Skip to main content
NCBI home page
Lippincott Open Access logoLink to Lippincott Open Access
. 2024 Mar 14;42(14):1687–1698. doi: 10.1200/JCO.23.02105

Efficacy of Poly(ADP-ribose) Polymerase Inhibitors by Individual Genes in Homologous Recombination Repair Gene-Mutated Metastatic Castration-Resistant Prostate Cancer: A US Food and Drug Administration Pooled Analysis

Jaleh Fallah 1,✉, Jianjin Xu 1, Chana Weinstock 1, Xin Gao 1, Brian L Heiss 1, William F Maguire 1, Elaine Chang 1, Sundeep Agrawal 1, Shenghui Tang 1, Laleh Amiri-Kordestani 1,2, Richard Pazdur 1,2, Paul G Kluetz 1,2, Daniel L Suzman 1
PMCID: PMC11095872  PMID: 38484203

Abstract

PURPOSE

We performed a pooled analysis of multiple trials of poly(ADP-ribose) polymerase inhibitors (PARPi) in metastatic castration-resistant prostate cancer (mCRPC) to investigate the efficacy of PARPi in each individual homologous recombination repair (HRR) mutated (m) gene.

PATIENTS AND METHODS

We pooled patient-level data from trials of PARPi in mCRPC that reported mutation status in individual HRR genes. Any HRR gene with available data across all the randomized trials of PARPi in first-line mCRPC was selected. The hazard ratios (HRs; 95% CI) for radiographic progression-free survival (rPFS; by blinded independent review) and overall survival (OS) of a PARPi plus an androgen receptor pathway inhibitor (ARPI) relative to placebo plus an ARPI in the pool of three randomized trials in first-line mCRPC were calculated using Kaplan-Meier estimates and a Cox proportional hazards model.

RESULTS

In ATMm (N = 268), rPFS HR was 1.05 (0.74 to 1.49) and OS HR was 1.18 (0.82 to 1.71). In BRCA1m (N = 64), rPFS HR was 0.51 (0.23 to 1.1) and OS HR was 0.74 (0.34 to 1.61). In BRCA2m (N = 422), rPFS HR was 0.31 (0.23 to 0.42) and OS HR was 0.66 (0.49 to 0.89). In CDK12m (N = 164), rPFS HR was 0.50 (0.32 to 0.80) and OS HR was 0.63 (0.39 to 0.99). In CHEK2m (N = 172), rPFS HR was 1.06 (0.67 to 1.66) and OS HR was 1.53 (0.95 to 2.46). In PALB2m (N = 41) rPFS HR was 0.52 (0.23 to 1.17) and OS HR was 0.78 (0.34 to 1.8).

CONCLUSION

In this pooled analysis, benefit from PARPi appeared greatest for patients with BRCA1m, BRCA2m, CDK12m, and PALB2m. Given limitations of this exploratory analysis, the apparent lack of benefit from PARPi in patients with CHEK2m or ATMm should be further explored in future clinical trials.

Efficacy of PARP inhibitors by individual genes in HRR gene-mutated mCRPC: A US FDA pooled analysis.

INTRODUCTION

Homologous recombination repair (HRR) is an essential pathway for DNA repair.1-3 Approximately 30% of patients with advanced prostate cancer have somatic or germline mutations (m) in genes directly or indirectly involved in the HRR pathway (HRRm).4-7

CONTEXT

  • Key Objective

  • Multiple genes are considered to be involved in the homologous recombination repair (HRR) pathway, and mutations in these genes may sensitize tumor cells to poly(ADP-ribose) polymerase inhibitors (PARPi). However, it is unclear if all HRR gene mutations result in similar efficacy from PARPi. This analysis evaluated the efficacy of PARPi in pooled subgroups with individual HRR gene mutations.

  • Knowledge Generated

  • HRR gene mutations are heterogeneous in terms of sensitivity to PARPi. This exploratory pooled analysis demonstrated notable difference in the magnitude of treatment effect from PARPi across mutated HRR genes.

  • Relevance (M.A. Carducci)

  • This pooled data analysis from the US Food and Drug Administration vantage point confirms that benefit from PARPi is not the same across DNA damage repair genes, and highlights certain mutations that are associated with limited benefit, such as ATM and CHEK2 mutations. This summary may force a change in the design of studies targeting this broad pathway seeking consistency in gene groupings.*

    *Relevance section written by JCO Associate Editor Michael A. Carducci, MD, FACP, FASCO.

Several poly(ADP-ribose) polymerase inhibitors (PARPi) have been approved by the US Food and Drug Administration (FDA) for treatment of patients with BRCA1m and/or BRCA2m across solid tumors, including metastatic castration-resistant prostate cancer (mCRPC).8-12 In addition to approvals of olaparib, rucaparib, and niraparib for patients with BRCAm-only mCRPC,13-15 the FDA approved olaparib and talazoparib for patients with HRRm mCRPC, with HRR panels consisting of BRCA and non-BRCA HRR genes.16,17 However, conclusions regarding PARPi activity by individual HRRm gene have been limited because of the small number of patients with each mutation in an individual trial and resultant challenges in controlling for baseline prognostic factors, differences in HRR panels across trials, and difficulty in isolating the contribution of effect of PARPi in add-on trials.

We conducted a pooled analysis to explore the correlation between individual HRRm genes and efficacy of PARPi in patients with each mutated gene.

PATIENTS AND METHODS

Study Design

Patient-level data were pooled by individual HRRm gene from trials of PARPi in patients with mCRPC submitted to the FDA in support of a marketing application.

Because of accumulating data across trials of various solid tumors including mCRPC showing strong association between BRCAm and efficacy of PARPi, patients with a BRCA comutation were excluded from the analysis of individual non-BRCA/HRR genes. For this analysis, the BRCA1 and BRCA2 subgroups were defined to exclude patients with a comutation in the other BRCA gene. Individual HRRm genes with available data across all the pooled randomized controlled trials (RCTs) were selected for efficacy analysis.

