Immunomodulatory Activity of Nivolumab in Metastatic Renal Cell Carcinoma

Metastatic renal cell carcinoma (mRCC) is a heterogeneous disease that is highly resistant to chemotherapy (1). High-dose IL2 produces durable, complete responses in a fraction of patients, providing proof of concept for the potential of immunotherapy in mRCC (2). Recently, targeted agents including tyrosine kinase inhibitors, VEGF inhibitors, and mTOR inhibitors have become available for treatment of mRCC (3–6). Outcomes with these agents are improved, but therapeutic resistance is inevitable and median overall survival (OS) is limited (<20 months; refs. 3–6).

Cellular immune responses may play a key role in modulating tumor progression in RCC and other cancers (7, 8). Tumors may co-opt immune checkpoint pathways to suppress the duration and amplitude of immune responses as a means of immune resistance (9). The introduction of the fully human programmed death-1 (PD-1) immune checkpoint inhibitor nivolumab has changed the therapeutic landscape for solid tumors, demonstrating durable responses in a subset of patients with multiple tumor types and improved OS in advanced melanoma and non–small cell lung cancer to date (10–15). In mRCC, objective response rates (ORR) of 20% to 22% and median OS of 18.2 to 25.5 months were reported in previously treated patients (12). A recent phase 3 study demonstrated significantly longer median OS with nivolumab versus everolimus (25.0 vs. 19.6 months, respectively; HR = 0.73, P = 0.002) in patients with previously treated with mRCC (16).

Examination of biomarkers may reveal prognostic or predictive factors relating to the disease or its treatment, which could provide the scientific rationale for combinations to address resistance. In RCC, treatment decisions still depend primarily on clinical criteria only (17). For anti-PD-1–directed therapies, increased tumor or stromal PD-1 ligand 1 (PD-L1) expression is a logical candidate to predict therapeutic response but does not appear to define the chance of response in a dichotomous way. Prior biomarker analysis, particularly in patients with melanoma, provided further evidence that evaluating the tumor microenvironment could provide insights into the mechanism of tumor responses to immunotherapies (18–22). However, most studies evaluated biomarkers only in baseline biopsies. We hypothesize that examination of on-treatment biopsies provides additional insights into key changes in the tumor microenvironment that may contribute to or hinder immune response during treatment in mRCC.

In this hypothesis-generating exploratory analysis, we sought to investigate the immunomodulatory activity of nivolumab in patients with clear cell mRCC. Here, we present an evaluation of changes in tumor-associated lymphocytes (e.g., CD3⁺, CD4⁺, and CD8⁺ T cells), tumor PD-L1 expression, immune gene expression in the tumor, and serum-soluble factors (e.g., CXCL9 and CXCL10)—known markers of T-cell activation and migration in patients in response to immunotherapies (22–24).

This was an open-label, parallel, four-group, phase 1b study of nivolumab (Bristol-Myers Squibb; Ono Pharmaceutical Company Limited). Previously treated patients were randomized 1:1:1 to receive nivolumab 0.3, 2, or 10 mg/kg; treatment-naïve patients received nivolumab 10 mg/kg. Nivolumab was administered as an intravenous infusion on day 1 of the treatment cycle every 3 weeks until confirmed complete response, progressive disease, intolerable adverse events (AE), or withdrawal of consent. Patients were permitted to continue nivolumab treatment beyond progression if the investigator noted clinical benefit and treatment tolerance.

Patients were required to provide fresh image-guided core needle biopsies of a soft-tissue site both at screening and on treatment as a condition of protocol participation. Tumor sites selected for biopsy must not have received previous radiation therapy. It was preferred that patients have at least one lesion large enough to undergo repeat core needle biopsies (one biopsy at baseline and one biopsy at C2D8). However, patients may also have had two distinct soft-tissue lesions eligible for core needle or excisional biopsies. One lesion was to be biopsied at baseline and the other was to be biopsied at C2D8.

Institutional review boards of participating institutions approved the protocol, and the study was conducted according to the Declaration of Helsinki and International Conference on Harmonisation. All patients gave written informed consent.

