
`10
`
`5
`
`3
`
`39
`
`54
`
`7
`
`6
`
`3
`
`35
`
`56
`
`8
`
`ⱖ 25
`
`ⱖ 30
`
`ⱖ 40
`
`ⱖ 50
`
`ⱖ 60
`
`ⱖ 70
`
`94
`
`83
`
`67
`
`55
`
`43
`
`25
`
`15.2
`
`15.4
`
`15.2
`
`15.2
`
`15.1
`
`14.5
`
`178
`
`167
`
`143
`
`108
`
`84
`
`59
`
`17.2
`
`17.2
`
`18.0
`
`19.7
`
`20.5
`
`22.6
`
`16.1
`
`16.2
`
`16.6
`
`16.9
`
`17.8
`
`17.2
`
`ⱖ 80
`
`ⱖ 90
`
`14
`
`7
`
`15.1
`
`10.6
`
`44
`
`21
`
`NA
`
`22.6
`
`22.6
`
`22.6
`
`Abbreviations: C ⫹ P, cabazitaxel plus prednisone; M ⫹ P, mitoxantrone plus
`prednisone; OS, overall survival; PSA, prostate-specific antigen.
`
`Disease extent, %
`
`Metastatic
`
`Bone
`
`Visceral
`
`Locoregional
`
`PSA, ng/mL
`
`Median
`
`95
`
`89
`
`25
`
`4
`
`97
`
`83
`
`23
`
`3
`
`96
`
`86
`
`24
`
`4
`
`169.5
`
`169.5
`
`169.5
`
`25th and 75th percentile
`
`68.4-479.2
`
`68.6-449.5
`
`68.4-465.0
`
`Alkaline phosphatase, U/L
`
`Median
`
`153.5
`
`149.5
`
`150.5
`
`25th and 75th percentile
`
`94.2-312.0
`
`82.2-288.5
`
`89.0-300.0
`
`Hemoglobin, g/L
`
`Median
`
`120
`
`119
`
`120
`
`25th and 75th percentile
`
`109.0-130.0
`
`109.4-129.2
`
`109.0-130.0
`
`54
`
`Measurable disease, %
`
`Baseline pain, %
`
`Prior hormonal therapy, %
`
`Hormonal
`
`Irradiation
`
`Surgery
`
`Biologic
`
`No. of chemotherapy lines, %
`
`1
`
`2
`
`ⱖ 2
`
`No. of docetaxel regimens, %
`
`1
`
`2
`
`ⱖ 2
`
`56
`
`46
`
`99
`
`62
`
`54
`
`10
`
`70
`
`22
`
`8
`
`86
`
`12
`
`2
`
`53
`
`49
`
`99
`
`65
`
`52
`
`8
`
`67
`
`26
`
`7
`
`84
`
`14
`
`2
`
`48
`
`99
`
`64
`
`53
`
`9
`
`69
`
`24
`
`8
`
`85
`
`13
`
`2
`
`Abbreviation: C ⫹ P, cabazitaxel plus prednisone; ECOG PS, Eastern Coop-
`erative Oncology Group performance status; M ⫹ P, mitoxantrone plus
`prednisone; PSA, prostate-specific antigen.
`
`arm significantly predicted ⱖ 30% decline in PSA for patients receiv-
`ing C ⫹ P (OR, 3.02; 95% CI, 2.17 to 4.21; P ⬍ .001) compared with
`patients receiving M ⫹ P.
`There were 449 deaths observed among 653 patients, and median
`follow-up time among 204 surviving patients was 16.4 months (95%
`CI, 14.8 to 18.5). Median OS by PSA decline by arm is listed in Table 2.
