Phase I/II Trial of Enzalutamide and Mifepristone, a Glucocorticoid Receptor Antagonist, for Metastatic Castration-Resistant Prostate Cancer

Although androgen deprivation therapy (ADT) initially controls metastatic prostate cancer, failure of ADT and progression to castration-resistance occurs in the vast majority of patients (1). This transition to castration-resistant prostate cancer (CRPC) is an important clinical landmark that correlates with an increased risk of morbidity and death (2). Despite CRPC development, androgen receptor (AR) signaling remains a key component driving CRPC progression (3, 4). Therapeutics that more potently block AR signaling, such as the highly selective AR antagonist enzalutamide and the androgen synthesis inhibitor abiraterone acetate are established standards of care for metastatic CRPC (mCRPC; 5–9). More recently, AR signaling inhibition (ARSI) has been shown to improve outcomes earlier in the prostate cancer continuum and such inhibitors are used heavily in combination with ADT in the first line castration-sensitive setting (10–12).

Depending on therapeutic context, the duration of ARSI efficacy varies before resistance emerges; however, resistance is a near universal eventuality (13–19). Beyond potent ARSI, therapeutic options are limited; targeting specific ARSI-resistance pathways is vital to reduce death from prostate cancer.

Multiple mechanisms may explain CRPC progression, including AR splice variants, AR mutations, and ligand-independent AR activation (15–19). There are also nuclear-hormone signaling independent resistance mechanisms, with some evidence supporting the hypothesis that alternative nuclear hormone signaling pathways, such as glucocorticoid receptor (GR) signaling, may compensate for AR signaling to enable prostate cancer cell survival despite potent ARSI. The GR and AR are in the same nuclear hormone receptor family and share target DNA sequence binding homology (20, 21). Although in primary prostate cancer specimens GR expression is relatively low, recent studies demonstrated that GR expression significantly increases after ARSI (22–24). Subsequent to ARSI, GR activation promotes prostate cancer cell survival and proliferation, and GR activation conferred protection from enzalutamide-associated growth suppression (25–27). In a subset of patients with CRPC, those who developed tumors with high GR expression correlated with a poor response to enzalutamide (27). In preclinical models, GR antagonism with mifepristone, a nonselective steroidal nuclear hormone antagonist, or other more selective GR modulators delay CRPC progression in combination with ARSI (26, 27). These data suggest increased GR expression compensates for diminished AR activity in prostate cancer cells treated with ARSI and may represent a therapeutic target for progressive CRPC.

However, since mifepristone is an inhibitor of CYP2C8 and CYP3A4 (responsible for enzalutamide metabolism) it can increase enzalutamide plasma exposure when given concurrently. The safety of this combination is unknown and requires further inquiry.

We hypothesized that after potent ARSI with enzalutamide, increased GR expression and function compensates for diminished AR signaling in CRPC, facilitating cell survival and castration-resistant progression. The safe coadministration of mifepristone, a potent GR inhibitor FDA-approved for Cushing Syndrome, with enzalutamide would block this pathway and improve patient outcomes. We thus conducted a phase I/II open label trial (NCT:02012296) of study of enzalutamide combined with mifepristone to assess the feasibility and impact on disease progression with dual AR and GR antagonism. The study started with a phase I portion focused on safety and pharmacokinetics (PK), followed by a randomized phase II portion.

Eligible patients had histologically confirmed CRPC defined according to prostate cancer working group (PCWG) criteria (28). Any prior systemic therapy for prostate cancer was acceptable except CYP17 antagonists or inhibitors that block androgen production (such as abiraterone) or prior ARSIs. Eligibility included an Eastern Cooperative Oncology Group (ECOG) performance status of ≤2, acceptable bone marrow, hepatic, and renal function, and adequate baseline blood pressure and electrolytes.

The phase I portion assessed the safety of the two-drug combination along with the PK impact of mifepristone on enzalutamide exposure. The primary objective was to determine the recommended phase II dose (RP2D) of enzalutamide combined with mifepristone. Patients were treated with single agent enzalutamide at 160 mg per day for 28 days. Baseline steady-state plasma levels of enzalutamide and its M2 metabolite (N-desmethyl enzalutamide) were determined [trough drug concentration (C_trough)]. Subjects then entered the combination portion of the study, at 300 mg per day of mifepristone combined with dose-reduced enzalutamide at 40 mg per day, chosen as a conservative starting dose due to enzalutamide–mifepristone drug–drug interactions (decreased enzalutamide clearance).