Eligible patients had histologically confirmed mRCC with a clear-cell component, measurable disease defined by Response Evaluation Criteria in Solid Tumors (RECIST) v1.1, Karnofsky performance score of ≥70%, presence of soft-tissue tumor lesions that could be biopsied at baseline and on treatment, and adequate organ and marrow function. To be eligible for the previously treated groups, patients must have been treated with between 1 and 3 previous systemic therapies for RCC, with progression following the most recent therapy within 6 months of study enrollment. For the treatment-naïve group, patients must not have received any previous systemic therapy in the metastatic or adjuvant setting. Exclusion criteria included active central nervous system metastases within 30 days of study enrollment; active or prior autoimmune disease; prior malignancy unless complete remission occurred ≥2 years prior to study enrollment; and previous treatment with anti–CTLA-4, anti–PD-1, anti–PD-L1, anti–PD-L2, anti-CD137, anti-CD40, or anti-OX40 antibodies.

The primary objective was to investigate the pharmacodynamic immunomodulatory activity of nivolumab on CD4⁺ and CD8⁺ tumor-associated lymphocytes and serum chemokines CXCL9 and CXCL10 in patients with clear cell mRCC. Secondary objectives were to assess the activity and safety of nivolumab. Exploratory objectives included analyses of serum-soluble factors and gene expression profiling in the tumor and whole blood. Data cutoff for all analyses was January 12, 2015.

Patients were evaluated for response every 6 weeks for the first 12 months from randomization, then every 12 weeks until documented progression. Patients who discontinued treatment for reasons other than progression continued to have tumor assessments according to schedule.

Biopsies obtained from metastatic lesions at baseline (screening) and cycle 2 day 8 (C2D8) were used to evaluate tumor-associated lymphocytes, PD-L1 expression, and gene expression. Biopsies were divided in half, with half formalin-fixed and paraffin-embedded for immunohistochemical analyses and half in RNAlater for gene expression analyses. Therefore, there was no need to prioritize tumor tissue sections for immunohistochemical or gene expression analyses. The number and composition of tumor-associated lymphocytes were assessed using two validated multiplex immunohistochemical assays provided by Mosaic Laboratories on formalin-fixed, paraffin-embedded sections. Assays included one dual stain, CD3⁺/CD8⁺, and one triple stain, CD3⁺/CD4⁺/FOXP3⁺. Each multiplex immunohistochemical assay was designed and validated to be compatible with Clinical Laboratory Improvement Amendments guideline class I test validation. A representative 20× staining field was spectrally imaged using the Nuance Multispectral Imaging System with software v.2.4.0 (Caliper Life Sciences) attached to a Nikon 90i microscope. The multispectral image was acquired in a spectral range of 420 to 720 nm using 20-nm wavelength steps. Image cubes were analyzed using inForm software v1.2 (Caliper Life Sciences). Image cubes were unmixed using spectral absorbance patterns for each chromogen and hematoxylin.

PD-L1 expression on the tumor cell surface was assessed in fresh samples at a central laboratory using an automated immunohistochemical assay (Bristol-Myers Squibb/Dako immunohistochemical assay using the 28-8 antibody), as described previously (25). RNA was extracted from fresh biopsies in parallel to immunohistochemistry and from whole blood at C1D1 (prior to nivolumab infusion), C1D2, and C2D8. RNA was labeled by WT-Pico Ovation (NuGEN). Gene expression profiling was performed using the HG-U219 array plate on the GeneTitan platform (Affymetrix). The robust multiarray analysis algorithm (26) was used to establish intensity values for each of 18,562 loci (BrainArray v.10; ref. 27). Data have been deposited in ArrayExpress (E-MTAB-3218 and E-MTAB-3219).

Assessment of serum chemokines (CXCL9, CXCL10) and other serum-soluble factors at baseline (C1D1 before treatment), C1D1 (after treatment), C2D1, C2D8, and C4D1 was performed for all treated patients for whom serum was available using a multiplex panel based on Luminex technology (Myriad Rules-Based Medicine).

Safety assessments were conducted at every visit and evaluated according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0 (28). Patients who received at least one dose of study drug were included in the safety population.