  1. Analysis of time-to-event end points (TTEs): all the RCTs of a PARPi plus an androgen receptor pathway inhibitor (ARPI) in the first-line mCRPC setting were included in the TTE analysis set. Patients were assigned to either of two treatment arms: (1) PARPi plus ARPI, or (2) placebo plus ARPI. The following TTEs were compared between the treatment arms:

  2. Radiographic progression-free survival (rPFS) by blinded independent central review (BICR). rPFS was defined as the time from random assignment until disease progression in soft tissue by RECIST v1.1,18 or in bone by Prostate Cancer Working Group 3 (PCWG3) criteria,19 or death.
    1. Overall survival (OS), defined as the time from random assignment until death.

To assess for the potential impact of comutation with other selected non-BRCA/HRRm genes, a sensitivity analysis of TTEs for each single HRRm gene was performed after exclusion of any comutation with the other five HRR genes.

  1. Analysis of objective response rate (ORR): all patients who received treatment and had BICR assessment of response were included in the ORR analysis set. All RECIST v1.1–based responses were confirmed with an imaging assessment ≥4 weeks from the initial response and required the absence of bone disease progression per PCWG3 criteria. ORR was reported for the following treatment groups: (1) PARPi plus ARPI and (2) placebo plus ARPI, both in first-line mCRPC; and (3) PARPi monotherapy in second-line mCRPC.

Statistical Analysis

Exploratory descriptive analyses were conducted for the pooled data set. The treatment effect of adding PARPi to ARPI on TTE including rPFS and OS for each individual gene was evaluated using Kaplan-Meier (KM) estimates and a Cox proportional hazards model. The hazard ratio (HR) of PARPi plus ARPI group versus control group (placebo plus ARPI) in the TTE analysis set and the corresponding Wald type 95% CI were estimated using a Cox model stratified by trial. The median rPFS and OS for each gene and each treatment group were estimated using the KM method with 95% CI calculated with the Brookmeyer-Crowley method using a log-log transform. Sensitivity analysis of rPFS and OS were conducted using a Cox regression model and further adjusted by several baseline covariates (baseline Eastern Cooperative Oncology Group performance status, metastasis site [bone only v visceral disease v others {lymph node ± bone}], and Gleason score [≤7 v >7]). An additional sensitivity analysis was conducted in each of the individual HRRm genes, after excluding comutation with all the other selected HRRm genes. Summary statistics including frequency and percentages were used for ORR assessment. All statistical analyses were done using R version 4.2.0 (R Core Team, Vienna, Austria).

RESULTS

Six clinical trials of PARPi in patients with mCRPC met the selection criteria and were included in this pooled analysis: three trials of PARPi plus ARPI in first-line mCRPC (all three were RCTs), and three trials of PARPi monotherapy (one RCT, and two single-arm trials).20,21,37,38,39,40 The characteristics of the pooled RCTs, including the treatment arms, diagnostic assays, and imaging intervals, are summarized in Appendix Table A1 (online only). Six HRR genes had data available across the three RCTs of PARP plus ARPI in first-line mCRPC and were included in the pooled analysis: ATM, BRCA1, BRCA2, CDK12, CHEK2, and PALB2. Selected HRR genes and treatment group assignment in the TTE and ORR analysis sets are shown in the CONSORT diagram in Figure 1.

FIG 1.

FIG 1.

Open in a new tab

Flow diagram. TTE analysis set: all RCTs were included in the TTE analysis set for evaluation of radiographic progression-free survival and overall survival. ORR analysis set: all patients with evaluable disease for response assessment per RECIST v1.1, who received treatment with PARPi plus ARPI and had BICR assessment of response were included in the analysis set for ORR. ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; HRR, homologous recombination repair; mCRPC, metastatic castration-resistant prostate cancer; ORR, objective response rate; PARPi, poly(ADP-ribose) polymerase inhibitor; RCT, randomized controlled trial; TTE, time-to-event end point.

Baseline characteristics of patients with each HRRm gene in the TTE analysis set are shown in Table 1. There were no large differences in baseline characteristics between the six HRRm gene groups.

TABLE 1.

Baseline Characteristics by Treatment Arm in the Pool of Three RCTs, by Individual HRR Gene Mutation

