The study (NCT02309177) was approved by the institutional review board or independent ethics committee at each study site (listed in Supplementary Table 1) and conducted in accordance with Good Clinical Practice, as described in the International Conference on Harmonisation E6 guidelines, and with the ethical principles outlined in the Declaration of Helsinki (10). All patients provided written informed consent before any study procedures were initiated.

A data review committee, comprising investigators and the sponsor, was established to monitor the conduct of the study and to confirm dose level and/or identify the need for dose de-escalation. A safety oversight subcommittee, made up of data review committee members, assessed safety and tolerability in part 1 and identified the recommended part 2 dose (RP2D) to be administered in the expansion phase.

Patients (aged ≥18 years) with histologically or cytologically confirmed stage IIIB or IV NSCLC who had not previously received chemotherapy or investigational therapy for the treatment of metastatic disease and who were not candidates for curative surgery or radiation were eligible. Adjuvant or neoadjuvant chemotherapy was allowed if it was completed >12 months prior to randomization, with no disease progression or recurrence during those 12 months. Patients were required to have measurable disease according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 guidelines and an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1. Patients who received prior therapy with immune checkpoint inhibitors or had grade ≥2 peripheral neuropathy or any lung disease that could potentially interfere with detection or management of suspected drug-related pulmonary toxicity were excluded.

This phase I, open-label, multicenter study evaluated the safety and efficacy of nivolumab plus nab-paclitaxel–based regimens in advanced solid tumors in 6 cohorts: 2 each for pancreatic cancer, NSCLC, and metastatic breast cancer. Here, we report the outcomes in the 2 NSCLC cohorts, in which nivolumab initiation was either concurrent with chemotherapy in cycle 1 or delayed to cycle 3.

All patients received nab-paclitaxel 100 mg/m² intravenously on days 1, 8, and 15 and carboplatin area under the curve 6 mg/mL/min intravenously on day 1 every 21 days for the first 4 cycles. Based on the available evidence at the time (11–13), nivolumab was dosed at 5 mg/kg intravenously; it was given on day 15 of each 21-day cycle, starting either with nab-paclitaxel/carboplatin in cycle 1 (concurrent) or in cycle 3 (delayed). Dosing of nivolumab on day 15 of each cycle was intended to minimize potential adverse events arising from interactions between chemotherapy and nivolumab (i.e., through dose adjustment of nab-paclitaxel on day 8, if warranted). Patients continued to receive nivolumab alone beyond 4 cycles. Nivolumab dose was determined based on dose-limiting toxicities (DLTs); once established, the nivolumab dose could not be reduced for individual patients, but it could be modified (see Supplementary Methods for details). Treatment could continue until disease progression (RECIST 1.1), unacceptable toxicity, or withdrawal of consent. If a treatment-related unacceptable toxicity was attributed to only 1 agent in the combination, patients were permitted to remain in the study and receive other agent(s) per the assigned schedule. Patients were permitted to continue nivolumab treatment beyond initial RECIST 1.1–defined progressive disease if they met a set of specific criteria (see Supplementary Methods for details). The study therapy was to be stopped if further disease progression was documented. Supportive care was permitted at the discretion of the investigator.

The study was designed with 2 parts: a dose-finding part and an expansion part (Figure 1). In the dose-finding part, DLTs were to be evaluated. In the expansion part, tolerability was to be further characterized and antitumor activity was to be explored. Patients were initially assigned to the concurrent cohort, and if the dose-finding part was deemed safe, the RP2D was to be declared and patients were to be assigned in a non-randomized manner to either the expansion part of the concurrent cohort or the dose-finding part for the delayed cohort (Figure 1; see Supplementary Methods for additional details).

The primary endpoints of the study were the number of patients with DLTs in the dose-finding part of each treatment cohort (see Supplementary Methods for definitions of DLT and DLT-evaluable population) and the percentage of patients with grade 3/4 treatment-emergent adverse events (TEAEs) or treatment discontinuation due to a TEAE during the study. Secondary endpoints included TEAEs leading to dose reduction, dose delay, or treatment discontinuation; investigator-assessed progression-free survival (PFS); overall survival (OS); disease control rate (DCR); overall response rate (ORR); and duration of response (DOR; all responses were based on RECIST 1.1 criteria. Exploratory endpoints included correlation between programmed death ligand 1 (PD-L1) expression prior to treatment and tumor response.