Activity analyses included best overall response (complete response, partial response, stable disease, progressive disease), progression-free survival (PFS), and OS, as well as ORR, that is, the proportion of patients whose best response was a complete response or partial response. The 95% confidence intervals (CI) for ORR were estimated using the Clopper–Pearson method (29). PFS was defined as the time from first dose to first documented disease progression, or death. PFS and OS functions were estimated by the Kaplan–Meier method with 95% CIs estimated using Greenwood's formula (30).

This study was not designed to statistically test a specific hypothesis; therefore, the sample size was not based on statistical power calculations. Pharmacodynamic effects of nivolumab on tumor-associated lymphocytes were described using summary statistics and changes or percent changes from baseline tabulated by cycle, visit, and dose.

Pharmacodynamic analyses of gene expression were based on an extended linear model, fit by restricted maximum likelihood (NLME version 3.1-109 under R 3.0.1 for Linux; ref. 31). For blood samples, the model included fixed effects of treatment group and time on study as categorical variables and treatment-by-time on study interactions. For tumor biopsy samples, the model also included fixed effects of process batch and sex (the latter because women were not equally represented in samples from each trial treatment group). Within-patient correlations were modeled by a spatial exponential structure with Euclidean distance. A multicontrast conditional F test was used to compare the null hypothesis that all time-related fixed-effect parameters were zero versus an alternative hypothesis that gene expression changed over time in at least one treatment group. The q value of each test (expected proportion of false positives incurred at that P value) was also estimated. Results presented are genes for which this null hypothesis was rejected (P for time on study < 0.01), and the change over time averaged over treatment groups was ≥1.3-fold (biopsy; 108 genes in Supplementary Table S1) or ≥1.2-fold (blood; 59 genes in Supplementary Table S2). Transcripts meeting these significance criteria for a test of time on study were evaluated for enrichment (ref. 32; P values provided were Bonferroni–corrected) of 1,539 genes from immune cell lineages (33) and for biologic impact (MetaCore; Thomson Reuters).

To examine possible treatment group-specific effects among genes where the null hypothesis was rejected (P for time on study, <0.01), a second such multicontrast test of all time-by-treatment interaction parameters was used to test whether the pattern of expression change differed between at least two treatment groups. If this null hypothesis was rejected (P < 0.01 for interaction between dose and time on study or between previous treatment status and time on study), then we examined the effect size in each treatment group. Genes for which the change over time for at least 1 treatment group was ≥1.3-fold (biopsy; 37 probesets) or ≥1.2-fold (blood; 24 probesets) are presented in Supplementary Tables S3 and S4. In all cases, the change over time for at least two of the other treatment groups did not meet those criteria.

Patients were enrolled from September 2011 to September 2012 at 14 participating international centers. Ninety-two patients were assigned to treatment, 91 of whom were treated (Supplementary Fig. S1). The median age was 61 years, 67% were male, and 66% had received previous therapy for metastatic disease (Table 1). Baseline characteristics were similar between previously treated (n = 67) and treatment-naïve (n = 24) patients (Table 1).

Responses were evaluated in the 91 treated patients (Table 2). The ORR was 15% (95% CI, 8.7–24.5). PFS rates were 43% (95% CI, 32–53) at 24 weeks and 25% (95% CI, 16–35) at 48 weeks. OS rates were 75% (95% CI, 64–83) at 12 months and 58% (95% CI, 46–68) at 24 months. Median OS (95% CI) was 16.4 months [10.1 to not reached (NR)], NR, and 25.2 months (12.0 to NR) for the nivolumab 0.3, 2, and 10 mg/kg groups for previously treated patients, respectively, and NR for treatment-naïve patients.