`
`Test for Surrogacy
`The Prentice criteria. Prentice operational criteria were applied.6
`First, treatment arm was a statistically significant predictor of OS (Fig
`1A). The observed hazard ratio (HR) for death for patients treated
`with C ⫹ P was 0.66 (95% CI, 0.55 to 0.79; P ⬍ .001) compared with
`
`patients treated with M ⫹ P. The observed median survival times were
`15.0 (95% CI, 14.0 to 16.3) and 12.7 months (95% CI, 11.2 to 13.6) for
`C ⫹ P and M ⫹ P, respectively. Second, ⱖ 30% decline in PSA was a
`statistically significant predictor of OS, with an HR for death of 0.52
`(95% CI, 0.43 to 0.64; P ⬍ .001) among patients who experienced
`ⱖ 30% PSA decline compared with those who did not (Fig 1B). Third,
`in a multivariable model with ⱖ 30% PSA decline and treatment arm,
`both PSA decline and treatment arm remained statistically significant.
`The adjusted HR for treatment arm was 0.76 (95% CI, 0.62 to 0.92;
`P ⬍ .005). Because of this, the third Prentice criterion was not met.
`In addition, ⱖ 50% decline in PSA was also tested for surrogacy
`of OS. Following the same steps described in the previous paragraph,
`treatment arm significantly predicted ⱖ 50% decline in PSA with
`patients treated with C ⫹ P having an OR of 2.41 (95% CI, 1.66 to 3.49;
`P ⬍ .001) compared with patients treated with M ⫹ P. PSA decline
`ⱖ 50% from baseline was also a statistically significant predictor of OS
`(HR for death, 0.56; 95% CI, 0.44 to 0.71; P ⬍ .001) among patients
`who experienced ⱖ 50% PSA decline compared with those who did
`not (Fig 2). The observed median survival times were 15.0 (95% CI,
`14.0 to 16.3) and 12.7 months (95% CI, 11.3 to 13.6) for C ⫹ P and
`M ⫹ P, respectively. Similar to the analysis for ⱖ 30% PSA decline in
`PSA, after adjusting for ⱖ 50% PSA decline, treatment arm remained
`a statistically significant predictor of survival. The adjusted HR for
`death for patients treated with C ⫹ P was 0.71 (95% CI, 0.59 to 0.86;
`P ⫽ .005) compared with patients treated with M ⫹ P. Thus, ⱖ 50%
`decline in PSA also failed to meet the third Prentice criterion.
`PSA decline as a continuous surrogate end point was also ex-
`plored as a potential surrogate of OS. In multivariable analysis, the
`adjusted HR for death decline for patients treated with C ⫹ P was 0.78
`(95% CI, 0.64 to 0.95; P ⫽ .01) compared with patients treated with
`M ⫹ P. Thus, PSA decline as a continuous outcome did not meet the
`third criterion of Prentice.
`PTE. As a measure of degree of surrogacy within this trial, the
`PTE analysis of the 0% to 90% decline in PSA within 3 months after
`treatment was undertaken. PTE for ⱖ 30% decline in PSA was 0.34
`(95% CI, 0.11 to 0.56), whereas PTE for ⱖ 50% decline in PSA was
`
`3946
`
`© 2013 by American Society of Clinical Oncology
`
`JOURNAL OF CLINICAL ONCOLOGY
`
`
`
`PSA Changes As Surrogate for OS
`
`< 30% decline
`≥ 30% decline, P < .001
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`(probability)
`
`Overall Survival
`
`B
`
`Mitoxantrone + prednisone
`Cabazitaxel + prednisone, P < .001
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`(probability)
`
`Overall Survival
`
`A
`
`0
`
`6
`
`12
`
`18
`
`24
`
`30
`
`0
`
`6
`
`12
`
`18
`
`24
`
`30
`
`Time (months)
`
`Time (months)
`
`No. at risk
`Mitoxantrone
` + prednisone
`Cabazitaxel
` + prednisone
`
`325
`
`328
`
`262
`
`286
`
`161
`
`205
`
`56
`
`79
`
`8
`
`26
`
`1
`
`3
`
`No. at risk
`< 30% decline
`≥ 30% decline
`
`403
`250
`
`317
`231
`
`193
`173
`
`68
`67
`
`8
`26
`
`0
`4
`
`Fig 1. (A) Treatment arm predicting for overall survival. (B) Greater than or equal to 30% decline in prostate-specific antigen predicting overall survival.