Interpatient dose escalation of enzalutamide in patient cohorts of (at least) 6 patients was based on safety and PK parameters performed by InVentiv/Medivation, utilizing a constant mifepristone dose. The RP2D was determined as the mifepristone dose combined enzalutamide such that ≤33% of patients (2/6 patients per cohort) experienced dose-limiting toxicities (DLT) and the ratio of day (57/29) enzalutamide plus active metabolites was ≥0.75 and ≤1.5, along with a doubling of serum cortisol, providing support that GR was systemically antagonized. DLTs were defined as grade 3 or 4 toxicities that were potentially therapy-related. An independent safety monitor oversaw the study conduct specifically with regards to safety.

This was a multicenter randomized open-label study conducted at five sites within the Department of Defense–supported Prostate Cancer Clinical Trials Consortium (PCCTC). As GR expression was reported to increase with enzalutamide (27), and to enrich for acquired enzalutamide-resistance (as opposed to de novo resistance), patients in the phase II portion began treatment with enzalutamide 160 mg per day as a single agent for 12 weeks. This was followed by randomization, in a one-to-one ratio, to receive either enzalutamide 160 mg/day or enzalutamide + mifepristone at the RP2D. To randomize, subjects needed stable disease or better at 12 weeks of single agent enzalutamide, defined by PSA ≤ 1.25 times the PSA at the start of enzalutamide, lack of radiographic progression as defined by PCWG criteria (28), clinical stability (by treating physician), and toleration of enzalutamide 160 mg/day.

The review boards of all participating institutions approved the study which was conducted according to the Declaration of Helsinki and Good Clinical Practice guidelines of the International Conference on Harmonization. All patients signed a written informed consent before the conduct of any study procedures and after a full explanation of the study to the patient by the study investigator.

The primary phase II endpoint was whether mifepristone combined with enzalutamide prolonged time to PSA progression compared with enzalutamide alone. PSA progression was defined according to PCWG criteria (28, 29) as a PSA that is ≥1.25 times (25% increase) the PSA at randomization (week 12). Time to PSA progression was used as a pharmacodynamic biomarker of GR antagonism within CRPC tumors as activation of both the GR and the AR can drive PSA expression in prostate cancer (21, 30). Secondary objectives included evaluating the effect of mifepristone on endocrine biomarkers such as cortisol, thyrotropin, and testosterone. Additional secondary objectives included PSA RR (≥50% reduction in PSA after 12 weeks of therapy), and time to radiographic and clinical progression, all according to PCWG criteria (28, 29).

Exploratory circulating tumor cell (CTC) studies were performed on CTCs collected once after 12 weeks of enzalutamide monotherapy.

Visits occurred every 2 weeks for the first 8 weeks, then monthly. Standard blood counts, chemistries, PSA, endocrine markers, and plasma samples were regularly collected.

Baseline and on-study electrocardiogram were obtained to monitor for QTc prolongation. Disease burden at baseline and every 12 weeks was assessed with standard nuclear medicine bone scans and abdominal/pelvic cross-sectional imaging.

Enzalutamide or enzalutamide + mifepristone continued until progression of disease was noted by PCWG criteria (28, 29). PSA was measured monthly, but did not determine study drug termination.

All CTC studies were performed centrally by Epic Sciences (Supplementary Fig. S1A). CTC identification was performed using Epic Sciences' CTC-specific platform for GR expression (31). Blood collected in Streck^TM tubes was shipped overnight to Epic Sciences. Nucleated cells from the blood were plated onto glass microscopy slides, fixed, and bio-banked at −80°C until analysis. Biomarker expression studies were performed on four slides per patient; two slides for GR. Two slides correspond to the analysis of 6 x 10⁶ nucleated cells within the blood draw. Each assay respectively stained for pan-cytokeratin (CK), CD45, and DAPI in addition to GR. A CTC is defined as any CK⁺, CD45⁻, DAPI⁺ cell. After staining the slide, each nucleated cell was imaged using high-throughput florescence microscopy, and CTCs were identified using Epic Sciences' proprietary digital pathology algorithms. Candidate CTCs were then confirmed by trained human technicians. The final cell counts in each subgroup (GR^±) were then tabulated.