Expression of PD-L1 in tumor biopsy samples was analyzed using a validated, automated immunohistochemistry assay with the rabbit monoclonal antibody 28-8 pharmDx kit (Agilent, Carpinteria, CA) and Signatera (RUO) kit (Natera, San Carlos, CA). The percentage of tumor cells with complete circumferential or partial linear membrane staining of any intensity, but not cytoplasmic staining, was used to score PD-L1 staining.

The treated population comprised patients who received ≥1 dose of any investigational product. Separately, analyses were also conducted on the subset of patients who received ≥1 dose of nivolumab. However, because nivolumab treatment in the delayed cohort depended on whether patients stayed on treatment through cycle 3, treatment efficacy in these 2 populations should not be compared.

At least 6 DLT-evaluable patients were needed at the given dose level to determine the RP2D. If a treatment was expanded to part 2, additional patients were enrolled to achieve a total of 20 nivolumab-treated patients. Twenty nivolumab-treated patients allowed accurate assessment of the incidence of adverse events (AEs).

TEAEs were summarized based on the Medical Dictionary for Regulatory Activities (MedDRA) version 21.0 guidelines and graded for severity using the National Cancer Institute's Common Terminology Criteria for Adverse Events version 4.0. The ORR and DCR were reported using proportions and exact 95% confidence intervals (CIs), and the PFS, OS, and DOR were summarized using the Kaplan-Meier method with median times and 95% CIs. Patients who discontinued treatment for reasons other than disease progression, consent withdrawal, or initiation of a new anticancer therapy were followed up for tumor response assessments and new anticancer therapies.

Among patients assigned to receive concurrent nivolumab, 22 were treated (20 received concurrent nivolumab) at 8 sites from May 2015 through June 2018: 12 patients were enrolled in the dose-finding part (6 of these were the DLT-evaluable patients), and 10 were enrolled in the expansion part. In this cohort, 2 patients did not receive nivolumab due to AEs. Patients in this cohort discontinued treatment due to progressive disease [11 patients (50.0%)], AEs [6 patients (27.3%)], withdrawal by patient [3 patients (13.6%)], or other reasons [2 patients (9.1%)].

Among patients assigned to receive delayed nivolumab, 10 were treated (6 received delayed nivolumab and were DLT-evaluable) at 7 sites from January 2016 through September 2018 and were enrolled in the dose-finding part. In this cohort, 4 patients discontinued treatment due to progressive disease before cycle 3 and, therefore, did not receive nivolumab. Patients in this cohort discontinued treatment due to progressive disease [7 patients (70.0%)], AEs [2 patients (20.0%)], or study termination [1 patient (10.0%)]. This arm was not expanded.

Demographic and baseline clinical characteristics are reported in Table 1. The median age was 65.0 years in the concurrent cohort and 70.0 years in the delayed cohort; among patients with confirmed histology, adenocarcinoma was most common (54.5 and 50.0%, respectively). Most patients in the concurrent cohort were female (72.2%), had ECOG PS of 1 (68.2%), and had PD-L1 expression ≥1% (54.5%). In the delayed cohort, most patients were male (70.0%), had ECOG PS of 0 (60.0%), and had PD-L1 expression <1% (50.0%) [among 70% with available data] (Table 1). The patients in the nivolumab-treated subset showed similar trends (Supplementary Table 2).

No DLTs were reported during the dose-finding part in the concurrent cohort. All patients had ≥1 all-grade TEAE; ≥1 grade 3/4 TEAE was reported in 90.9% of patients (Table 2). At least 1 serious TEAE was reported in 36.4% of patients. Grade 3/4 TEAEs that occurred in ≥10% of patients were (MedDRA preferred terms) anemia, neutropenia, neutrophil count decreased, white blood cell count decreased, hypokalemia, and vomiting. Grade 1/2 peripheral neuropathy was observed in 11 patients (50%). Results were generally similar in the nivolumab-treated subset (Supplementary Table 3). At least 1 all-grade or grade 3/4 TEAE of special interest attributable to nivolumab was reported in 16 (80%) and 3 (15%) of nivolumab-treated patients, respectively; grade 1/2 hypothyroidism and pneumonitis occurred in 3 (15.0%) and 2 (10.0%) patients, respectively (Supplementary Table 3).