Repeat core needle biopsies were obtained for the majority of patients (same lesion: 60 of 91; distinct or unknown second lesion: 13 of 91), with metastatic sites including liver (16 of 73), lymph nodes (12 of 73), kidney (9 of 73), lung (6 of 73), adrenal (4 of 73), pancreas (2 of 72), and other locations (soft tissue, 24 of 73). Immunohistochemical analysis showed enrichment of CD3⁺, CD4⁺, and CD8⁺ cells from baseline to C2D8 (Fig. 1A and B). For all nivolumab-treated patients who had baseline and C2D8 values (n = 36), median increases from baseline to C2D8 in the proportion of CD3⁺, CD4⁺, and CD8⁺ cells were 9.83%, 0.44%, and 2.64%, respectively. Median percent changes from baseline of CD3⁺, CD4⁺, and CD8⁺ cells were 69%, 180%, and 117%, respectively, with most patients experiencing increases. Baseline percentages and increases from baseline were greater for CD3⁺ and CD8⁺ than for CD4⁺ cells (Fig. 1B). Most patients had very low or undetectable levels of CD4⁺ cells at baseline and only modest changes from baseline. Fourteen patients had both baseline percentage levels of CD4⁺ cells and changes from baseline values of <0.45%. These results did not appear to vary with nivolumab dose or previous treatment status (data not shown). The proportion of CD3⁺, CD4⁺, and CD8⁺ cells and their relationships to each other are shown in Supplementary Fig. S2.

RNA expression analysis from tumor biopsies obtained in parallel (n = 114) showed that mean levels of transcripts for subunits of CD3 and CD8, but not CD4, significantly increased on treatment (1.7-fold for 915_at/CD3D, P = 0.006; 1.7-fold for 925_at/CD8A, P = 0.002; 1.2-fold for 920_at/CD4, P = 0.175; Fig. 1C). These changes were not dependent on nivolumab dose or previous treatment status.

PD-L1 expression on tumor cells was assessed by immunohistochemistry in fresh biopsies obtained at baseline and C2D8. Of 56 evaluable baseline biopsies, 18 (32%) had ≥5% PD-L1 expression. In patients with fresh matched biopsies at baseline and on treatment, there was no consistent change in tumor PD-L1 expression following nivolumab treatment relative to baseline (Supplementary Fig. S3).

Expression profiling data were obtained from 59 tumor biopsies at baseline and 55 at C2D8. A total of 42 patients had samples at both timepoints, of which at least 34 were repeat biopsies of the same lesion. Statistical analysis identified 108 transcripts that changed over time (≥1.3-fold change in mean expression, P < 0.01; Supplementary Table S1). The 108 transcripts included 71 previously associated with immune lineages (ref. 33; P = 1.3 × 10⁻⁷⁶), all of which increased at C2D8. Of these 71 transcripts, 43 have been defined as lymphoid lineage–specific (e.g., GZMA/G/H and KLRB1/D1/G1) or myeloid–specific (e.g., CXCL11, PD-L1, and IDO1; Fig. 2A). A subset of the lymphoid lineage transcripts are completely specific to the T-cell lymphoid subset (e.g., CTLA-4, CD8A/B, CD3D/E/G, and ICOS). Sixteen of the 108 transcripts were previously identified as IFN-regulated (34), including CXCL9 and CD274 (PD-L1). IFNγ was the only IFN represented in the 108 genes. Forty-seven of the transcripts were previously identified as showing differential expression in baseline biopsies of patients who subsequently responded to ipilimumab (P = 2.7 × 10⁻⁸⁶).

To evaluate whether pharmacodynamic transcriptional changes were observed in the periphery, microarray analysis was performed on whole blood samples (N = 82 at C1D1 and 74 at C1D2, with 70 having matched samples; N = 73 at C2D8). Expression of 59 transcripts changed from baseline (C1D1) to C1D2 (≥1.2-fold change in mean expression, P < 0.01; Supplementary Table S2), including 30 previously associated with immune lineages (ref. 33; P < 0.001; Fig. 2B). These included transcripts for T-cell receptor α and β subunits and the CD3γ subunit, all of which decreased relative to baseline. The 59 transcripts included 29 previously identified as IFN-regulated (34), all of which increased. No transcripts from IFN genes were regulated or detectable in blood.

These transcriptional effects were generally similar between dose groups and between previously treated and treatment-naïve patients (P > 0.01 for interaction between time and dose group or previous treatment status). The analyses of pharmacodynamic transcriptional effects that differ between treatment groups are presented in Supplementary Tables S3 and S4.