Gene Mutation Variable Level Control Treatment
ATM Age, years, median ± SD 72 ± 10 73 ± 10
Race, No. (%) Asian 26 (20) 27 (20)
Black/African American 2 (2) 4 (3)
White 89 (67) 95 (70)
Other 15 (11) 10 (7)
Ethnicity, No. (%) Hispanic 22 (17) 23 (17)
Non-Hispanic 98 (74) 104 (76)
Other or unknown 12 (9) 9 (7)
ECOG PS, No. (%) 0 87 (66) 83 (61)
1 45 (34) 53 (39)
2 0 0
Gleason score, No. (%) ≤7 34 (26) 43 (32)
8-10 96 (73) 90 (66)
Missing 2 (2) 3 (2)
Site of metastasis, No. (%) Bone only 65 (49) 54 (40)
Visceral metastasis 17 (13) 24 (18)
Other 50 (38) 58 (43)
Missing 0 0
BRCA1 Age, years, median ± SD 74 ± 15 67 ± 9
Race, No. (%) Asian 7 (26) 4 (11)
Black/African American 0 (0) 2 (5)
White 19 (70) 29 (78)
Other 1 (4) 2 (5)
Ethnicity, No. (%) Hispanic 3 (11) 9 (24)
Non-Hispanic 21 (78) 28 (76)
Other or unknown 3 (11) 0 (0)
ECOG PS, No. (%) 0 17 (63) 26 (70)
1 10 (37) 11 (30)
2 0 0
Gleason score, No. (%) ≤7 7 (26) 9 (24)
8-10 19 (70) 27 (73)
Missing 1 (4) 1 (3)
Site of metastasis, No. (%) Bone only 9 (33) 17 (46)
Visceral metastasis 3 (11) 10 (27)
Other 15 (56) 10 (27)
BRCA2 Age, years, median ± SD 69 ± 12 68 ± 12
Race, No. (%) Asian 41 (18) 34 (17)
Black/African American 1 (0) 7 (4)
White 162 (73) 138 (69)
Other 19 (9) 20 (10)
Ethnicity, No. (%) Hispanic 26 (12) 18 (9)
Non-Hispanic 177 (79) 157 (79)
Other or unknown 20 (9) 24 (12)
ECOG PS, No. (%) 0 150 (67) 129 (65)
1 73 (33) 70 (35)
2 0 0
Gleason score, No. (%) ≤7 68 (30) 39 (20)
8-10 148 (66) 148 (74)
Missing 7 (3) 12 (6)
Site of metastasis, No. (%) Bone only 89 (40) 75 (38)
Visceral metastasis 45 (20) 35 (18)
Other 89 (40) 89 (45)
Missing 0 0
CDK12 Age, years, median ± SD 69 ± 10 70 ± 10
Race, No. (%) Asian 24 (33) 28 (38)
Black/African American 2 (3) 0 (0)
White 42 (58) 43 (59)
Other 5 (7) 2 (3)
Ethnicity, No. (%) Hispanic 8 (11) 8 (11)
Non-Hispanic 61 (84) 65 (89)
Other or unknown 4 (5) 0 (0)
ECOG PS, No. (%) 0 47 (64) 46 (63)
1 26 (36) 27 (37)
2 0 0
Gleason score, No. (%) ≤7 12 (16) 8 (11)
8-10 59 (81) 64 (88)
Missing 2 (3) 1 (1)
Site of metastasis, No. (%) Bone only 36 (49) 28 (38)
Visceral metastasis 12 (16) 17 (23)
Other 25 (34) 28 (38)
Missing 0 0
CHEK2 Age, years, median ± SD 72 ± 10 73 ± 11
Race, No. (%) Asian 13 (15) 15 (18)
Black/African American 1 (1) 0 (0)
White 69 (79) 68 (80)
Other 4 (5) 2 (2)
Ethnicity, No. (%) Hispanic 13 (15) 7 (8)
Non-Hispanic 71 (82) 76 (89)
Other or unknown 3 (3) 2 (2)
ECOG PS, No. (%) 0 51 (59) 51 (60)
1 36 (41) 34 (40)
2 0 0
Gleason score, No. (%) ≤7 29 (33) 27 (32)
8-10 57 (66) 56 (66)
Missing 1 (1) 2 (2)
Site of metastasis, No. (%) Bone only 40 (46) 42 (49)
Visceral metastasis 16 (18) 7 (8)
Other 31 (36) 36 (42)
PALB2 Age, years, median ± SD 68 ± 16 67 ± 14
Race, No. (%) Asian 4 (22) 3 (13)
Black/African American 0 0
White 13 (72) 19 (83)
Other 1 (6) 1 (4)
Ethnicity, No. (%) Hispanic 3 (17) 2 (9)
Non-Hispanic 15 (83) 19 (83)
Other or unknown 0 (0) 2 (9)
ECOG PS, No. (%) 0 9 (50) 14 (61)
1 9 (50) 9 (39)
Gleason score, No. (%) ≤7 2 (11) 10 (43)
8-10 16 (89) 13 (57)
Site of metastasis, No. (%) Bone only 5 (28) 8 (35)
Visceral metastasis 3 (17) 7 (30)
Other 10 (56) 8 (35)
Open in a new tab

Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance status; HRR, homologous recombination repair; RCTs, randomized controlled trials; SD, standard deviation.

Results for each of the six individual HRRm genes are discussed below. Because of concern for overinterpretation of the results in a small number of patients and to avoid cross-trial comparisons, results of individual trials are not shown.

ATMm

Of 297 patients with a ATMm in the pool of three RCTs in first-line mCRPC, 29 patients (10%) with BRCA comutation were excluded and the remaining 268 patients were included in the TTE analysis set.

The median rPFS by BICR was 19 months in the PARPi plus ARPI arm and 19 months in the placebo plus ARPI arm, and the HR for rPFS was 1.05 (95% CI, 0.74 to 1.49). The median OS was 31 months in the PARPi plus ARPI arm and 31 months in the placebo plus ARPI arm, and the HR for OS was 1.07 (95% CI, 0.81 to 1.56). The results of sensitivity analysis of rPFS and OS in the ATMm subgroup after adjustment for covariates were consistent with the results of the primary analyses (Table 2; Fig 2A). The results were consistent across the three individual RCTs in first-line mCRPC (results of individual trials are not shown).

TABLE 2.

rPFS and OS by HRR Gene Mutation

Gene Mutation End Point Arm No. No. of Events Median Survival, Months HRa (95% CI) HRb (95% CI)
ATMm (N = 268) rPFS (BICR) PARPi + ARPI 136 67 19 (16-28) 1.05 (0.74 to 1.49) 1.02 (0.71 to 1.45)
Placebo + ARPI 132 59 19 (16-NE)
OS PARPi + ARPI 136 63 33 (26-NE) 1.18 (0.82 to 1.71) 1.12 (0.77 to 1.62)
Placebo + ARPI 132 55 33 (28-NE)
BRCA1m (N = 64) rPFS (BICR) PARPi + ARPI 37 14 20 (14-NE) 0.51 (0.23 to 1.1) 0.52 (0.2 to 1.33)
Placebo + ARPI 27 15 12 (4-NE)
OS PARPi + ARPI 37 16 29 (15-NE) 0.74 (0.34 to 1.61) 0.73 (0.28 to 1.89)
Placebo + ARPI 27 13 26 (13-28)
BRCA2m (N = 422) rPFS (BICR) PARPi + ARPI 199 60 NA (22-NE) 0.31 (0.23 to 0.42) 0.27 (0.19 to 0.37)
Placebo + ARPI 223 144 10 (8-11)
OS PARPi + ARPI 199 77 33 (29-NE) 0.66 (0.49 to 0.89) 0.6 (0.44 to 0.82)
Placebo + ARPI 223 112 24 (22-28)
CDK12m (N = 146) rPFS (BICR) PARPi + ARPI 73 32 17 (16-NE) 0.5 (0.32 to 0.8) 0.51 (0.32 to 0.82)
Placebo + ARPI 73 44 14 (8-17)
OS PARPi + ARPI 73 32 36 (25-NE) 0.63 (0.39 to 0.99) 0.64 (0.4 to 1.03)
Placebo + ARPI 73 44 27 (20-33)
CHEK2m (N = 172) rPFS (BICR) PARPi + ARPI 85 37 14 (14-25) 1.06 (0.67 to 1.66) 1.00 (0.63 to 1.59)
Placebo + ARPI 87 43 18 (13-22)
OS PARPi + ARPI 85 39 26 (23-NE) 1.53 (0.95 to 2.46) 1.48 (0.92 to 2.4)
Placebo + ARPI 87 32 34 (26-NE)
PALB2m (N = 41) rPFS (BICR) PARPi + ARPI 23 12 14 (8-NE) 0.52 (0.23 to 1.17) 0.43 (0.15 to 1.21)
Placebo + ARPI 18 13 9 (2-20)
OS PARPi + ARPI 23 12 25 (15-NE) 0.78 (0.34 to 1.8) 0.59 (0.21 to 1.65)
Placebo + ARPI 18 11 20 (11-NE)
Open in a new tab