For GR evaluation (Supplementary Fig. S1A), the mAb specific to the GR (D6H2L, Cell Signaling Technology; Rabbit IgG, catalog no. 12041) C-terminal domain was utilized. If a patient had greater than 0 detectable CTC per milliliter of tested blood, they were classified in a binary fashion as “CTC-positive.” The threshold for a CTC to be positive for GR is not formally known, and it is not known to what extent GR expression correlates with activity, therefore cell-line cells spiked into healthy donor blood from DU145 [GR-positive (GR⁺); RRID: CVCL_0105) and LNCaP [GR-negative (GR⁻)] were analyzed in parallel. For each cell detected on the slide the mean fluorescence intensity from the GR antibody detection was recorded and classified cells as GR⁺ if it had GR expression detectable above background fluorescence within identified CTCs. Otherwise, cells were classified GR⁻.

Phase II enzalutamide was supplied by Astellas and Medivation (now Pfizer) as 40-mg capsules, 120 capsules per bottle. Mifepristone was provided by Corcept Therapeutics as 300-mg tablets, 30 tablets per bottle. Enzalutamide and mifepristone were taken concurrently once daily.

In phase I, steady-state C_trough for enzalutamide and its metabolite were determined after 28 days of enzalutamide alone and after an additional 28 days of combination therapy. The mean of the day 57/29 ratio of trough concentrations for enzalutamide and its metabolite were calculated for each dosing cohort and used for dose adjustment determinations in the following cohort. SD and range of the C_trough and ratio were calculated.

Randomization for phase II was done at the University of Chicago (Chicago, IL) in a 1:1 block fashion, using block sizes of 4, 6, and 8. There was no specific stratification as the 12-week standard-of-care enzalutamide lead-in was hypothesized to add homogeneity to the population. For the randomized phase II study, the primary endpoint was PSA progression-free survival (PFS) postrandomization, defined as time to PSA progression or death, whichever came first. We assumed median time to PSA progression in the control (enzalutamide alone) arm would be 6 months. This was based on the phase III enzalutamide trial AFFIRM (13), in which approximately 10% of treated patients had PSA progression prior to 12 weeks, the PSA PFS was approximately 42% at 9 months, corresponding to a 6-month rate, conditional on no event at 12 weeks, of 0.42/0.9 = 0.47. Thus, our enrichment strategy was projected to lead to a patient population with a median PFS of 6 months (24 weeks) in the control arm postrandomization. To detect an HR of 0.60, corresponding to an increase in the median from 24 to 40 weeks, with 80% power, a sample size of 84 patients (42/arm) was required, using a one-sided test at the α = 0.10 significance level. This assumed a 2-year accrual period and a subsequent 1-year follow-up period. Since it was anticipated that approximately 10% of enrolled patients in phase II would not be randomized due to progression or intolerance, 92 patients were to be enrolled. An interim futility analysis was conducted after 50% of PSA progression events were observed. A conditional power for the primary endpoint of less than 25% at this analysis would lead to study termination for futility. As shown in Results, using the protocol-defined PSA criteria above, futility was reached and accrual to the trial was terminated early. For subsequent analyses, the PSA progression endpoint was modified to conform with PCWG criteria to require both a 25% increase from baseline or nadir and an absolute 5-ng/mL increase. As will be seen, the conclusions were unaltered.

For time to PSA progression, Kaplan–Meier (32–35) curves were generated for the two treatment arms and compared using a log-rank test. Median time-to-event in each group was estimated along with 90% confidence intervals using the method of Brookmeyer and Crowley (34). Radiographic PFS was analyzed with Kaplan–Meier curves. For patients without progression, PFS was censored at the date of the patient's last assessment.

Relative PSA change from baseline within each arm was also reported using a waterfall plot (28). Adverse events were summarized by grade and type. Group comparisons used the χ² or Fisher exact tests.

We also compared endocrine pharmacodynamic marker differences between the enzalutamide-alone versus the enzalutamide + mifepristone treatment arms. In previous placebo-controlled trials with mifepristone, serum cortisol levels reliably doubled from a baseline of approximately 15 μg/dL to more than 30 μg/dL (38, 39). Based on the reported interquartile ranges and assuming normality of the distribution, the coefficient of variation was estimated at 50%. At time of interim analysis, the difference in cortisol levels was analyzed. A lack of biomarker effect could provide justification for closing the trial. The PK data was obtained as above and were summarized using standard descriptive methods (means, SDs, medians, and ranges).

The primary objective of the CTC correlative study was to assess intra- and interpatient variability in GR within CTCs from patients with progressive CRPC. The secondary objective of this aim was to explore the correlation between baseline GR expression and PSA PFS in patients treated with enzalutamide ± mifepristone. The components of variability were estimated using analysis of variance. Finally, Cox (33) regression models for time to PSA progression were fit using GR expression as a binary covariate and incorporating treatment-by-marker interaction terms to determine the marker's prognostic and/or predictive value.

Six patient dose cohorts escalated through the pre-planned dose combinations of mifepristone and enzalutamide (see Fig. 1). During the phase I portion of the study, patients could potentially receive mifepristone doses of either 300 mg daily or 300 mg every other day combined with enzalutamide at either 40, 80, 120, or 160 mg daily. Per protocol guidelines, the starting dose for the first cohort was 40 mg of enzalutamide and 300 mg of mifepristone daily. PK analyses were then performed by InVentiv/Medivation. The first PK levels of enzalutamide and its main M2 metabolite were then taken after 28 days of enzalutamide at standard dosing to reach a steady state with less variability of enzalutamide drug levels, after which time combination dosing with mifepristone occurred to find dose escalation cohorts. After 28 days of combination, another trough PK was taken of both enzalutamide and its active metabolite (M2). A ratio of trough drug concentration after 28 days of combined dosing (day 57 of the study) divided by trough drug concentrations after 28 days of enzalutamide-alone dosing (day 29 of the study) was then performed, defined as r_i. An average of r_i for all patients within a dose cohort (i.e., mean of 6 or 12 patients) was then calculated, defined as ř. For ř > 1.5, enzalutamide was then decreased by one level (e.g., 80 mg daily to 40 mg daily) and the patient reentered the dose escalation algorithm. If ř < 0.75, enzalutamide was then increased by one level (e.g., 40 mg daily to 80 mg daily) as below and the patient reentered the dose escalation algorithm. If ř ≤ 1.5, ≥ 0.75, then mifepristone dose changes were determined by the rate of DLT. DLT was defined as any grade 3 or 4 toxicity that was potentially related to the therapy. There was a 28-day DLT monitoring period after combination treatment during the phase I study in which safety was examined prior to dose escalation. A 28-day DLT period was chosen as the majority of DLT seen with enzalutamide, including rare seizures, were seen in previous phase I/II studies within 6 weeks of beginning enzalutamide (patients had been on enzalutamide for 4 weeks prior to combination dosing already). The DLT period therefore began at day 29, upon the onset of concurrent mifepristone and enzalutamide drug daily administration, and ended at day 57, 4 weeks later. If ř ≤1.5 and ≥0.75 mifepristone was escalated one level and the patient reentered the dose escalation algorithm provided the DLT rate was less than 33%. The decision to escalate mifepristone was made by investigators and an independent safety monitor after evaluation of toxicity data from the current dose level. If the DLT rate was greater than 33% (>2 in 6 patients per cohort), then it was concluded the maximum tolerated dose (MTD) of mifepristone was exceeded and the prior dose level was determined to be the MTD which was then expanded to 12 patients. If the MTD was exceeded with mifepristone 300 mg every day, the dose was reduced to 300 mg every other day. The RP2D was defined as the highest dose tested where at most 2 of the 6 patients developed DLT during the first cycle of treatment and ř ≤ 1.5.

The data generated in this study are available upon request from the corresponding author.

As above, the clinical trial had a two-part study design with phase I focused on safety and PK followed by phase II. Between January 2014 and January 2019, 140 total patients were assessed for trial eligibility; 106 patients were ultimately accrued (Fig. 1). The most common reason for not meeting eligibility were (N, %) prolonged QTc (7, 21%) and elevated liver function tests (2, 6%).

Eighteen patients were enrolled in phase I dose escalation and treated in three combination dosing cohorts. These include (i) enzalutamide 40 mg and mifepristone 300 mg, (ii) enzalutamide 80 mg and mifepristone 300 mg, and (iii) enzalutamide 120 mg and mifepristone 300 mg. Six patients were in each dosing cohort. Demographics were consistent with a general metastatic prostate cancer population (Table 1).

Eighty-eight patients were enrolled in phase II (Fig. 1). After the 12-week enzalutamide lead-in, 66 patients ultimately underwent randomization. Thirty-three patients continued to receive enzalutamide alone (160 mg/day), while 33 patients were randomized to receive the RP2D of 120 mg/day enzalutamide and 300 mg/day mifepristone (enzalutamide + mifepristone). Fifteen (17%) patients were not randomized due to progressive disease prior to 12 weeks, 2 (2%) patients not randomized due to adverse events, and 5 (6%) not randomized due to study closure during their 12-week enzalutamide lead-in. One patient was excluded from postrandomization analysis due to a protocol deviation of combined treatment initiation. Patient characteristics were well balanced between the two phase II groups (Table 1). Despite the study being written anticipating postdocetaxel enzalutamide, most patients received treatment in the predocetaxel setting (34% of patients received docetaxel previously).

Patients who failed to randomize after the 12-week enzalutamide lead-in had a higher proportion of prior docetaxel therapy and a higher percentage of visceral disease relative to those randomized.

The main PK outcomes are summarized in Fig. 2. The day 57/day 29 PK trough ratio of enzalutamide and metabolites concentration (Fig. 2) for cohort 1, enzalutamide 40 mg and mifepristone 300 mg, was 0.6 (SD: 0.2, range: 0.4–0.8). Per protocol, the dose of enzalutamide was then escalated in the second cohort to enzalutamide 80 mg, mifepristone 300 mg. Day 57/29 mean PK ratios were still suboptimal per protocol at 0.7 (SD: 0.1, range: 0.5–0.8). Cohort 3 was then dosed enzalutamide 120 mg, mifepristone 300 mg and had a mean day 57/29 PK ratio of 1.0 (SD: 0.2, range: 0.8–1.2), reaching the protocol goal of 0.75 to 1.5 as being acceptable for phase II.

To ensure adequate GR antagonism, we assessed change in cortisol after 4 weeks of mifepristone treatment. Serum cortisol levels were measured prior to enzalutamide initiation, after 28 days of enzalutamide monotherapy, and then at day 57 after randomization to combination dosing with mifepristone. Cortisol routinely doubled as expected after combination dosing with mifepristone for each cohort (Supplementary Fig. S2).

Study drug–attributable adverse events in at least 15% of patients in any trial phase are seen in Table 2. Overall, the combination of mifepristone + enzalutamide was generally well-tolerated. There were no DLTs in any cohort and no study drug–attributable grade 4 or 5 adverse events. In phase I the most common all-grade side effects were fatigue (72%), anorexia (28%), and diarrhea (28%). The only notable study drug–attributable grade 3 side effects were fatigue (6%), cognitive disturbance (6%), and hypertension (6%). There were no significant differences in adverse events when comparing the various enzalutamide dosing cohorts.

In the phase II study, the combination of mifepristone + enzalutamide was well tolerated with minimal significant differences associated with the addition of mifepristone (study drug–attributable adverse events listed in Table 2). When comparing the enzalutamide-alone arm to enzalutamide + mifepristone arm, the most common any-grade side effects were fatigue (76% vs. 76%), hot flashes (45% vs. 45%), anorexia (12% vs. 33%; P > 0.05), and generalized muscle weakness (6% vs. 24%; P > 0.05). Grade 3 fatigue was higher in the enzalutamide + mifepristone arm (12%) as compared with the enzalutamide-alone group (3%). Like phase I, there were no grade 4 or 5 attributable adverse events. There were very few treatment discontinuations due to adverse events (2 patients in the enzalutamide arm, 1 patient in the enzalutamide + mifepristone arm).

PSA PFS was the primary endpoint. Per protocol, an interim futility analysis was performed after 50% of the planned PSA-progression events (35 events). At this analysis the HR (enzalutamide + mifepristone/enzalutamide alone) was 1.34 in favor of the control arm and the conditional power for finding a benefit to combination therapy if the trial continued was only 0.12. Based on lack of efficacy at interim analysis, the study was closed to accrual.

As described under Methods, additional analysis using a more stringent definition of PSA progression was performed. In this analysis, which included 26 events, median PSA PFS after randomization was 20.8 months in the enzalutamide-alone arm compared with 16.5 months in the enzalutamide + mifepristone arm. Cox regression analysis yielded an HR of 1.09 when comparing the two arms (log-rank P = 0.83; Figure 3A).

With respect to PSA response, the bulk of the PSA change occurred within the first 12 weeks of the initial enzalutamide lead-in (Supplementary Fig. S3A). The prerandomization PSA RR (defined as PSA decline >50%) was 81% in those subsequently randomized to enzalutamide-alone group and 72% in the enzalutamide + mifepristone group. Very few patients eligible for randomization experienced primary PSA progression (posttreatment PSA increase as best response during the lead-in: 2 patients in the enzalutamide + mifepristone group, 1 in the enzalutamide-alone).

The maximum percentage decrease in PSA that patients achieved after randomization to either enzalutamide-alone or enzalutamide + mifepristone at week 12 was similar comparing the two treatment arms (Supplementary Fig. S3B). The mean postrandomization decrease in PSA change to best response was −30.3% in the enzalutamide arm and −28.8% in enzalutamide + mifepristone arm (P = 0.86). Nine patients (28%) in each arm had a postrandomization PSA decline more than 50%. With respect to radiographic response, 4 of 33 patients in each arm had documented RECIST response.

In sum, these data suggest that the addition to mifepristone to enzalutamide did not improve PSA-PFS or response.

Time to radiographic progression was a secondary endpoint of phase II. Patients that came off study for either physician discretion, PSA progression, toxicity or data lock were censored at the time taken off study. There were no significant differences in time to radiographic progression between the two arms (Fig. 3B). Median radiographic PFS was 16.5 months in the enzalutamide + mifepristone group versus not reached in the enzalutamide-only group (HR = 1.70; P = 0.22).

The goal of the CTC analysis was to test whether the presence of GR⁺ CTCs would predict benefit from the addition of mifepristone to enzalutamide. Week 12 blood samples drawn immediately prior to randomization were available for CTC analyses on 24 of 33 patients in the enzalutamide + mifepristone group (73%), 29 of 33 in the enzalutamide-alone group (88%), and 6 of 17 (35%) of patients who did not proceed to randomization after the initial enzalutamide lead-in. The majority (68%) of randomized patients were noted to have detectable CTCs compared with 6 of 6 (100%) of patients in the group that did not proceed to randomization (P = 0.17). A higher percentage of patients in the enzalutamide group (79%) had detectable CTCs at the time of randomization, compared with (54%) of those randomized to enzalutamide + mifepristone arm (P = 0.08). Adjusting the treatment arm comparison of PSA progression for the presence of CTCs, the HR decreased from 1.09 to 0.93, but remained nonstatistically significant (P = 0.87). Consistent with prior data (31), patients with detectable CTC at time of randomization were less likely to have had a PSA response (Supplementary Fig. S1B) and more likely to have worse PSA-PFS (Supplementary Fig. S1C).

Amongst the entire study population, 30 of 42 (71%) patients with detectable CTCs had GR⁺ CTCs at week 12. A numerically higher percentage of patients who were not eligible to randomize had GR⁺ CTCs (5/6; 83%) compared with 25 of 36 (69%) randomized patients (Supplementary Fig. S1D). The incidence of GR⁺ CTC was similar in the two arms (74% enzalutamide alone, 62% enzalutamide + mifepristone, as expected from randomization) and the presence of GR⁺ CTC was not, contrary to the study hypothesis, a predictor of prolonged PSA-PFS (HR = 1.15; P = 0.84; Supplementary Fig. S1E). In the converse analysis of PSA-PFS in patients who, at week 12 had GR⁻ CTCs or lacked CTCs, there was no difference between arms (HR = 0.69; P = 0.59).

As observed in phase I, serum cortisol was expected to increase by as much as double after treatment with mifepristone (36, 37). The ratio of week 16 to week 12 cortisol was 1.45 (0.47–2.43; enzalutamide alone) versus 2.40 (1.90–2.89; enzalutamide + mifepristone) indicating cortisol nearly doubled after introducing mifepristone, as expected (P = 0.06). Given positive skewness, a logarithmic comparison of week 16 with week 12 cortisol demonstrated strong statistical significance (P = 0.0002).

Prior studies demonstrated that mifepristone markedly increases serum androgen levels in castrate patients (32). This increase was thought secondary to mifepristone's inhibition of GR resulting in an increase in adrenocorticotropic hormone leading to an increase in adrenal androgen production. As shown in Fig. 4, testosterone increased slightly after 12 weeks of enzalutamide lead-in, but significantly increased with mifepristone after randomization. The average postrandomization testosterone was 16.0 ng/dL in the enzalutamide-alone group versus 38.8 ng/dL in the enzalutamide + mifepristone group (P = 0.0012). Therefore, mifepristone led to a nearly 2.5× increase in androgen production for the enzalutamide + mifepristone population.

This was the first reported randomized clinical trial to test the hypothesis that continuous GR pathway inhibition would be safe and improve outcomes in combination with ARSI for mCRPC. We found that a 25% dose reduction in enzalutamide, when added to mifepristone, resulted in equivalent drug levels compared with full-dose enzalutamide. Although generally well-tolerated, the addition of mifepristone to enzalutamide following a 12-week enzalutamide lead-in did not delay time to PSA or radiographic PFS.

Of note, this study demonstrated a median PSA PFS after randomization of 20.8 months in the enzalutamide-alone arm and 16.5 months in the enzalutamide + mifepristone arm. This implies a total median PSA PFS of 23.8 months with enzalutamide, considerably longer than the median time to PSA progression seen in prior studies (13, 38). However, our study did not allow patients who progressed on enzalutamide monotherapy in the first 12 weeks to proceed to randomization. By focusing on such “nonprogressors” we were selecting for patients with significantly above-average PSA PFS from enzalutamide.

This trial had several strengths. While there is significant preclinical evidence that the GR contributes to enzalutamide resistance (23, 26, 27, 39, 40), this is the first ever prospective randomized study to study combined GR and AR antagonism in a CRPC clinical population. In addition, our study was multi-institutional, involving over 100 patients with CRPC across four cancer centers throughout the United States. Our study provides some of the first clinical evidence of the therapeutic combination of GR modulators with AR antagonism in prostate cancer. Furthermore, our phase I study was able to demonstrate how safe, pharmacologically guided dosing of enzalutamide combined with a CYP3A4, CYP3C8/9 inhibitor could achieve equivalent therapeutic effect. Given enzalutamide is associated with increased risk of seizures at higher plasma concentrations and may be less effective at lower blood doses, careful attention to PK is of paramount importance (41, 42). This study can be used to suggest how to safely combine enzalutamide with other agents that have PK properties like mifepristone. Furthermore, this study was the first to show that long-term GR blockade, at a pharmacologically active dose, along with ARSI would be safe or tolerable (43).

A unique feature of this trial was that it was the first ever to utilize the interrogation of CTCs to analyze the GR as a biomarker in a prospective CRPC trial. The goal of this analysis was to determine if GR expression could be a predictive biomarker for mifepristone efficacy. Our study did show that patients with CTCs had a less robust response and a shorter PSA PFS from enzalutamide regardless of the treatment they ultimately received, consistent with prior data validating pretreatment CTCs as a prognostic biomarker in men starting either abiraterone or enzalutamide (31). However, our study did not demonstrate that the presence of GR⁺ CTCs was predictive of a specific lack of response to enzalutamide, nor that individuals with GR⁺ CTCs benefited from GR directed therapy with mifepristone.

Our study had several shortcomings. First, the phase II portion was open-label and lacked a placebo-controlled arm. Secondly, the primary endpoint of phase II was PSA PFS, chosen since PSA is a pharmacodynamic biomarker of nuclear hormone activity. Instead, radiographic PFS may have better reflected clinical status. Finally, the definition of GR positivity in CTCs is not well established. We took a binary approach of classifying CTCs as GR⁺ should they exhibit any expression beyond background. There are potential pitfalls to this approach. GR expression is likely on a spectrum. Focusing instead on only high GR expression (e.g., highest decile) rather than a binary approach may better reflect GR positivity. GR expression may also not correspond with downstream pathway activation. Future studies could use paired biopsies as more accurate correlative markers of GR activation. A more nuanced view of GR expression within progressive mCRPC may better support GR's utility as a therapeutic biomarker. Additionally, our trial collected CTCs only at randomization with the hypothesis that GR⁺ CTC would predict benefit of dual antagonism. CTCs at baseline, randomization, and end of study could determine if mifepristone helped clear GR⁺ CTCs. Future studies could collect CTCs at more time points.

Although mifepristone did not improve PSA PFS, this trial does not refute the hypothesis that GR antagonists have a benefit in combination with ARSI in prostate cancer. The exact mechanisms of castration resistance and subsequent resistance mutations is an active area of investigation. We may have randomized patients to mifepristone after too long a period of enzalutamide exposure. Our trial design originally posited that GR expression would be enriched after 12 weeks of enzalutamide monotherapy. Emerging evidence suggests that GR enrichment may occur much earlier in prostate cancer (44). A recent study of androgen biosynthesis inhibition in the preoperative setting demonstrated that GR enrichment may occur as early as the neoadjuvant castrate-sensitive stage (44). Conducting this study in the CRPC setting after an additional 12 weeks of enzalutamide may have inadvertently enriched for intrapatient pleotropic ARSI-resistance mechanisms beyond GR. Prior to becoming castration-resistant, prostate cancer disease biology is likely more homogenous. Introducing GR inhibition in an earlier clinical stage before the activation of adaptive cell survival mechanisms (such as p53 mutations or Rb loss) may have therapeutic merit.

AR mutations (such as AR-amplification, AR splice variant expression, and aberrant AR co-regulator activities) have been implicated in castration resistance and could have also contributed to the study's failure (45, 46). Despite randomization, the trial may have had an imbalance in such other AR mutations, (e.g., AR-v7) between the two arms.

Although mifepristone is a potent GR antagonist it has several characteristics that may have made it a suboptimal GR antagonist for prostate cancer therapy. It is a nonselective steroidal nuclear hormone receptor antagonist that binds other NR3C family members including AR and the progesterone receptor. Modulation of these other receptors potentially led to unintended consequences. For example, in the setting of mutated AR ligand binding domains (LBD), other steroidal compounds can ligand and activate LBD-mutant AR. There are reports that mifepristone can activate AR (depending on AR LBD mutations and the mifepristone dose used; refs. 47, 48) blunting our therapeutic approach. Future studies should study the evolution of LBD mutations in the context of GR antagonism.

Mifepristone can also raise testosterone levels (32). Our study showed that testosterone levels significantly increased with mifepristone, potentially hindering enzalutamide's therapeutic effect in the enzalutamide + mifepristone arm. Highly specific nonsteroidal GR antagonists may not raise testosterone levels and thus may be more effective in prostate cancer (39, 49). While mifepristone may have been a good starting point given its broad nuclear hormone receptor antagonism and clinical availability, phase I studies of more selective agents like relacorilant (NCT 03674814) or exicorilant (NCT 03437941) combined with enzalutamide are ongoing, and can be used in future dual AR-GR antagonism prostate cancer studies.

While we achieved systemic cortisol receptor blockade with mifepristone as demonstrated by the increase in cortisol subsequent to mifepristone, other studies have shown that local tumor glucocorticoid levels stimulate the GR and contribute to enzalutamide resistance (50). Given mifepristone binds and antagonizes GR it is unlikely that upstream glucocorticoid production would be sufficient to overcome mifepristone; however, it is possible that very high local glucocorticoid production outcompeted mifepristone for the GR. Future studies with imbedded tissue correlative studies can help answer whether this is a major contributing factor underlying insufficient GR antagonist activity.

In conclusion, this is the first prospective clinical trial reported of dual AR-GR antagonism in CRPC. Daily dosing of enzalutamide combined with mifepristone was safe and well tolerated. The development of more specific GR antagonists, combined with AR antagonists, studied in an earlier stage more homogenous population may lead to more effective future therapeutic regimens. Additional work should clarify which biomarkers can identify patients who would benefit most from this approach.

E.I. Heath reports other support from University of Chicago during the conduct of the study; other support from Pfizer, Astellas, Janssen, and Bayer HealthCare; and personal fees from Janssen and Bayer HealthCare outside the submitted work. J. Schonhoft reports other support from Epic Sciences during the conduct of the study. T. Karrison reports grants from NCI during the conduct of the study. W.M. Stadler reports personal fees from AstraZeneca, Bayer HealthCare, Eisai, Merck, Pfizer, Treadwell Therapeutics, Calico Life Sciences, Caremark/CVS, and EMA Wellness outside the submitted work. R.Z. Szmulewitz reports nonfinancial support from Astellas/Pfizer and Corcept Therapeutics during the conduct of the study; personal fees from Astellas/Pfizer, Merck, Janssen and AstraZeneca outside the submitted work; in addition, R.Z. Szmulewitz has a patent for US9801893B2 issued and licensed to Corcept Therapeutics. No disclosures were reported by the other authors.

A.V. Serritella: Conceptualization, resources, data curation, formal analysis, validation, methodology, writing–original draft, project administration, writing–review and editing. D. Shevrin: Investigation, writing–review and editing, patient enrollment. E.I. Heath: Investigation, patient enrollment. J.L. Wade: Investigation, patient enrollment. E. Martinez: Resources, investigation. A. Anderson: Formal analysis, writing–review and editing. J. Schonhoft: Formal analysis, writing–review and editing. Y.-L. Chu: Formal analysis, writing–review and editing. T. Karrison: Formal analysis, writing–review and editing. W.M. Stadler: Supervision, investigation, writing–review and editing. R.Z. Szmulewitz: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

Support for the clinical trial was provided by the Department of Defense CDMRP W81XWH-14-1-0021 and the CTC correlatives were supported by Prostate Cancer Foundation-Movember Challenge Award. Biostatistics and clinical trial office supported by University of Chicago NCI Cancer Center Support Grant (5P30CA014599-46). This work was supported by grants to R.Z. Szmulewitz.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.