As observed increases in IFNγ-regulated chemokine transcripts in tumor could potentially result in an increase in circulating chemokines in the periphery, several serum-soluble factors were quantified (Supplementary Table S5). In serum, increases were noted in CXCL9 and CXCL10, with median changes of 1,861 pg/mL (range, −2,000 to 22,890) and 157 pg/mL (range, −398 to 3,930), respectively, from baseline (C1D1 before treatment) to C2D8 (N = 83). Median percent changes from baseline of these chemokines were 101% (range, 45–1,730%) and 37% (range, −30% to 936%), respectively. Most patients had increases in both CXCL9 and CXCL10, observed across all baseline values (Fig. 3A). Within-patient changes in CXCL9 tended to be greater than within-patient changes in CXCL10 (Fig. 3B), and changes in CXCL9 and CXCL10 were highly correlated. In tumor, mean levels of CXCL9 and CXCL10 transcripts increased from baseline to C2D8 (2.4-fold for 4283_at/CXCL9, P < 0.001; 2-fold for 3627_at/CXCL10, P = 0.011; Fig. 3C). These observed changes were not associated with nivolumab dose or previous treatment status (data not shown). Serum levels of CXCL9 and CXCL10 cytokines at C2D8 showed correlation with their transcript levels in the corresponding patient biopsy (N = 54; CXCL9: r = 0.37, P = 0.006; CXCL10: r = 0.30, P = 0.029; Fig. 3D). No correlation was observed between serum cytokine levels at C1D1 (before treatment) and transcript levels in the corresponding patient biopsy obtained at screening (r < 0.21, P > 0.1).

Treatment-related AEs of any grade occurred in all previously treated and treatment-naïve patients (Table 3). The most common AEs (all grades) were fatigue and nausea, mainly grade 1 or 2. Grade 3 to 4 AEs occurred in 54% of previously treated patients overall and 50% of treatment-naïve patients, mainly fatigue.

Categories of select AEs, or AEs with potential immunologic causes, were also assessed (Table 3). Of interest, select pulmonary AEs (all pneumonitis) were only reported in 3 patients in the previously treated group (1 in the 0.3 mg/kg group and 2 in the 10 mg/kg group) and 3 patients in the treatment-naïve group.

This prospective exploratory study highlights the importance of these pharmacodynamic studies for future investigations of immunotherapeutic agents in mRCC. This is the first prospective translational study involving analysis of both baseline and on-treatment biopsies in RCC, aimed specifically at understanding the immunomodulatory activity of nivolumab. The immunomodulatory effect of PD-1 inhibition with nivolumab was demonstrated through multiple lines of evidence across all doses studied.

Clinical activity was observed in both treatment-naïve and previously treated patients at each dose. Median OS was between 16.4 and 25.2 months for previously treated patients, similar to that seen in a phase 2, randomized, dose-ranging study of nivolumab (18.2–25.5 months) and in a recent phase 3 study of nivolumab versus everolimus (25.0 months) in advanced clear cell RCC (12, 16). Median OS was not reached for treatment-naïve patients. The type and frequency of AEs were similar in previously treated and treatment-naïve patients and consistent with previous reports of nivolumab in solid tumors; the frequency of severe AEs was low (10–12). As with the dose-ranging study of nivolumab (12), there was no obvious dose–response relationship.

The current study of the pharmacodynamic effects of nivolumab in RCC has important implications to further our understanding of its mechanism of action in this setting and to select combination therapies. We have demonstrated that nivolumab reverses T-cell exhaustion within the tumor microenvironment as hypothesized. Immunohistochemical analysis of tumor-associated lymphocyte markers (i.e., CD3⁺ and CD8⁺) showed increased lymphocytic presence in biopsies at C2D8 in the majority of treated patients across all nivolumab doses. This was accompanied by significant increases in the expression of genes that are hallmarks of T_H1 inflammatory response and cytotoxic T-cell function, such as ICOS, IFNγ, granzymes, and perforin. Transcripts for T-cell receptor subunits (e.g., CD3γ, TCRα, TCRβ) rapidly and transiently decreased in whole blood after treatment with nivolumab, suggesting that nivolumab may prompt T cells to exit the periphery. These data suggest that nivolumab either increased tumor trafficking or infiltration of T cells, facilitated the expansion of T cells already within the tumor microenvironment, or both.