`
`ⱖ 50% decline in PSA were 0.30 (95% CI, 0.27 to 0.32; Fig 3B) and
`0.27 (95% CI, 0.25 to 0.3; Fig 3D), respectively. The association
`between PSA decline as a continuous surrogate and OS is shown in
`Appendix Figure A2 (online only). R2s were 0.62 (95% CI, 0.61 to
`0.62) and 0.50 (95% CI, 0.47 to 0.52) at the individual and trial
`levels, respectively. The values of R2 were ⬍ 1, suggesting that PSA
`decline is not a surrogate for OS.
`
`DISCUSSION
`
`In this analysis, surrogacy for any PSA-based end point could not be
`demonstrated using either the Prentice criteria or PTE. In addition, an
`analysis based on split sample in random subgroups did not demon-
`strate trial-level surrogacy. Although surrogacy for some PSA-based
`end points has been met in patients with mCRPC receiving primary
`docetaxel-based chemotherapy, surrogacy does not seem to be main-
`tained for second-line chemotherapy used in the postdocetaxel set-
`ting. To our knowledge, this is the first analysis of these surrogate end
`points in patients receiving second-line chemotherapy.
`OS remains the gold-standard end point in phase III trials of
`mCRPC, and although retrospective analyses have demonstrated
`some modest degree of surrogacy for PSA decline, these intermediate
`end points have not been prospectively validated.11,12,19 The need for
`surrogate markers of OS will only increase as more agents are ap-
`proved for the treatment of patients with mCRPC.20-23 Fortunately,
`these agents lead to prolonged survival for patients; however, their
`open-label use will push survival farther out. Moreover, their use
`subsequent to clinical-trial treatment may reduce the hypothesized
`effect size of the therapies being evaluated.20-23 The dilution will re-
`quire larger trial sizes, longer follow-up periods, or larger effect sizes if
`OS is to be used as the primary end point, all of which make the OS end
`point less desirable.
`It has become increasingly common to use ⱖ 30% decline in PSA
`as an end point for all patients with mCRPC based on the original
`first-line data. It is with these concerns in mind that we evaluated the
`
`0.20 (95% CI, 0.05 to 0.35). The lower bound of the 95% CI did not
`exceed 0.50, suggesting a lack of surrogacy (Appendix Fig A1A, online
`only). A similar analysis using a confirmatory PSA value after a decline
`of either ⱖ 30% or ⱖ 50% failed to provide evidence of surrogacy
`for survival.
`PSA rise. Exploratory analyses of a 0% to 90% rise in PSA within
`3 months after treatment were performed (Appendix Fig A1B, online
`only). The lower bounds of the 95% CI were ⬍ 0.50, implying a lack of
`surrogacy for PSA rise. The results were similar when confirmation of
`PSA rise was required, with the lower bound not meeting the PTE
`requirement for surrogacy (data not shown).
`Meta-analytic approach. Associations between ⱖ 30% and
`ⱖ 50% decline in PSA and OS are presented in Figure 3. At the
`individual level, global ORs for ⱖ 30% (Fig 3A) and ⱖ 50% decline
`in PSA were 2.46 (95% CI, 2.45 to 2.47) and 2.08 (95% CI, 2.07 to
`2.09), respectively (Fig 3C). At the trial level, R2s for ⱖ 30% and
`
`< 50% decline
`≥ 50% decline, P < .001
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`(probability)
`
`Overall Survival
`
`0
`
`6
`
`12
`
`18
`
`24
`
`30
`
`Time (months)
`
`No. at risk
`< 50% decline 490
`≥ 50% decline 163
`
`399
`149
`
`256
`110
`
`88
`47
`
`15
`19
`
`0
`4
`
`Fig 2. Greater than or equal to 50% decline in prostate-specific antigen
`predicting overall survival.