NOTE. For ATM, CDK12, CHEK2, and PALB2, analysis set included any patient who had each of these mutations and excluded those with BRCA (BRCA1 and/or BRCA2) comutation, regardless of comutation with other HRR genes. For BRCA1 and BRCA2, analysis set excluded patients who had comutation with the other BRCA mutation, regardless of comutation with other HRR genes.

Abbreviations: ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio; HRR, homologous recombination repair; NA, not available; NE, not estimable; OS, overall survival; PARPi, poly(ADP-ribose) polymerase inhibitor; rPFS, radiographic progression-free survival.

a

HR estimated from a Cox proportional hazards model stratified by trial (HR <1 favors the treatment group).

b

HR estimated from a Cox proportional hazards model stratified by trial and adjusted by baseline ECOG PS, metastasis site (bone only v visceral disease v others), and Gleason score (≤7 v >7; HR <1 favors the PARPi group).

FIG 2.

FIG 2.

Open in a new tab

Kaplan-Meier curves of rPFS by BICR and OS by each HRR gene mutation (A: ATM; B: BRCA1; C: BRCA2; D: CDK12; E: CHEK2; F: PALB2). ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; HR, hazard ratio; HRR, homologous recombination repair; OS, overall survival; PARPi, poly(ADP-ribose) polymerase inhibitor; rPFS, radiographic progression-free survival.

The ORR with PARPi monotherapy was 7% (five of 70 patients; 95% CI, 2 to 16; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 42% (17 of 40 patients; 95% CI, 27 to 59) and 59% (32 of 54 patients; 95% CI, 45 to 72), respectively (Table 4).

TABLE 3.

ORR of PARPi in mCRPC, by HRR Gene Mutation

HRR Gene Mutation ORR by BICR, % (x/y),
(95% CI)a 		Difference Between PARPi + ARPI v Placebo + ARPI in First-Line mCRPC (∆)
Second-Line mCRPC First-Line mCRPC
PARPi Monotherapyb Placebo + ARPIc PARPi + ARPIc
ATMm 7 (5/70)
(3 to 16)		 42 (17/40)
(19 to 66)		 59 (32/54)
(46 to 71)		 ∆ = +17%
BRCA1m 20 (2/10)
(1 to 83)		 18 (2/11)
(5 to 51)		 46 (6/13)
(18 to 80)		 ∆ = +28%
BRCA2m 47 (51/108)
(38 to 57)		 27 (25/94)
(19 to 36)		 60 (53/89)
(44 to 79)		 ∆ = +33%
CDK12m 5 (2/44)
(1 to 16)		 32 (7/22)
(16 to 53)		 62 (20/32)
(45 to 77)		 ∆ = +30%
CHEK2m 0 (0/17)
(0 to 100)		 33 (9/27)
(14 to 60)		 43 (13/30)
(27 to 61)		 ∆ = +10%
PALB2m 14 (1/7)
(2 to 58)		 45 (5/11)
(20 to 73)		 43 (3/7)
(14 to 77)		 ∆ = –2%
Open in a new tab

NOTE. Confirmed ORR. x: number of patients with response per RECIST v 1.1 criteria. y: number of patients with evaluable disease for response assessment.

Abbreviations: ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; HRR, homologous recombination repair; mCRPC, metastatic castration-resistant prostate cancer; ORR, objective response rate; PARPi, poly(ADP-ribose) polymerase inhibitor.

a

Calculated based on a generalized linear mixed effect model with logit transformation.

b

Pool of three trials of PARPi monotherapy, that is, PROfound, TRITON2, and TALAPRO1 (all were trials of a PARPi monotherapy in second-line mCRPC).

c

Pool of three trials of PARPi plus ARPI, that is, PROpel, MAGNITUDE, and TALAPRO2 (all were randomized trials of an ARPI plus PARPi or placebo).

TABLE 4.

Considerations for Efficacy by Individual HRR Gene Mutation

HRR Gene Mutation Interpretation of Efficacy Results
ATMm ATMm was the second largest subgroup of this pooled data set. In this exploratory analysis, an apparent lack of treatment effect of PARPi for patients with ATMm mCRPC was observed (HR for rPFS and OS >1) when adding PARPi to ARPI, which should be further investigated. The ORR in the subgroup of patients with ATMm mCRPC who received PARPi monotherapy was only 7% and the difference in ORR between ARPI with or without PARPi was +17%, both indicating a modest treatment effect of PARPi either as monotherapy or with an APRI.
BRCA1m Interpretation of the results in patients with BRCA1m tumor is limited because of the small number of patients in this subgroup. Results of rPFS analysis and OS analysis suggest favorable outcomes of PARPi therapy in patients with BRCA1m mCRPC. The ORR in PARPi monotherapy group, and the increase in ORR by adding PARPi to ARPI, provides further supportive evidence of efficacy of PARPi in patients with BRCA1m mCRPC. Although the 95% CIs for rPFS and OS are wide because of small number of patients in this subgroup, overall, the results suggest a favorable treatment effect of PARPi in patients with BRCA1m mCRPC, consistent with results of PARPi trials across other tumors.22
BRCA2m BRCA2m was the largest subgroup in this pooled data set. Results of both PARPi plus ARPI and PARPi monotherapy trials were consistent and demonstrated a strong treatment effect of PARPi therapy. These results are consistent with the results of PARPi in other BRCA2m solid tumors.22
CDK12m In the CDK12m subgroup, both rPFS and OS favored the PARPi arm. The low response rate of PARPi monotherapy (5%) and 30% increase in the point estimate of ORR in PARPi plus ARPI compared with ARPI + placebo may be indicative of a synergistic effect of PARPi and ARPI or may be merely a chance finding.
CHEK2m In this exploratory analysis, an apparent lack of treatment effect of PARPi for patients with CHEK2m mCRPC was observed (HR for rPFS and OS >1) when adding PARPi to ARPI, which should be further investigated. There was no objective response with PARPi monotherapy per BICR assessment in 17 patients with CHEK2m who had evaluable disease for response assessment, and the difference in ORR for ARPI with or without PARPi was 10%, both indicating a modest treatment effect of PARPi either as monotherapy or with an ARPI.
PALB2m PALB2m was the smallest subgroup in our analysis, with only 49 patients. The point estimates for HRs for rPFS and OS were favorable, suggesting potential efficacy for treatment of patients with PALB2m mCRPC with PARPi. Similar to the BRCA1m subgroup, the small number of patients in PALB2m subgroup resulted in wide confidence intervals for rPFS and OS, and the number of patients with evaluable disease for response was even smaller, which made the interpretation of ORR results in this subgroup difficult.
Open in a new tab

Abbreviations: ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; HR, hazard ratio; HRR, homologous recombination repair; mCRPC, metastatic castration-resistant prostate cancer; ORR, objective response rate; OS, overall survival; PARPi, poly(ADP-ribose) polymerase inhibitor; rPFS, radiographic progression-free survival.

BRCA1m

Of 71 patients with a BRCA1m in the pool of three RCTs in first-line mCRPC, seven patients (10%) with BRCA2 comutation were excluded and the remaining 64 patients were included in the TTE analysis set.

The median rPFS by BICR was 20 months in the PARPi plus ARPI arm and 12 months in the placebo plus ARPI arm, and the HR for rPFS was 0.51 (95% CI, 0.23 to 1.1). The median OS was 29 months in the PARPi plus ARPI arm and 26 months in the placebo plus ARPI arm, and the HR for OS was 0.74 (95% CI, 0.34 to 1.61). The results of sensitivity analyses of rPFS and OS in the BRCA1m subgroup after adjustment for covariates were consistent with the results of the primary analyses (Table 2; Fig 2B). In two of three RCTs in first-line mCRPC, the results of the BRCA1m subgroup were favorable in the PARPi plus ARPI arm (results of individual RCTs are not shown).

The ORR with PARPi monotherapy was 20% (two of 10 patients; 95% CI, 3 to 56; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 18% (two of 11 patients; 95% CI, 2 to 52) and 46% (six of 13 patients; 95% CI, 19 to 75), respectively (Table 4).

BRCA2m

Of 429 patients with a BRCA2m in the pool of three RCTs, seven patients (2%) with BRCA1 comutation were excluded and the remaining 422 patients were included in the TTE analysis set.

The median rPFS by BICR was not reached in the PARPi plus ARPI arm and was 10 months in the placebo plus ARPI arm; the HR for rPFS was 0.31 (95% CI, 0.23 to 0.42). The median OS was 33 months in the PARPi plus ARPI arm and 24 months in the placebo plus ARPI arm; the HR for OS was 0.66 (95% CI, 0.49 to 0.89). The results of sensitivity analysis of rPFS and OS in the BRCA2m subgroup after adjustment for covariates were consistent with the results of the primary analyses (Table 2; Fig 2C). These findings were consistent across all three RCTs in first-line mCRPC that were included in the pooled analysis (results of individual RCTs are not shown).

The ORR with PARPi monotherapy was 47% (51 of 108 patients; 95% CI, 38 to 57; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 26% (25 of 95 patients; 95% CI, 18 to 37) and 60% (55 of 91 patients; 95% CI, 49 to 70), respectively (Table 4).

CDK12m

Of 160 patients with a CDK12m in the pool of three RCTs, 14 patients (9%) with BRCA comutation were excluded and the remaining 146 patients were included in the TTE analysis set.

The median rPFS by BICR was 17 months in the PARPi plus ARPI arm and 14 months in the placebo plus ARPI arm, and the HR for rPFS was 0.50 (95% CI, 0.32 to 0.80). The median OS was 36 months in the PARPi plus ARPI arm and 27 months in the placebo plus ARPI arm, and the HR for OS was 0.63 (95% CI, 0.39 to 0.99). The results of sensitivity analysis of rPFS and OS in the CDK12m subgroup after adjustment for covariates were consistent with the results of primary analyses (Table 2; Fig 2D). In two of three RCTs in first-line mCRPC, the results in the CDK12m subgroup were favorable for the PARPi plus ARPI arm (results of individual RCTs are not shown).

The ORR with PARPi monotherapy was 5% (two of 44 patients; 95% CI, 1 to 15; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 32% (seven of 22 patients; 95% CI, 14 to 55) and 62% (20 of 32 patients; 95% CI, 44 to 79), respectively (Table 4).

CHEK2m

Of 197 patients with a CHEK2m in the pool of three RCTs, 25 patients (13%) with BRCA comutation were excluded and the remaining 172 patients were included in the TTE analysis set.

The median rPFS by BICR was 17 months in the PARPi plus ARPI arm and 18 months in the placebo plus ARPI arm, and the HR for rPFS was 1.06 (95% CI, 0.67 to 1.66). The median OS was 26 months in the PARPi plus ARPI arm and 34 months in the placebo plus ARPI arm, and the HR for OS was 1.53 (95% CI, 0.95 to 2.46). The results of sensitivity analysis of rPFS and OS in the CHEK2m subgroup after adjustment for covariates were consistent with the results of primary analyses (Table 2; Fig 2E). The results were consistent across the three individual RCTs in first-line mCRPC (results of individual trials are not shown).