At the tumor site, increased transcription was observed for genes encoding CXCL9 and CXCL10, key IFNγ-regulated chemokines that guide the trafficking behavior of T cells, and for which elevated expression is a favorable prognostic factor in RCC (35). Two-thirds of patients had repeat core needle biopsies of the same soft-tissue lesions at baseline and on treatment, but we do not feel that multiple biopsies at the same site affected the conclusions of our study. In fact, a number of studies have shown the feasibility of undertaking repeat biopsies for the pharmacodynamic evaluation of targeted agents (36). Clinical response to ipilimumab, which promotes proliferation and activation of peripheral T cells, has been associated with pretreatment levels of CXCL9 and CXCL10 transcripts at the tumor site (21). The concomitant increased serum concentrations of CXCL9 and CXCL10 at C2D8, in the absence of detectable transcripts in whole blood, are in agreement with a model in which nivolumab induces CXCL9 and CXCL10 production by myeloid cells at inflammatory sites in tissue to recruit immunocompetent T cells to the tumor microenvironment (37). Notably, nivolumab produces a pharmacodynamic increase in transcripts for CTLA-4 itself and for a significant number of genes whose expression is associated with clinical response to ipilimumab (21), providing support for combination with anti-CTLA-4 therapies.

Transcripts for PD-L1 (CD274) are among those showing a pharmacodynamic effect in tumor biopsies. This PD-L1 expression likely originates from myeloid cell infiltrates because we show that nivolumab treatment does not appear to consistently modulate PD-L1 expression at the tumor cell surface. Nivolumab-induced T-cell reactivation was expected to increase IFNγ-regulated transcripts and proteins due to initiation of adaptive immune response. Using gene expression profiling, IFNγ transcripts and IFN-regulated transcripts were found to have increased in tumors, whereas only IFN-regulated transcripts were seen to rapidly and transiently increase in whole blood following nivolumab treatment. These effects are in line with observations from previous studies that demonstrated enhanced T-cell activation and increased expression of IFNγ at the tumor site after PD-L1/PD-L2 blockade or PD-1 inhibition (38–40).

Interestingly, the pharmacodynamic effects of nivolumab also include significant increases in the expression of genes linked to innate immunity and natural killer (NK) cell function (e.g., KLRB1/D1/G1, CD69, and NKG7), suggesting the testable hypothesis that nivolumab may boost T-cell–mediated antitumor immune activity through enhanced infiltration and/or stimulation of NK cells. Das and colleagues similarly found that nivolumab modulated genes involved in NK cell function (41). These data suggest that nivolumab may be effectively combined with NK cell–directed therapies, such as lirilumab, in metastatic RCC.

This analysis identified immune pharmacodynamic effects that are produced across nivolumab doses and in most patients. Additional analyses of these data are being conducted to investigate the relationship between the immunomodulatory activity of nivolumab and clinical outcomes.

T.K. Choueiri reports receiving commercial research support from Bristol-Myers Squibb and Pfizer, and is a consultant/advisory board member for Bayer, Bristol-Myers Squibb, Merck, Novartis, and Pfizer. M.N. Fishman reports receiving commercial research support from Merck, Pfizer, and Prometheus, and is a consultant/advisory board member for Alkermes and Prometheus. B. Escudier is a consultant/advisory board member for Bristol-Myers Squibb, Exelixis, GlaxoSmithKline, Novartis, and Pfizer. D.F. McDermott is a consultant/advisory board member for Bristol-Myers Squibb, Genentech, Merck, Novartis, and Pfizer. C.G. Drake has ownership interest (including patents) in AZ Medimmune, Bristol-Myers Squibb, and Potenza, is a consultant/advisory board member for Agenus, Bristol-Myers Squibb, Compugen, Eli Lilly, ImmunExcite, Merck, Potenza Biotherapeutics, Roche/Genentech, and Tizona Biotherapeutics, and reports receiving commercial research grants from Aduro Biotech, Bristol-Myers Squibb, and Janssen. H. Kluger reports receiving commercial research support from Bristol-Myers Squibb. W.M. Stadler is a consultant/advisory board member for Genentech/Roche, and reports receiving commercial research grants from Bristol-Myers Squibb. J.L. Perez-Gracia reports receiving commercial research grants from Bristol-Myers Squibb. D.G. McNeel has ownership interest (including patents) in and is a consultant/advisory board member for Madison Vaccines Inc. M.R. Harrison reports receiving speakers bureau honoraria from Genentech, is a consultant/advisory board member for Genentech and Novartis, and reports receiving commercial research grants from Argos, Bristol-Myers Squibb, Genentech, and Pfizer. E.R. Plimack is a consultant/advisory board member for Acceleron, Bristol-Myers Squibb, GlaxoSmithKline Novartis, and Pfizer. L. Appleman reports receiving commercial research grants from Astellas, Aveo, Bristol-Myers Squibb, Exelixis, Medivation, Novartis, and Pfizer. L. Fong reports receiving commercial research grants from AbbVie, Bristol-Myers Squibb, Janssen, Merck, and Roche/Genentech. L. Albiges is a consultant/advisory board member for Amgen, Bayer, Bristol-Myers Squibb, Novartis, Pfizer, and Sanofi. S. Srivastava has ownership interest (including patents) in Bristol-Myers Squibb. J.F. Kurland is an employee of MedImmune and has ownership interest (including patents) in Bristol-Myers Squibb. M. Sznol has ownership interest (including patents) in Adaptive Biotechnologies, and is a consultant/advisory board member for AstraZeneca, Bristol-Myers Squibb, Genentech-Roche, Immune Design, Kyowa-Kirin, Lion, Merck, Novartis, Pfizer, and Symphogen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T.K. Choueiri, M.N. Fishman, B. Escudier, D.F. McDermott, C.G. Drake, H. Kluger, B. Curti, M. Jure-Kunkel, J.F. Kurland, J.S. Simon, M. Sznol

Development of methodology: T.K. Choueiri, M.N. Fishman, J.L. Perez-Gracia, T.C. Young, M. Jure-Kunkel, J.F. Kurland

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.K. Choueiri, M.N. Fishman, D.F. McDermott, C.G. Drake, H. Kluger, W.M. Stadler, J.L. Perez-Gracia, D.G. McNeel, B. Curti, M.R. Harrison, E.R. Plimack, L. Appleman, L. Fong, L. Albiges, S. Srivastava, M. Jure-Kunkel, J.F. Kurland, J.S. Simon, M. Sznol

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.K. Choueiri, M.N. Fishman, B. Escudier, H. Kluger, J.L. Perez-Gracia, M.R. Harrison, E.R. Plimack, L. Appleman, L. Fong, L. Albiges, L. Cohen, T.C. Young, S.D. Chasalow, P. Ross-Macdonald, S. Srivastava, M. Jure-Kunkel, J.F. Kurland, J.S. Simon, M. Sznol

Writing, review, and/or revision of the manuscript: T.K. Choueiri, M.N. Fishman, B. Escudier, D.F. McDermott, C.G. Drake, H. Kluger, W.M. Stadler, J.L. Perez-Gracia, D.G. McNeel, B. Curti, M.R. Harrison, E.R. Plimack, L. Appleman, L. Fong, L. Albiges, L. Cohen, T.C. Young, S.D. Chasalow, P. Ross-Macdonald, S. Srivastava, M. Jure-Kunkel, J.F. Kurland, J.S. Simon, M. Sznol

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.K. Choueiri, T.C. Young, S.D. Chasalow, P. Ross-Macdonald, M. Jure-Kunkel

Study supervision: T.K. Choueiri, L. Cohen, S. Srivastava, J.F. Kurland, M. Sznol

The authors thank the patients and their families; clinical faculty and personnel of investigating sites; Jenny J. Kim, who was at Johns Hopkins University School of Medicine and the Sidney Kimmel Comprehensive Cancer Center, Baltimore, Maryland, at the time the study was conducted and data analyzed; The Center for Immuno-Oncology (CIO) at DFCI and its director, F. Stephen Hodi, MD; Quan Hong, PhD, for statistical analyses; Dako for collaborative development of the PD-L1 immunohistochemistry assay; and Bristol-Myers Squibb/Ono Pharmaceutical Company Limited for supporting the work. Writing/editorial assistance was provided by Shala Thomas, PhD, of PPSI, funded by Bristol-Myers Squibb.

The analyses and studies described in this report were funded by Bristol-Myers Squibb/Ono Pharmaceutical Company Limited.

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.