`
`www.jco.org
`
`© 2013 by American Society of Clinical Oncology
`
`3947
`
`
`
`Halabi et al
`
`B
`
`200
`
`150
`
`100
`
`50
`
`Frequency
`
`2.3
`
`2.4
`
`2.5
`
`2.6
`
`2.7
`
`0
`
`0
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`1.0
`
`Odds Ratio
`
`R 2
`
`D
`
`200
`
`150
`
`100
`
`50
`
`Frequency
`
`1.9
`
`2.0
`
`2.1
`
`2.2
`
`0
`
`0
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`1.0
`
`Odds Ratio
`
`R 2
`
`200
`
`150
`
`100
`
`50
`
`0
`
`200
`
`150
`
`100
`
`50
`
`0
`
`A
`
`Frequency
`
`C
`
`Frequency
`
`Fig 3. (A, C) Individual and (B, D) trial-level effects for (A, B) ⱖ 30% and (C, D) ⱖ 50% decline in prostate-specific antigen. Dashed line indicates empirical mean.
`
`utility of PSA kinetics as surrogates for OS in patients with mCRPC in
`a trial involving second-line chemotherapy. Our analysis impugns the
`utility of PSA or PSA kinetics as surrogates for OS in patients with
`mCRPC receiving second-line chemotherapy. Furthermore, the data
`demonstrate that there are different disease states within the group of
`patients with mCRPC. There are a number of potential explanations
`for why PSA kinetics may have some utility as a surrogate in patients
`treated with first-line chemotherapy, but not those treated with
`second-line therapy. First, the benefit of cabazitaxel in improving OS
`may not be mediated through PSA-dependent mechanisms.24 PSA
`may decline for reasons unassociated, or not linearly associated, with
`cell killing by second chemotherapy treatment.25 Stated alternatively,
`patients with mCRPC previously treated with first-line docetaxel may
`have selected prostate cancer cells with a greater degree of dissociation
`between PSA decline and cancer-cell killing. Although there are no
`data to support this possibility, it is also possible that cabazitaxel, as
`opposed to docetaxel, has a narrower spectrum of activity while still
`resulting in PSA declines.
`Although the results of these data suggest that measures of PSA
`are not appropriate as surrogate markers of clinical benefit in this
`setting, it should be recognized that the only reason it was even possi-
`ble to evaluate PSA kinetics using the Prentice criteria for surrogacy in
`this setting was because cabazitaxel prolonged OS in men with
`mCRPC.6 However, the Prentice criteria do not determine trial-level
`surrogacy. Several authors have used different approaches to the vali-
`
`dation of surrogate end points, such as individual-level surrogacy
`based on individual patient data.26-30 Thus, an intermediate end point
`of OS may be of great clinical trial utility, even if it does not meet the
`Prentice criteria. The successful identification of a surrogate, such as
`progression-free survival, for OS would have wide-ranging implica-
`tions for the design, conduct, and analysis of trials in this population.
`There are many strengths to the analysis reported in this article.
`First, post-therapy changes in PSA have been considered as both
`binary and continuous outcomes. Second, different analytic ap-
`proaches were undertaken where both individual-level and trial-level
`surrogacy associations were assessed. Although the Prentice criteria
`establish only individual-level surrogacy after adjusting for treatment,
`the meta-analytic approaches consider association at both individual
`and trial levels. Finally, because the data are from a trial in which
`patients were treated with cabazitaxel, an innovative approach was
`implemented where the data were randomly divided into five clusters,
`and each cluster was assumed to come from an independent trial.
`Although data splitting is a useful tool, it cannot substitute for a true
`meta-analysis. As new drugs and new paradigms are introduced, ad-
`ditional validation will be warranted.