In this cohort, 21 patients (95.5%) who received any combination of nab-paclitaxel, carboplatin, or nivolumab had ≥1 TEAE that resulted in dose reduction and/or interruption, and 7 patients (31.8%) discontinued treatment due to TEAEs (Table 3a). The most common (incidence ≥10%) TEAEs leading to dose reduction and/or interruption were hematologic in nature. Two patients required nivolumab interruption due to pneumonitis. The most common TEAE leading to treatment discontinuation was decreased neutrophil count attributable to nab-paclitaxel. Results were generally similar in the nivolumab-treated subset (Supplementary Table 4A).

Patients received treatment for a median of 32.25 weeks in a median of 9.0 cycles (see Supplementary Table 5 for additional treatment exposure and dose modification data, including those for the nivolumab-treated subset).

For the PFS analysis, 13 patients (59.1%) had disease progression or died. The investigator-assessed median PFS was 10.5 months (95% CI: 4.93–28.42) (Figure 2A), and the 1-year estimated PFS rate was 43.0% (95% CI: 19.76–64.43). The results were generally similar in the nivolumab-treated subset (Supplementary Table 6). Among patients with known baseline PD-L1 status (n = 16), the median PFS in the nivolumab-treated subset was similar in patients with PD-L1 expression <1% (10.2 months) and PD-L1 expression ≥1% (10.5 months) (Figure 3A). Among patients with confirmed histology (n = 21), the median PFS was 10.5 months in patients with squamous histology and 7.4 months in those with non-squamous histology (Figure 4A).

For the OS analysis, 15 patients (68.2%) had died. The median OS was 29.3 months (95% CI: 9.13–38.47) (Figure 2B), and the 1-year estimated OS rate was 68.2% (95% CI: 44.62–83.38). The results were similar in the nivolumab-treated subset (Supplementary Table 6). Among patients with known baseline PD-L1 status in the nivolumab-treated subset (n = 16), the median OS was numerically longer in patients with PD-L1 expression ≥1% (30.3 months) vs. PD-L1 expression <1% (18.5 months) (Figure 3B). Among patients with confirmed histology (n = 21), the median OS was numerically longer in patients with non-squamous (38.5 months) vs. squamous histology (12.1 months) (Figure 4B).

The confirmed ORR was 45.5%, with 1 complete response and 9 partial responses (PRs) (Table 4); all responses occurred in nivolumab-treated patients. Among patients in the nivolumab-treated subset with available baseline PD-L1 expression levels (n = 16), the ORR was 40.0% in patients with PD-L1 expression <1% (PRs in 2 of 5 patients) and 63.6% in those with PD-L1 expression ≥1% [7 (1 complete response, 6 PRs) of 11 patients]. The DCR was 90.9%, and the median DOR was 9.2 months [95% CI: 3.25–not evaluable (NE)] (Table 4). The responses were generally similar in the nivolumab-treated population (Supplementary Table 7). The median best percent change from baseline in total length of target lesions was −35.1%; Figure 5A shows individual values. In this cohort, 4 patients were treated with nivolumab beyond the initial RECIST-defined progressive disease, and the best percent changes in total length of target lesions from the first disease progression event in these patients were −22, 0, 15, and 40%.

One DLT (grade 2 pneumonitis that did not resolve with dose delay and systemic steroids) was reported in the delayed cohort. All patients had a grade 3/4 TEAE, and 20.0% had ≥1 serious grade 3/4 TEAE (Table 2). Grade 3/4 TEAEs that occurred in ≥10% of patients were neutropenia, anemia, neutrophil count decreased, thrombocytopenia, and hypokalemia (Table 2). One patient (10%) experienced grade 3/4 peripheral neuropathy; grade 1/2 peripheral neuropathy was reported in 3 patients (30%). In this cohort, 8 patients (80.0%) who received any combination of nab-paclitaxel, carboplatin, or nivolumab had ≥1 TEAE that resulted in dose reduction and/or interruption, and 2 (20.0%) patients discontinued treatment due to TEAEs (Table 3b). The results were generally similar in the nivolumab-treated subset (Supplementary Table 4B).

Patients in the delayed cohort received treatment for a median of 17.55 weeks in a median of 4.5 cycles (see Supplementary Table 5 for additional treatment exposure and dose modification data, including those for the nivolumab-treated subset).

For the PFS analysis, 7 patients (70.0%) in the delayed cohort had disease progression or died. The investigator-assessed median PFS was 4.1 months (95% CI: 1.25–NE; Figure 2C), and the 1-year estimated PFS rate was 24.0% (95% CI: 3.77–53.73). In patients with non-squamous histology, the investigator-assessed median PFS was 2.9 months; it was not evaluable in patients with squamous histology. The investigator-assessed median PFS was not evaluable in the nivolumab-treated subset of patients with PD-L1 expression <1%; in those with PD-L1 expression ≥1%, it was 4.1 months.

For the OS analysis, 7 patients (70.0%) had died. The median OS was 8.2 months (95% CI: 2.20–NE; Figure 2D), and the 1-year estimated OS rate was 40.0% (95% CI: 12.27–67.02). Overall survival results for the nivolumab-treated subset are summarized in Supplementary Table 6. In patients with non-squamous histology, the median OS was 7.3 months; it was not evaluable in patients with squamous histology. The median OS was not evaluable in the nivolumab-treated patients with PD-L1 expression <1%; in those with PD-L1 expression ≥1%, it was 7.6 months.

As noted in the “Patients and Treatment” section, 4 of the 10 patients assigned to the delayed cohort discontinued treatment before cycle 3. The PFS and OS outcomes in the nivolumab-treated subset of patients in the delayed cohort (Supplementary Table 6) were markedly different from those in all patients in the delayed cohort described above in this section.

The ORR was 30.0%; 3 patients—all of whom received nivolumab—had a confirmed PR; 2 of the 3 PRs were noted before nivolumab administration. Among patients with PD-L1 data available (6 patients: 5 with PD-L1 <1% and 1 with PD-L1 ≥1%), the ORR was 60.0% in nivolumab-treated patients with PD-L1 expression <1% (PRs in 3 of 5 patients); no response was observed in the patient with PD-L1 expression ≥1%. The DCR was 60.0%, and the median best percent change from baseline in total length of target lesions was −13.1%; Figure 5B shows individual values. In this cohort, 2 patients were treated with nivolumab beyond the initial RECIST-defined progressive disease, and the best percent changes in total length of target lesions from the first disease progression event in these patients were 0 and 16%.

JG has received research grants from Bristol-Myers Squibb, Roche/Genentech, AstraZeneca/MedImmune, and Merck and has served as a speaker for Merck. DW has served in advisory roles for AbbVie, Amgen, AstraZeneca, Bristol-Myers Squibb, Celgene, and Janssen and has served as a consultant for CTI Biopharma and McGivney Global and as a speaker for Bristol-Myers Squibb, Celgene, Janssen, Roche/Genentech, and Lilly. BG has served in consulting/advisory roles for Celgene, Cook Medical, Foundation Medicine, Ipsen, Merrimack, Exelixis, Taiho Oncology, Bristol-Myers Squibb, and Terumo Interventional Systems. PO'D has received research support from Pfizer, Genentech, Bristol-Myers Squibb, GlaxoSmithKline, Five Prime, Forty Seven, Boston Biomedical Inc., Novartis, Celgene, Incyte, Lilly/ImClone, Array, H3 Biomedicine, and Taiho; has served as a consultant for Genentech and Celgene; and has given expert testimony for Bayer and Lilly. RB was an employee of Celgene at the time of conduct of the study. SB, LL, CL, and TO are employees of Celgene and hold stock options. KK has served in advisory roles for Bristol-Myers Squibb.