The ORR with PARPi monotherapy was 0% (zero of 17 patients; 95% CI, 0 to 20; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 33% (nine of 27 patients; 95% CI, 17 to 54) and 43% (13 of 30 patients; 95% CI, 25 to 63), respectively (Table 4).

PALB2m

Of 44 patients with a PALB2m in the pool of three RCTs, three patients (7%) with BRCA comutation were excluded and the remaining 41 patients were included in the TTE analysis set.

The median rPFS by BICR was 14 months in the PARPi plus ARPI arm and 9 months in the placebo plus ARPI arm, and the HR for rPFS was 0.52 (95% CI, 0.23 to 1.17). The median OS was 25 months in the PARPi plus ARPI arm and 20 months in the placebo plus ARPI arm, and the HR for OS was 0.78 (95% CI, 0.34 to 1.80). The results of sensitivity analysis of rPFS and OS in the PALB2m subgroup after adjustment for covariates were consistent with the results of primary analyses (Table 2; Fig 2F). These findings were consistent across all three RCTs in first-line mCRPC that were included in the pooled analysis (results of individual trials are not shown).

The ORR with PARPi monotherapy was 14% (one of seven patients; 95% CI, 0 to 58; Table 3). The ORRs with placebo plus ARPI, and PARPi plus ARPI were 45% (five of 11 patients; 95% CI, 17 to 77) and 43% (three of seven patients; 95% CI, 10 to 82), respectively (Table 4).

DISCUSSION

In this exploratory pooled analysis, individual tumor gene mutations involved in the HRR pathway demonstrated heterogeneous sensitivity to treatment with PARPi. Benefit from PARPi appeared greatest for patients with BRCA1m, BRCA2m, CDK12m, and PALB2m, while no strong treatment effect of PARPi was observed for patients with mutations in either ATMm or CHEK2m when adding a PARPi to ARPI. Considering the small sample sizes and lack of randomization/stratification by gene, observations around the apparent lack of treatment effect of PARPi for patients with CHEK2m or ATMm require confirmation in future clinical trials. More detailed discussion of the results for each subgroup are presented in Table 4. The results of sensitivity analysis for each of these genes, after excluding patients with mutations in any of the other five HRR genes (Appendix Table A2), were consistent with the results of the primary analysis, demonstrating a large difference in magnitude of treatment effect from adding PARPi to ARPI across these HRR gene mutations. This is an important consideration for the design of future trials of PARPi in this setting. The rationale for grouping several subgroups with different HRR-associated gene mutations as one cohort in clinical trials has primarily been based on the shared mechanistic pathway, with very limited preclinical/clinical data from small early-phase studies. However, emerging data from late-phase clinical trials of PARPi indicate that the treatment effect in HRRm cohorts appears to be primarily attributed to the BRCAm subgroup nested in the HRRm cohorts. These findings indicate that pooling a diverse set of HRR genes as one cohort without separate prespecified analysis of BRCAm and non-BRCAm groups in clinical trials evaluating PARPi no longer appears justifiable. Additionally, it may not be justifiable to pool ATMm and CHEK2 with other HRR genes as one cohort, given the large differences in treatment effect in patients with ATMm and CHEK2m compared with those with mutations in other HRR genes. For these genes, a separate analysis may be warranted in future trials and consideration given to early futility analyses if adequate efficacy in patients with these genes is not demonstrated.

One limitation of our pooled analysis is that the three pooled RCTs varied with regards to the PARPi and ARPI drugs as well as the patients' baseline characteristics (Appendix Table A1). To account for the potential impact of these differences on results, we stratified the TTE analyses by trial. Additionally, we evaluated the efficacy results of each individual trial for consistency and took these into consideration when interpreting the results of the pooled analysis.

The results of our pooled analysis demonstrating heterogeneity in the sensitizing effect of HRRm genes to PARPi are consistent with published results from two other trials of PARPi in mCRPC that were not included in the pooled analysis. TRITON3 trial (ClinicalTrials.gov identifier: NCT02975934) was an open-label RCT in patients with BRCA1m, BRCA2m, or ATMm mCRPC that randomly assigned patients to receive either rucaparib or the physician's choice of docetaxel or ARPI. Randomization was stratified by HRRm gene. Although the results in the BRCAm subgroup showed evidence of efficacy (rPFS HR, 0.5 [95% CI, 0.36 to 0.69]), the results in the ATMm subset in TRITON3 were unfavorable (rPFS HR, 0.95 [95% CI, 0.52 to 1.52] and OS HR, 1.20 [95% CI, 0.74 to 1.95]).23 In TOPARP B (ClinicalTrials.gov identifier: NCT01682772), a trial of olaparib in patients with HRRm mCRPC, the confirmed ORR (per RECIST v1.1) and PSA50 response (≥50% decrease in PSA from baseline) in the ATMm subgroup were both <10%, while the confirmed ORR and PSA50 response rate were 52% and 77%, respectively, in the BRCAm subgroup. There was no objective response or PSA50 response in the CDK12m subgroup. In six patients with PALB2m, the ORR was 33% and the PSA50 response rate was 67%.24

The heterogeneity of the PARPi treatment effect by HRRm gene may be related to the role of each of these genes in the HRR pathway. ATM and CHEK2 are sensors of DNA damage; CDK12 regulates the transcription of long DNA repair genes; and BRCA1, BRCA2, and PALB2 are HRR effectors that are recruited by the sensors to the DNA break site to repair the break.25-30 On the basis of the results of our pooled analysis, mutations in HRR effector genes appeared to be associated with greater efficacy from PARPi therapy compared with mutations in sensors of DNA breaks (eg, ATM and CHEK2). The function of CDK12 remains poorly defined and, given the heterogeneity of the treatment effect across the pooled trials for patients with CDK12m and the large difference between the response rates for PARPi monotherapy and plus an ARPI for these patients, further preclinical and clinical investigation is warranted to better understand potential interactions between PARPi and ARPI.

Another potential indicator of sensitivity to PARPi is the level of homologous recombination deficiency (HRD) that results from each HRR gene mutation. In a retrospective study, the tumor HRD score was evaluated in three cohorts of primary prostate cancer. In a cohort of 64 patients with germline mutations in ATM, BRCA2, or CHEK2, BRCA2m tumors had a higher HRD score compared with ATMm and CHEK2m tumors,31 providing a potential explanation for the apparent difference in PARPi sensitivity among these mutated genes.

Presence of clonal hematopoiesis of indeterminate potential (CHIP) in blood samples can result in false-positive results of HRR testing using circulating tumor DNA (ctDNA) in blood.32 Although a small proportion of patients with HRRm in the pooled trials are expected to have false-positive results because of reasons such as presence of CHIP in blood, most patients in these trials also had a positive tumor tissue test result. Given the relatively small number of patients enrolled solely on the basis of a positive ctDNA result and the high level of concordance between positive results for ctDNA and tumor tissue,33 CHIP is unlikely to explain the heterogeneity in PARPi sensitivity between different gene subgroups.

This analysis evaluated the six HRRm genes most prevalent in studies of PARPi in mCRPC. Although evaluation of less common HRRm genes was not possible for this analysis on the basis of the limited numbers of patients, additional data from ongoing trials of PARPi in HRRm prostate cancer may allow sufficient sample sizes to evaluate these rare gene mutations.34-36

In conclusion, in this exploratory pooled analysis, benefit from PARPi appeared greatest for patients with BRCA1m, BRCA2m, CDK12m, and PALB2m mCRPC. Although there was an apparent lack of treatment effect in patients with either CHEK2m or ATMm, limitations of the analysis necessitate further prospective evaluation. The heterogeneity in PARPi sensitivity among HRR-associated genes suggests that a shared mechanistic pathway alone may not be adequate to justify pooling a diverse set of genes for clinical trials evaluating PARPi. Future clinical trials should consider this heterogeneity in their design. Potential approaches may include stratification by BRCAm versus other non-BRCA HRR genes or evaluation of these populations in separate cohorts or other gene panel groupings that account for available evidence for the by-gene treatment effect.

In all the trials that were pooled for this analysis, the investigators performed the human investigations after approval by a local human investigations committee. All the patient-level data from these trials were anonymized to protect the identities of patients involved in the research.

In all the trials that were pooled for this analysis, the investigators obtained informed consent from each participant or each participant's guardian.

APPENDIX

TABLE A1.

Characteristics of the Pooled Trials

Clinical Trial Study Population PARPi Arm Control Arm Stratification Factors HRR Genes Diagnostic Assays Imaging Interval
PROfound37 Second-line mCRPC Olaparib Abiraterone or enzalutamide Previous taxane (yes v no)
Measurable disease (yes v no)		 ATM
BRCA1
BRCA2
BRIP1
BARD1
CDK12
CHEK1
CHEK2
FANCL
PALB2
PPP2R2A
RAD51B
RAD51C
RAD51D
RAD54L 		FoundationOne CDx; FoundationOne Liquid CDx CT/MRI and bone scan; every 8 weeks (±7 days) relative to the date of random assignment, until disease progression by BICR
PROpel20 First-line mCRPC Olaparib plus abiraterone Abiraterone plus placebo Metastases (bone only v visceral v other)
Docetaxel treatment at mCSPC stage (yes v no)		 ATM
BRCA1
BRCA2
BARD1
BRIP1
CDK12
CHEK1
CHEK2
FANCL
PALB2
RAD51B
RAD51C
RAD51D
RAD54L 		FoundationOne CDx; FoundationOne Liquid CDx CT/MRI and bone scan; every 8 weeks (±7 days) for the first 24 weeks, then every 12 weeks (±7 days) relative to the date of random assignment
TALAPRO-238 First-line mCRPC Talazoparib plus enzalutamide Enzalutamide plus placebo HRR gene alteration status (HRRm v non-HRRm v unknown)
Previous treatment with life-prolonging therapy (docetaxel or abiraterone, or both) in the castration-sensitive setting (yes v no)		 ATM
ATR
BRCA1
BRCA2
CDK12
CHEK2
FANCA
NBN
MLH1
MRE11A
PALB2
RAD51C 		FoundationOne CDx; FoundationOne Liquid CDx CT/MRI and bone scan; every 8 weeks through week 25 and then every 12 weeks until radiographic progression is determined by investigator (part 1) or by investigator and BICR (part 2)
MAGNITUDE39 First-line mCRPC Niraparib plus abiraterone Abiraterone plus placebo Cohort 1 and 2:
 Previous treatment with taxane-based chemotherapy (yes v no)
 Previous treatment with novel antiandrogen therapy such as enzalutamide, apalutamide, and darolutamide (yes v no)
 Previous treatment with abiraterone (yes v no).
Cohort 1 only:
 Gene alteration group (BRCAm v non-BRCA HRRm)		 ATM
BRCA1
BRCA2
BRIP1
CDK12
CHEK2
FANCA
HDAC2
PALB2 		FoundationOne CDx; Resolution Bioscience HRD plasma test; AmoyDx blood and tissue assays; Accredited local laboratory tests CT/MRI and bone scan; at cycle 3 day 1, cycle 5 day 1, cycle 7 day 1, and then every 12 weeks until radiographic progression (a treatment cycle was defined as 28 days)
TRITON-240 ≥Second-line mCRPC Rucaparib NA (single-arm trial) NA (single-arm trial) ATM
BRCA1
BRCA2
BARD1
BRIP1
CDK12
CHEK2
FANCA
NBN
PALB2
RAD51
RAD51B
RAD51C
RAD51D
RAD54L 		Central testing by Foundation Medicine. Germline testing was performed by Color Genomics CT and bone scan (MRI, X-ray, PET-CT, and ultrasound if required); every 8 weeks for up to 24 weeks, then every 12 weeks
TALAPRO-121 ≥Second-line mCRPC Talazoparib NA (single-arm trial) NA (single-arm trial) ATM
ATR
BRCA1
BRCA2
CHEK2
FANCA
MLH1
MRE11A
NBN
PALB2
RAD51C 		FoundationOne CDx CT/MRI and bone scan; every 8 weeks through week 25, then every 12 weeks thereafter
Open in a new tab

Abbreviations: BICR, blinded independent central review; CT, computed tomography; HRR, homologous recombination repair; mCRPC, metastatic castration-resistant prostate cancer; mCSPC, metastatic castration-sensitive prostate cancer; MRI, magnetic resonance imaging; NA, not available; PARPi, poly(ADP-ribose) polymerase inhibitor; PET, positron emission tomography.

TABLE A2.

rPFS and OS by Each Single HRR Gene Mutation (without comutation with the other five genes)

Gene Mutation End Point Arm No. No. of Events Median Survival, Months HRa (95% CI) HRb (95% CI)
ATMm rPFS BICR PARP + ARPI 111 52 19 (17-NE) 0.96 (0.65 to 1.42) 0.93 (0.62 to 1.39)
Placebo + ARPI 112 49 19 (14-NE)
OS PARP + ARPI 111 46 37 (29-NE) 0.95 (0.63 to 1.43) 0.89 (0.59 to 1.36)
Placebo + ARPI 112 48 33 (28-NE)
BRCA1m rPFS BICR PARP + ARPI 28 9 NE (16-NE) 0.44 (0.17 to 1.12) 0.23 (0.06 to 0.91)
Placebo + ARPI 18 10 12 (4-NE)
OS PARP + ARPI 28 10 NE (24-NE) 0.63 (0.24 to 1.68) 0.46 (0.12 to 1.69)
Placebo + ARPI 18 8 26 (15-NE)
BRCA2m rPFS BICR PARP + ARPI 178 54 NE (22-NE) 0.29 (0.21 to 0.4) 0.26 (0.18 to 0.36)
Placebo + ARPI 183 121 10 (8-11)
OS PARP + ARPI 178 69 33 (28-NE) 0.66 (0.48 to 0.9) 0.63 (0.46 to 0.88)
Placebo + ARPI 183 96 24 (21-28)
CDK12m rPFS BICR PARP + ARPI 55 24 22 (16-NE) 0.53 (0.31 to 0.89) 0.54 (0.31 to 0.93)
Placebo + ARPI 61 33 16 (11-19)
OS PARP + ARPI 55 20 NE (29-NE) 0.47 (0.27 to 0.83) 0.49 (0.28 to 0.88)
Placebo + ARPI 61 36 29 (20-34)
CHEK2m rPFS BICR PARP + ARPI 58 21 NE (14-NE) 0.84 (0.47 to 1.5) 0.8 (0.44 to 1.44)
Placebo + ARPI 64 31 18 (11-NE)
OS PARP + ARPI 58 21 NE (25-NE) 0.97 (0.54 to 1.74) 0.96 (0.53 to 1.75)
Placebo + ARPI 64 26 34 (24-NE)
PALB2m rPFS BICR PARP + ARPI 19 10 29 (8-NE) 0.41 (0.16 to 1.04) 0.27 (0.07 to 0.98)
Placebo + ARPI 13 10 11 (2-14)
OS PARP + ARPI 19 9 29 (15-NE) 0.63 (0.24 to 1.65) 0.49 (0.14 to 1.73)
Placebo + ARPI 13 9 20 (11-38)
Open in a new tab

Abbreviations: ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; HR, hazard ratio; HRR, homologous recombination repair; NE, not estimable; OS, overall survival; PARP, poly(ADP-ribose) polymerase; rPFS, radiographic progression-free survival.

a

HR estimated from a Cox proportional hazards model stratified by trial (HR <1 favors the treatment group).

b

HR estimated from a Cox proportional hazards model stratified by trial and adjusted by baseline Eastern Cooperative Oncology Group performance status, metastasis site (bone only v visceral disease v others) and Gleason score (≤7 v >7; HR < 1 favors the PARPi group).

William F. Maguire

Employment: University of Pittsburgh Medical Center, University of Pittsburgh Medical Center/Center for Connected Medicine

Patents, Royalties, Other Intellectual Property: US-11660357-B2, One-step labeling of antibodies to high specific activity with actinium-225. I have officially declined personal royalties resulting from this patent.

Sundeep Agrawal

Employment: United Health Group

Daniel L. Suzman

Other Relationship: Included Health

No other potential conflicts of interest were reported.

DISCLAIMER

This is a US Government work. There are no restrictions on its use.

*

J.F. and J.X. contributed equally to this work.

DATA SHARING STATEMENT

The sharing of individual patient data from each participating trial will be subject to the policies and procedures of the institution(s) and group(s) who did the original studies.

AUTHOR CONTRIBUTIONS

Conception and design: Jaleh Fallah, Jianjin Xu, Chana Weinstock, Brian L. Heiss, Elaine Chang, Sundeep Agrawal, Shenghui Tang, Richard Pazdur, Paul G. Kluetz, Daniel L. Suzman, William F. Maguire

Administrative support: Richard Pazdur

Collection and assembly of data: Jaleh Fallah, Jianjin Xu, Brian L. Heiss, William F. Maguire, Elaine Chang, Richard Pazdur

Data analysis and interpretation: Jaleh Fallah, Jianjin Xu, Chana Weinstock, Xin Gao, Brian L. Heiss, William F. Maguire, Elaine Chang, Shenghui Tang, Laleh Amiri-Kordestani, Richard Pazdur, Daniel L. Suzman

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Efficacy of Poly(ADP-ribose) Polymerase Inhibitors by Individual Genes in Homologous Recombination Repair Gene-Mutated Metastatic Castration-Resistant Prostate Cancer: A US Food and Drug Administration Pooled Analysis

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

William F. Maguire

Employment: University of Pittsburgh Medical Center, University of Pittsburgh Medical Center/Center for Connected Medicine

Patents, Royalties, Other Intellectual Property: US-11660357-B2, One-step labeling of antibodies to high specific activity with actinium-225. I have officially declined personal royalties resulting from this patent.

Sundeep Agrawal

Employment: United Health Group

Daniel L. Suzman

Other Relationship: Included Health

No other potential conflicts of interest were reported.

REFERENCES

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The sharing of individual patient data from each participating trial will be subject to the policies and procedures of the institution(s) and group(s) who did the original studies.

Articles from Journal of Clinical Oncology are provided here courtesy of Wolters Kluwer Health

RESOURCES