`In summary, based on this extensive analysis, there is no evidence
`that PSA kinetics are appropriate markers of clinical benefit, and as
`such, they cannot be used as surrogates for OS in patients with
`mCRPC receiving second-line chemotherapy after progression with
`
`3948
`
`© 2013 by American Society of Clinical Oncology
`
`JOURNAL OF CLINICAL ONCOLOGY
`
`
`
`PSA Changes As Surrogate for OS
`
`docetaxel. Decisions to stop treatment should not be guided by short-
`term, isolated changes in PSA measurements, and the identification
`and validation of surrogate end points for OS for the approximately
`29,720 men31 who will die as a result of this disease in 2013 remain
`unmet needs in mCRPC trials.
`
`AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS
`OF INTEREST
`
`Although all authors completed the disclosure declaration, the following
`author(s) and/or an author’s immediate family member(s) indicated a
`financial or other interest that is relevant to the subject matter under
`consideration in this article. Certain relationships marked with a “U” are
`those for which no compensation was received; those relationships marked
`with a “C” were compensated. For a detailed description of the disclosure
`categories, or for more information about ASCO’s conflict of interest policy,
`please refer to the Author Disclosure Declaration and the Disclosures of
`Potential Conflicts of Interest section in Information for Contributors.
`Employment or Leadership Position: None Consultant or Advisory
`Role: Andrew J. Armstrong, sanofi-aventis (C), Medivation (C); Johann
`
`de Bono, sanofi-aventis (C), AstraZeneca (C), Johnson & Johnson (C)
`Stock Ownership: None Honoraria: Andrew J. Armstrong,
`sanofi-aventis; Johann de Bono, Astellas Pharma, Johnson & Johnson,
`Medivation Research Funding: Susan Halabi, sanofi-aventis; Andrew J.
`Armstrong, Medivation, Janssen Pharmaceuticals, sanofi-aventis; Johann
`de Bono, AstraZeneca, sanofi-aventis, Genentech; Ellen Kaplan,
`sanofi-aventis; Nicole C. Solomon, sanofi-aventis Expert Testimony:
`None Patents: None Other Remuneration: None
`
`AUTHOR CONTRIBUTIONS
`
`Conception and design: Susan Halabi, Andrew J. Armstrong, Eric Small
`Provision of study materials or patients: Oliver Sartor, Johann de Bono
`Collection and assembly of data: Johann de Bono
`Data analysis and interpretation: Susan Halabi, Andrew J. Armstrong,
`Oliver Sartor, Ellen Kaplan, Chen-Yen Lin, Nicole C. Solomon,
`Eric J. Small
`Manuscript writing: All authors
`Final approval of manuscript: All authors
`
`REFERENCES
`
`1. Kelly WM, Scher HI, Mazumdar M, et al:
`Prostate-specific antigen as a measure of disease
`outcome in metastatic hormone-refractory prostate
`cancer. J Clin Oncol 11:607-615, 1993
`2. Smith DC, Dunn RL, Strawderman MS, et al:
`Change in serum prostate-specific antigen as a
`marker of
`response to cytotoxic therapy for
`hormone-refractory prostate cancer. J Clin Oncol
`16:1835-1843, 1998
`3. Scher HI, Kelly WM, Zhang ZF, et al: Post-
`therapy serum prostate-specific antigen level and
`survival in patients with androgen-independent pros-
`tate cancer. J Natl Cancer Inst 91:244-251, 1999
`4. Small EJ, Halabi S, Ratain MJ, et al: Random-
`ized study of three different doses of suramin ad-
`ministered with a fixed dosing schedule in patients
`with advanced prostate cancer: Results of Inter-
`group 0159, Cancer and Leukemia Group B 9480.
