Atezolizumab with enzalutamide versus enzalutamide alone in metastatic castration-resistant prostate cancer: a randomized phase 3 trial

Abstract

Early clinical data indicate that some patients with castration-resistant prostate cancer may benefit from program death ligand-1 (PD-L1) inhibition, especially with enzalutamide. The IMbassador250 trial (no. NCT03016312) enrolled 759 men with metastatic castration-resistant prostate cancer whose disease progressed on abiraterone. The addition of atezolizumab to enzalutamide in an open-label randomized trial did not meet the primary endpoint of improved overall survival in unselected patients (stratified hazard ratio 1.12, 95% confidence interval (0.91, 1.37), P = 0.28), despite an acceptable safety profile. In archival tumor samples, prostate tumors showed comparatively low expression of key immune biomarkers. DNA damage-response alterations, phosphatase and tensin homolog status and PD-L1 expression levels were similar between hormone-sensitive and castration-resistant prostate cancers. In planned biomarker analysis, longer progression-free survival was seen with atezolizumab in patients with high PD-L1 IC2/3, CD8 expression and established immune gene signatures. Exploratory analysis linked progression-free survival in the atezolizumab arm with immune genes such as CXCL9 and TAP1, together with other potentially relevant biomarkers including phosphatase and tensin homolog alterations. Together these data indicate that the expected biology associated with response to immune checkpoint inhibitors is present in prostate cancer, albeit in fewer patients. Careful patient selection may be required for immune checkpoint inhibitors to identify subgroups of patients who may benefit from this treatment approach.

Extended Data Fig. 1. Forest plot of subgroup analysis

(a) PFS and (b) rPFS in the intention-to-treat (ITT) population. P value and HR are from the unstratified Cox regression model. Prior local therapy included prior radical prostatectomy or radiotherapy. PD-L1–positive immune cells (IC) defined as: IC0, <1%; IC1/2/3, ≥1%; IC2/3, ≥5%.

Extended Data Fig. 2. Forest plot of known biomarkers in urothelial carcinoma, renal cell carcinoma and prostate cancer

Biomarkers shown among urothelial carcinoma (IMvigor210), renal cell carcinoma (IMmotion150), and prostate cancer (IMbassador250) for PFS. DDR, DNA damage response; PD-L1, programmed death- ligand 1; PFS, progression-free survival; T_eff, effector T cell, TMB, tumour mutational burden. Med refers to median PFS in months. HRs and CIs were calculated using Cox proportional hazards regression model, and P values were calculated using unstratified log-rank test without adjustment for multiplicity.

Extended Data Fig. 3. Exploratory analysis of DNA alterations in progression free survival

PFS in the atezolizumab + enzalutamide vs enzalutamide treatment arms. 325 samples were included for analysis. DNA alteration biomarkers included Androgen Receptor (AR) amplification status, v-ets erythroblastosis virus E26 oncogene homolog (ERG) fusions, alterations of TP53, BRCA2, SPOP, CDK12 (at least a frameshift, nonsense or splice-site alteration) and ATM. In addition, frameshift mutation burden (FSB) was included. FSB was calculated by the number of frameshift mutations divided by length of genome examined. It was reported as the number of frameshift mutations per megabase (mut/Mb). The cutoff of FSB (8.7 mut/Mb) was previously established in prostate cancer. HRs and CIs were calculated using Cox proportional hazards regression model, and P values were calculated using log-rank test. MST refers to median survival time (PFS) in months.

Extended Data Fig. 4. Distribution of biomarkers and PD-L1 IC status.

Analysis of PTEN loss, Teff signature, DNA Damage Response (DDR) alterations and Androgen Receptor (AR) amplification status and PD-L1 IC status. IC0/1 were considered low IC whereas IC2/3 were considered high IC. Numbers on the bars indicate the number of patients being analysed.

Figure 1|

IMbassador250 consort diagram. A total of 759 patients with metastatic castration-resistant prostate cancer who progressed on abiraterone and were ineligible for or declined taxane chemotherapy gave consent and entered the study between June 2017 and May 2018. The number of immunohistochemistry (IHC) and next generation sequencing (NGS) samples of biomarker evaluable population in both arms is shown. FMI: Foundation Medicine.

Figure 2|

Clinical efficacy in the atezolizumab + enzalutamide vs enzalutamide treatment arms. (a) OS Kaplan-Meier curve. 55.7% of patients (211 of 379) in the atezolizumab + enzalutamide arm and 47.9% of patients (182 of 380) in the enzalutamide arm had an event. The stratified HR interaction P value for OS from the stratified Cox regression model was 0.28. The primary comparison of OS between treatment arms was based on a stratified log-rank test. The HR for death in the experimental arm compared with the control arm was estimated using a stratified Cox regression model, and the 95% CI was provided. (b) Forest plot. P value is from the unstratified Cox regression model. HR is shown for OS in the intention-to-treat (ITT) population. Prior local therapy included prior radical prostatectomy or radiotherapy. PD-L1–positive immune cells (IC) defined as: IC0, <1%; IC1/2/3, ≥1%; IC2/3, ≥5%. Median survival follow-up, 13 months. Minimum follow-up, 12 months. (c) rPFS Kaplan-Meier curve. rPFS was assessed by the investigator and adapted from the Prostate Cancer Working Group 3 (PCWG3) criteria. 74.4% of patients (282 of 379) in the atezolizumab + enzalutamide arm and 72.4% of patients (275 of 380) in the enzalutamide arm had an event. HR is stratified by the Cox proportional hazards model. (d) Time to PSA progression Kaplan-Meier curve for per PCWG3 criteria. 72.3% of patients (274 of 379) in the atezolizumab + enzalutamide arm and 75.0% of patients (285 of 380) in the enzalutamide arm had an event. HR was stratified by Cox proportional hazards model. (e) Waterfall plots showing maximum PSA decline in atezolizumab + enzalutamide and Enzalutamide arm. CR/PR were considered responders. SD, PD, non CR/PD and unable to evaluate were considered non responders. Clinical cutoff for b-e, 24 June 2019.

Figure 3|. Ad hoc analysis:

A comparison of DDR alterations, PTEN status, Teff. AR alteration status and PD-L1 expression between hormone-sensitive prostate cancer (HSPC) and castration-resistant prostate cancer (CRPC) biopsy samples. Patient samples were not sequential from the same patient. Each patient contributed only one sample. Numbers in the bars indicate the number of patients being analysed. DDR genes included for analysis were BRCA1, BRCA2, ATM, CDK12, PALB2, ARID1A, ATRX, CHEK1, CHEK2, FANCA, FANCF, FANCG, FANCI, FANCM, MSH2, NBN, RAD52 and WRN. Two-sided Fisher’s exact test with no adjustment of multiplicity was performed in each comparison with P values shown on top.

Figure 4|. Ad hoc analysis:

(a) A comparison of T_eff, MHC I, macrophage levels and immune checkpoint among 3 genitourinary cancers (renal cell carcinoma, urothelial carcinoma, and prostate cancer). P values were calculated by 2-sided Mann-Whitney u test. ***P < 0.001; **P < 0.01; *P < 0.05. Individual samples analysed: n=263 for RCC; n=348 for UBC and n=400 for prostate. The box plots follow standard Tukey representation by depicting the median at the middle line, with the lower and upper hinges at the first and third quartiles, respectively, the whiskers showing the minima to maxima no greater than 1.5× the interquartile range, and the remaining outlying data points plotted individually. Values for maxima, minima, Q1, Q2, Q3, upper whisker, lower whisker: Teff-RCC: 1.64, −5.53, 0.084, 0.35, 0.60, 1.37, −0.69. UBC: 2.17, −5.48, 0.088, 0.51, 0.84, 1.97, −1.04. Prostate: 0.60, −1.23, −0.58, −0.39, −0.24, 0.28, −1.10. APM- RCC: 2.10, −0.78, 0.35, 0.75, 1.18, 2.43, −0.91. UBC: 2.34, −3.14, −0.53, 0.09, 0.83, 2.88, −2.58. Prostate: 1.13, −2.22, −0.90, −0.66, −0.37, 0.43, −1.70. Macrophage-RCC: 2.02, −1.57, 0.70, 0.98, 1.32, 2.25, −0.24. UBC: 1.74, −2.65, −0.66, 0.05, 0.60, 2.50, −2.56. Prostate: 1.86, −4.62, −0.82, −0.54, −0.35, 0.35, −1.52. Immune checkpoint-RCC: 3.59, −1.42, −0.15, 0.18, 0.49, 1.45, −1.11. UBC: 2.27, −4.44, −0.27, 0.16, 0.68, 2.11, −1.71. Prostate: 0.85, −1.13, −0.38, −0.16, −0.04, 0.66, −1.00. (b) A comparison of PD-L1 IC, TMB and (c) percentage of CD8 in tumours. Individual samples analysed: n=224 for RCC; n=342 for UBC and n=400 for prostate. Numbers on the bars indicate the number of patients. P values of PD-L1 IC, TMB were calculated by Chi square test. P values of CD8 in tumours were calculated by 2-sided Mann-Whitney u test. Values for maxima, minima, Q1, Q2, Q3, upper whisker, lower whisker in (c): RCC: 1.38, −1.70, −0.26, 1.34, 0.45, 1.52, −1.34. UBC: 1.37, −3.10, −0.72, −0.28, 0.13, 1.40, −1.99. Prostate: 0.97, −Inf, −0.82, −0.47, −0.11, 0.95, −1.89.

Figure 5|

Pre-planned biomarker analysis: PFS in the atezolizumab + enzalutamide vs enzalutamide treatment arms. The numbers with DNA analysis were 325 and RNA analysis were 396. (a) T_eff levels (median cutoff), PD-L1 expression, CD8 levels (median cutoff), presence of DDR, and TMB levels. Median was defined by the prostate cancer repository in Foundation Medicine as TMB 4.5 mut/mb. 2.5 mut/mb is the median of TMB of the BEP of this study. The high frequency of the median created imbalance of the two arms. HRs and CIs were calculated using Cox proportional hazards regression model, and P values were two-sided, calculated using unstratified log-rank test without adjustment for multiplicity. MST refers to median survival time (PFS) in months. (b) Venn diagram shows the overlap between patients with ≥median CD8 levels in tumours and high PD-L1 IC (IC2/3). Numbers indicate the number of patients in respective groups. (c) Forest plots summarise the results of PFS for expression of genes, which show an association with outcome. The genes associated with efficacy were ordered based on HR. HRs and CIs were calculated using Cox proportional hazards regression models. (d) Unbiased analysis of PFS modelling based on pathways from Reactome. The pathways that were associated with prognosis were ordered by HRs. Immune signatures that favoured atezo were highlighted in red. Sizes of bubbles represent the two-sided P values calculated using log-rank test with no adjustment for multiplicity. Colours represent the HRs calculated using Cox proportional hazards regression models.