Background

Non-small cell lung cancer (NSCLC) represents ~ 84% of all lung cancers [1]. The majority of patients have metastatic disease at diagnosis [1, 2]. Chemotherapy combined with pembrolizumab, an immune checkpoint inhibitor that inhibits programmed cell death protein 1 (PD-1), is an efficacious option for first-line treatment of advanced NSCLC lacking targetable mutations [3, 4]. However, NSCLC cells evade immune surveillance via multiple mechanisms, so approaches targeting multiple immune pathways may provide enhanced efficacy in this setting.

Indoleamine 2,3-dioxygenase 1 (IDO1), an enzyme that catalyzes the breakdown of tryptophan to kynurenine, is constitutively expressed by a wide variety of human tumor cell types as well as by dendritic cells that localize to tumor-draining lymph nodes [5, 6]. Some NSCLC tumors co-express programmed death-ligand 1 (PD-L1) and IDO1 [7,8,9]. Overexpression of IDO1 has been shown to result in suppression of T cell-mediated immune responses [5, 10]. IDO1, which is induced by interferon, can also suppress inflammatory responses targeting tumor cells [5, 11,12,13,14]. Thus, inhibition of IDO1 may help restore an effective antitumor response and may also augment checkpoint inhibitor activity by reducing resistance to anti-tumor inflammatory responses.

Epacadostat is a potent selective inhibitor of IDO1 [15, 16] that has been shown to normalize plasma kynurenine levels in patients with advanced solid tumors at twice-daily (BID) doses ≥ 100 mg as monotherapy [16]. The combination of epacadostat and pembrolizumab has been investigated in NSCLC and other types of advanced solid tumors. In the phase I/II ECHO-202/KEYNOTE-037 study of epacadostat plus pembrolizumab in patients with advanced tumors, promising responses were observed in melanoma (ORR 60.5%) and NSCLC with PD-L1 tumor proportion score (TPS) < 50% (ORR 24.4%), and treatment was generally well tolerated [17]. We present results from the final analysis of a phase II randomized study (NCT03322566) assessing the safety and efficacy of epacadostat plus pembrolizumab with platinum-based chemotherapy (E + P + C) vs. placebo plus pembrolizumab with platinum-based chemotherapy (PBO + P + C) in patients with metastatic NSCLC.

Methods

Study design and conduct

ECHO-306/KEYNOTE-715 was an active comparator, partial double-blind, parallel-group, multicenter randomized phase II study. The study was originally planned as a phase III study. However, after the study was initiated, it was amended to a phase II study design by a protocol amendment based on data from the ECHO-301/KEYNOTE-252 study in unresectable/metastatic melanoma that showed that epacadostat 100 mg BID in combination with pembrolizumab did not improve the primary endpoint of progression-free survival (PFS) compared with pembrolizumab monotherapy [18]. Other changes in study methods based on this protocol amendment are described in their respective sections below.

This study was conducted in accordance with the Declaration of Helsinki, Good Clinical Practice guidelines, applicable country and/or local statutes, regulations regarding independent ethics committee review, and the protection of human patients in biomedical research. All patients provided written informed consent before initiating treatment.

Study population

Adults ≥ 18 years of age with confirmed stage IV NSCLC not indicated for EGFR-, ALK-, or ROS1-directed therapy, measurable disease per Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1), life expectancy ≥ 3 months, Eastern Cooperative Oncology Group (ECOG) performance status 0 or 1 and adequate organ function based on laboratory values were eligible for enrollment. Exclusion criteria included prior systemic chemotherapy or other targeted/biological antineoplastic therapy for metastatic NSCLC; prior treatment with any anti-PD-1, anti-PD-L1, anti-PD-L2 or anti-IDO1 agent or with an agent directed to another stimulatory or co-inhibitory T-cell receptor; radiotherapy ≤ 14 days or lung radiation therapy > 30 Gy ≤ 6 months before the first study treatment dose; systemic steroid therapy ≤ 7 days prior to the first dose of study treatment or any other form of immunosuppressive medication; history of serotonin syndrome after receiving serotonergic drugs; history of (non-infectious) pneumonitis requiring systemic steroids or current pneumonitis/interstitial lung disease and untreated central nervous system (CNS) metastases and/or carcinomatous meningitis. Tumor PD-L1 expression levels were determined using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies, Carpinteria, CA).

Study procedure and interventions

In the original design, patients were randomized to the E + P + C, PBO + P + C or open-label E + P treatment arms and stratified by PD-L1 TPS (< 50% vs. ≥ 50%) and tumor histology (squamous vs. non-squamous). In the original study design, enrollment of 1062 patients was planned. The open-label E + P group was closed by protocol amendment 05, which also changed the design to a phase II study, and the target was reduced to 148 planned randomized patients into the existing two blinded treatment arms (E + P + C and PBO + P + C). Patients already randomized to the E + P group at amendment 5 were allowed to continue study treatment if they were deriving clinical benefit, but were not included in the primary efficacy analysis. The efficacy analysis population consisted of all patients randomized to a blinded study intervention (i.e., the intent-to-treat population), and the safety analysis population consisted of all patients as treated who received ≥ 1 dose of study intervention.

Pembrolizumab 200 mg was administered intravenously (IV) every 21 days (Q3W) and epacadostat 100 mg or matching placebo orally BID for up to 35 3-week cycles. Epacadostat could be reduced to 50 or 25 mg BID to mitigate immune-related adverse events (AEs). Platinum-based chemotherapy consisted of one of three regimens selected by the investigator prior to randomization. For non-squamous tumor histology, pemetrexed 500 mg/m2 IV with either cisplatin 75 mg/m2 IV or carboplatin area under the curve (AUC) 5 mg/mL/min IV was given Q3W for four cycles, followed by pemetrexed 500 mg/m2 IV Q3W for up to 35 cycles. For squamous tumor histology, paclitaxel 175 − 200 mg/m2 IV + carboplatin AUC 5 − 6 mg/mL/min IV was given Q3W for four cycles.

Fasted patients had blood drawn before dosing on day 1 of cycle 1 (C1D1) and day 1 of cycle 2 (C2D1). Serum kynurenine levels were determined by a proprietary, validated liquid chromatography-tandem mass spectrometry assay using calibrated standards at Worldwide Clinical Trials, Morrisville, NC.

Study objectives and endpoints

The original primary endpoints were overall survival (OS) and PFS. The primary endpoint was changed to ORR when a protocol amendment changed the study from a phase III to a phase II study, with the primary objective of the study comparing the ORR of combination E + P + C versus PBO + P + C. ORR was defined per RECIST v1.1 based on confirmed response by blinded independent central review (BICR). The secondary objectives were to compare PFS, OS, duration of response (DOR) and safety/tolerability of E + P + C versus PBO + P + C. AEs were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) version 4.0. Assessment of pharmacodynamic activity of epacadostat based on changes in circulating kynurenine levels from baseline was among the exploratory objectives.

Statistical analyses

ORR was compared using the Miettinen and Nurminen method [19] stratified by PD-L1 TPS (< 50% vs ≥ 50%) and tumor histology (squamous vs non-squamous). Because of the small sample size, the strata ‘PD-L1 TPS ≥ 50 percent non-squamous’ and ‘PD-L1 TPS ≥ 50 percent squamous’ were combined into one stratum for the primary analysis. Based on the assumption of 74 patients randomized per arm with at least 12 weeks of follow-up, the study had 81.7% power to detect a clinically meaningful (ie, 20 percentage-point difference) in ORR between the E + P + C and PBO + P + C groups at α = 5% (one-sided). PFS and OS were compared using a stratified log-rank test. Hazard ratios (HRs) were estimated using a stratified Cox regression model with Efron’s method of tie handling. Median PFS, median OS, PFS rates and OS rates were estimated with the Kaplan–Meier method. Circulating kynurenine levels within each treatment group were compared using paired t-tests.

Results

Patient characteristics

This study was conducted from January 09, 2018—October 16, 2020. A total of 411 patients were screened and 233 met inclusion criteria and were randomized to E + P + C (n = 91), PBO + P + C (n = 87) or E + P (n = 55) (Fig. 1). The E + P + C and PBO + P + C treatment groups were generally well balanced; more patients had stage M1C tumors in the E + P + C group, and more patients had stage M1B tumors in the PBO + P + C group (> 10% difference) (Table 1). Most patients in both groups were former smokers with predominantly non-squamous tumor histology, PD-L1 TPS < 50% and did not have baseline brain metastasis.

Fig. 1
figure 1

Patient disposition AE adverse event; PD progressive disease

Full size image
Table 1 Patient and disease characteristics
Full size table

Treatment duration

Compared with the PBO + P + C group, the E + P + C group had less exposure to study medication, with a lower median number of days on treatment, fewer cycles on treatment, fewer days on epacadostat and fewer patients completing all four initial chemotherapy cycles (Supplementary Table 1). The median duration of follow-up was 5.1 months (range 0.1–10.4) in the E + P + C group and 7.2 months (range 1.4–10.4) in the PBO + P + C group.

Efficacy

The open-label E + P group was excluded from the efficacy analysis. Based on BICR assessment, superiority of ORR for the E + P + C group compared with the PBO + P + C group could not be claimed (Table 2). The E + P + C group had a lower confirmed ORR (26.4%; 95% confidence interval [CI] 17.7–36.7) than the PBO + P + C group (44.8%; 95% CI 34.1–55.9), with a difference of − 18.5% (95% CI − 32.0 – (− 4.3); one-sided P = 0.9948). For PD-L1 TPS < 50%, ORR was 27.8% (95% CI 17.9–39.6) and 40.9% (95% CI 29.0–53.7) in the E + P + C and PBO + P + C groups, respectively. For PD-L1 TPS ≥ 50%, ORR was 21.1% (95% CI 6.1–45.6) and 57.1% (95% CI 34.0–78.2) in the E + P + C and PBO + P + C groups, respectively. No complete responses were observed and the proportion of patients with best response of stable disease was similar in the E + P + C and PBO + P + C groups (41.8%; 95% CI 31.5–52.6 vs. 40.2%; 95% CI 29.9–51.3). The disease control rate was 68.1% (95% CI 57.5–77.5) in the E + P + C group and 85.1% (95% CI 75.8–91.8) in the PBO + P + C group. The open-label E + P treatment arm was terminated by the protocol amendment described above; although efficacy data of the open-label E + P group were not compared with the other two groups, the ORR for this group was 20.0% (95% CI 10.4–33.0).

Table 2 Summary of objective response
Full size table

Because the prespecified success criterion for the primary ORR hypothesis was not met, the study was unblinded, all epacadostat and placebo administration was stopped (all remaining patients had continued access to pembrolizumab), and the subsequent efficacy hypotheses were not formally tested. The median DOR was not reached (range 1.1 + to 7.0 +) in the E + P + C group and was 7.0 months (range 1.2 + to 8.0 +) in the PBO + P + C group (plus symbols [ +] indicate no progressive disease by the time of last disease assessment); the median TTR was similar for the E + P + C and PBO + P + C groups (Table 2). Among confirmed responders, similar proportions of patients had ongoing responses in both groups. However, the E + P + C group had a numerically higher percentage of confirmed responders with an extended response ≥ 6 months (29.2% vs. 15.4%).

The median PFS was 8.0 months for the E + P + C group and 8.2 months for the PBO + P + C group (HR 1.47; 95% CI 0.91–2.36) (Fig. 2A). The PFS rates at 3, 6 and 9 months were numerically lower in the E + P + C versus the PBO + P + C group. The median OS was not reached in either group (Fig. 2B). The OS rates were numerically lower in the E + P + C versus the PBO + P + C group at 3, 6 and 9 months.

Fig. 2
figure 2

Kaplan–Meier curves for a PFSa and b OS. aBased on BICR assessment per RECIST v1.1. One-sided p-value based on log-rank test stratified by PD-L1 TPS (< 50% vs ≥ 50%) and predominant tumor histology (squamous vs non-squamous), because of small sample size, the strata ‘PD-L1 TPS greater than or equal to 50 percent Non-squamous’ and ‘PD-L1 TPS greater than or equal to 50 percent. Squamous’ were combined into one stratum. BICR blinded independent central review; OS overall survival; PFS progression-free survival; RECIST v1.1 Response Evaluation Criteria in Solid Tumors version 1.1

Full size image

A subgroup analysis did not identify any baseline characteristics associated with improved responses (Supplementary Table 2). Subgroups included age, baseline ECOG status, gender, smoking status, baseline brain metastasis, PD-L1 TPS (≥ 50% or < 50%), investigator choice of chemotherapy, predominant tumor histology, race (white vs. non-white), metastatic stage and non-East Asian geographic region.

Safety and tolerability

The frequency of AEs, drug-related AEs, grade 3 − 5 AEs, drug-related grade 3 − 5 AEs, serious AEs (SAEs) and drug-related SAEs were slightly higher (< 10% difference) in the E + P + C group compared with the PBO + P + C group (Table 3). Anemia was less frequent in the E + P + C group (24.4%) than in the PBO + P + C group (38.4%). In the E + P + C group, vomiting (22.2%) and rash (25.6%) were more frequent than in the PBO + P + C group (vomiting, 9.3%; rash, 18.6%). The most frequent SAEs (> 2 patients) in the E + P + C group were febrile neutropenia (4.4%), pneumonia (4.4%) and lower respiratory tract infection (3.3%). Of these, only pneumonia was reported in the PBO + P + C group (12.8%). The most frequent (> 2%) drug-related SAEs reported in the E + P + C group were febrile neutropenia (4.4%), diarrhea (2.2%), lower respiratory tract infection (2.2%) and neutropenia (2.2%). Of these, only diarrhea was reported in the PBO + P + C group at a similar rate (2.3%). The most frequent drug-related SAE in the PBO + P + C group, pneumonia (3.5%), was not reported in the E + P + C group. No serotonin syndrome related to epacadostat was observed in this trial.

Table 3 Summary of adverse events
Full size table

A higher percentage of patients in the E + P + C group had dose modifications of study drugs due to an AE compared with the PBO + P + C group (Table 3). Specifically, dose modification of pembrolizumab, epacadostat/placebo, carboplatin and pemetrexed were more frequent in the E + P + C group versus the PBO + P + C group. Similar proportions of patients in the E + P + C and PBO + P + C groups discontinued one or more study interventions due to drug-related AEs/SAEs (Table 3). Discontinuations of pembrolizumab or epacadostat/placebo were similar in both groups, while discontinuations of chemotherapy were more frequent in the E + P + C group versus the PBO + P + C group. Two deaths due to drug-related AEs (pneumonitis and sepsis) were reported in the E + P + C group and none in the PBO + P + C group (Table 3).

Pharmacodynamic activity of epacadostat

Median circulating kynurenine levels were measured at baseline (C1D1) and after one cycle of treatment (C2D1) (Fig. 3). In all treatment arms, circulating kynurenine levels were similar at C1D1 and C2D1 (PBO + P + C: 2.2 µM vs. 2.3 µM; E + P + C: 1.9 µM vs. 1.8 µM; E + P: 2.2 µM vs. 2.1 µM; all P = NS) and above the median levels observed in healthy subjects (1.5 μM) [16].

Fig. 3
figure 3

Circulating kynurenine levels at baseline (C1D1) and after one cycle of treatment (C2D1). The number of samples assessed was 83 in the placebo plus pembrolizumab and chemotherapy group, 74 in the epacadostat plus pembrolizumab and chemotherapy group (73 for C1), and 48 in the pembrolizumab plus epacadostat group. C1D1 vs. C2D1 were compared within each treatment arm using paired t-tests. Vertical lines represent maximum and minimum values, horizontal lines represent the median values, bars represent the interquartile range (25th-75th) percentiles, and dots represent outlier values. The dotted line indicates the median kynurenine levels in healthy subjects (1.5 μM) [16]. C cycle; D day; ns not significant

Full size image

Discussion

In ECHO-306/KEYNOTE-715, the addition of epacadostat to pembrolizumab + chemotherapy did not meet the prespecified success criterion for the primary ORR hypothesis in patients with previously untreated metastatic NSCLC. Subgroup analyses did not suggest any patient subgroup who experienced an improved response. PFS and OS differences between treatment arms were not conclusive to confirm an effect of the combination of epacadostat + pembrolizumab + chemotherapy. The median PFS and ORR in the placebo arm were consistent with those reported in other pembrolizumab + chemotherapy NSCLC studies [4, 20]. The addition of epacadostat to pembrolizumab + chemotherapy was generally well tolerated with a safety profile consistent with those previously observed for epacadostat and pembrolizumab + chemotherapy. No new AEs of safety concern were identified, but more patients in the epacadostat group discontinued chemotherapy.

In the current study, the addition of epacadostat to pembrolizumab plus chemotherapy did not provide additional benefit based on ORR, and the ORR was lower in the E + P + C arm. We cannot discount the possibility that epacadostat had a negative impact on outcomes. Although this trend is potentially of concern, it has not been consistently seen in other trials of epacadostat combinations (see accompanying articles, this issue). Also, given the lack of pharmacodynamic effect based on peripheral blood kynurenine analysis, there is no obvious biologic reason for why epacadostat 100 mg BID would reduce the quality of response to PD-1 inhibition in combination with chemotherapy. It also is unlikely that levels of epacadostat were altered due to drug-drug interactions. In a separate study (ECHO-207/KEYNOTE-723), no effects of pembrolizumab with various chemotherapies on epacadostat exposure were observed [21]. Instead, the reduction in ORR might have been the result of the differences in tolerability and safety observed with epacadostat versus placebo, imbalance between treatment arms in baseline prognostic factors, or the result of chance. Patients in the epacadostat arm had lower exposure to study drugs than those in the placebo arm as shown by more frequent dose modification of study drugs due to an AE, higher chemotherapy discontinuation rates (specifically, for carboplatin or pemetrexed) and fewer days and cycles of study treatments. In addition, there was a higher percentage of patients with stage M1C tumors at baseline in the epacadostat treatment group. The E + P + C arm also had a slightly higher percentage of patients with squamous NSLC histology, which is known to have a poorer prognosis than non-squamous NSCLC [4, 20, 22].

Studies that assessed addition of epacadostat to pembrolizumab for other types of cancers also did not meet prespecified efficacy endpoints. In the phase III ECHO-301/KEYNOTE-252 study evaluating epacadostat plus pembrolizumab versus placebo plus pembrolizumab in metastatic melanoma, the primary endpoint of PFS was not met [18]. Other reports in this supplement present similar findings in NSCLC (PD-L1 TPS ≥ 50%), urothelial carcinoma, renal cell carcinoma and head and neck squamous cell carcinoma (see accompanying articles, this issue).

Epacadostat 100 mg BID was chosen for evaluation in this and other studies on the basis of a potential well-tolerated safety profile in combination with pembrolizumab and chemotherapy, robust ORRs, durable disease control rates, and the observation that epacadostat monotherapy at BID doses ≥ 100 mg has been shown to normalize plasma kynurenine levels in patients with advanced solid tumors [16]. However, in this study, epacadostat 100 mg BID did not appear to reduce the levels of circulating kynurenine in patients with NSCLC when administered with pembrolizumab plus chemotherapy or pembrolizumab alone. Although not directly observed in the PBO + P + C arm of the current study, pembrolizumab has been reported to enhance kynurenine production [23]. Furthermore, a retrospective analysis of epacadostat clinical studies showed that epacadostat doses of ≥ 600 mg BID may be needed to normalize plasma and intratumoral kynurenine levels when given in combination with a checkpoint inhibitor [23]. Thus, when used in combination with pembrolizumab, higher doses of epacadostat may be needed to normalize kynurenine levels. Evaluation of higher epacadostat doses may shed light on the potential efficacy of combination treatment with epacadostat and pembrolizumab.

Other studies are also evaluating the combination of IDO1 and PD-1/PD-L1 inhibition for the treatment of cancer. Epacadostat plus the PD-1 inhibitor nivolumab is being assessed in patients with advanced solid tumors, including NSCLC, in the phase I/II ECHO-204 study; promising preliminary antitumor activity has been reported, particularly in squamous cell carcinoma of the head and neck with epacadostat 300 mg BID in combination with nivolumab and in melanoma with epacadostat 100 mg BID or 300 mg BID in combination with nivolumab [24]. A phase I trial assessing the IDO1 inhibitor navoximod with the PD-L1 inhibitor atezolizumab showed antitumor activity in various tumor types, including NSCLC; however, no clear benefit of adding navoximod was seen [25]. Identification of biomarkers associated with a benefit of combined IDO1 and PD-1/PD-L1 inhibition is needed to determine the usefulness of this approach in specific patient populations and to guide further studies.

Limitations of our study include the small sample size, the change in study design from phase III to phase II during the study and the early study discontinuation. Because ORR is not a validated surrogate endpoint for OS with checkpoint inhibitors, this study did not directly assess the potential clinical benefit. Subgroup analyses based on epacadostat pharmacodynamics was not possible due to the small fraction of patients with reduced serum kynurenine.

Conclusions

The addition of epacadostat 100 mg BID to pembrolizumab and platinum-based chemotherapy was generally well tolerated, but this study did not meet the pre-specified success criterion for the primary ORR hypothesis in this population of frontline patients with NSCLC. Additional biomarker analyses and clinical studies are needed to determine if a combination strategy including epacadostat (possibly at higher doses) could provide clinical benefit in selected populations.