`J Clin Oncol 20:3369-3375, 2002
`5. Halabi S, Vogelzang NJ, Ou SS, et al:
`Progression-free survival as a predictor of overall
`survival
`in men with castrate resistant prostate
`cancer. J Clin Oncol 27:2766-2771, 2009
`6. Prentice RL: Surrogate endpoints in clinical
`trials: Definition and operational criteria. Stat Med
`8:431-440, 1989
`7. Lin DY, Fleming TR, De Gruttola V: Estimating
`the proportion of treatment effect explained by a
`surrogate marker. Stat Med 16:1515-1527, 1997
`8. Kantoff PW, Higano CS, Shore ND, et al:
`Sipuleucel-T immunotherapy for castration-resistant
`prostate cancer. N Engl J Med 363:411-422, 2010
`9. Kantoff PW, Schuetz TJ, Blumenstein BA, et
`al: Overall survival analysis of a phase II randomized
`controlled trial of a poxviral-based PSA-targeted im-
`munotherapy in metastatic castration-resistant pros-
`tate cancer. J Clin Oncol 28:1099-1105, 2010
`10. Petrylak DP, Tangen CM, Hussain MH, et al:
`Docetaxel and estramustine compared with mitox-
`
`antrone and prednisone for advanced refractory
`prostate cancer. N Engl J Med 351:1513-1520, 2004
`11. Petrylak DP, Ankerst DP, Jiang CS, et al:
`Evaluation of prostate-specific antigen declines for
`surrogacy in patients treated on SWOG 99-16. J Natl
`Cancer Inst 98:516-521, 2006
`12. Armstrong AJ, Garrett-Mayer E, Ou Yang YC,
`et al: Analysis of prostate-specific antigen decline as
`a surrogate for overall survival
`in metastatic
`hormone-refractory prostate cancer
`(HRPC): A
`TAX327 analysis. J Clin Oncol 25:3965-3970, 2007
`13. Tannock IF, de Wit R, Berry WR, et al: Do-
`cetaxel plus prednisone or mitoxantrone plus pred-
`nisone for advanced prostate cancer. N Engl J Med
`351:1502-1512, 2004
`14. de Bono JS, Oudard S, Ozguroglu M, et al:
`Prednisone plus cabazitaxel or mitoxantrone for
`metastatic castration-resistant prostate cancer pro-
`gressing after docetaxel treatment: A randomised
`open-label trial. Lancet 376:1147-1154, 2010
`15. Scher HI, Halabi S, Tannock I, et al: Design
`and end points of clinical trials for patients with
`progressive prostate cancer and castrate levels of
`testosterone: Recommendations of the Prostate
`Cancer Clinical Trials Working Group. J Clin Oncol
`26:1148-1159, 2008
`16. Schoenfeld D: Chi-squared goodness of fit
`tests for proportional hazards regression model.
`Biometrika 67:145-153, 1980
`17. Burzykowski T, Molenberghs G, Buyse M:
`The validation of surrogate end points by using data
`from randomized clinical trials: A case-study in ad-
`vanced colorectal cancer. J R Stat Soc A 167:103-
`124, 2004
`18. Buyse M, Molenberghs G, Burzykowski T, et
`al: The validation of surrogate endpoints in meta-
`analyses of randomized experiments. Biostatistics
`1:49-67, 2000
`19. Collette L, Burzykowski T, Carroll K, et al: Is
`prostate-specific antigen a valid surrogate endpoint
`for survival in hormonally treated patients with met-
`astatic prostate cancer? J Clin Oncol 23:6139-6148,
`2005
`
`■ ■ ■
`
`20. de Bono JS, Logothetis CJ, Molina A, et al:
`Abiraterone and increased survival
`in metastatic
`prostate cancer. N Engl J Med 364:1995-2005, 2011
`21. Scher HI, Fizazi K, Saad F, et al: Increased
`survival with enzalutamide in prostate cancer after
`chemotherapy. N Engl J Med 367:1187-1197, 2012
`22. Ryan CJ, Smith MR, de Bono JS, et al: Abi-
`raterone in metastatic prostate cancer with no prior
`chemotherapy. N Engl J Med 368:138-148, 2013
`23. Parker C, Nilsson S, Heinrich D, et al: Alpha
`emitter radium-223 and survival in metastatic pros-
`tate cancer. N Engl J Med 369:213-23, 2013
`24. Fleming TR, DeMets DL: