Baseline and post-treatment modulation of serum levels of chemokines, cytokines and other immune mediators were assessed by techniques that included ELISA or other multiplex-based assay methods. Primary analysis included IFN-gamma and IFN-gamma inducible factors, including chemokine [C-X-C motif] ligand 9 (CXCL9) and CXCL10

All response evaluable participants who were treated per the protocol

Biomarkers examined were percent positive CD8 and percent positive CD4, both using the Mosaic Singleplex IHC assay. Analyses are presented with the medians at baseline and on-treatment, rather than the median change because the baseline values differed across groups. Baseline was defined as the last non-missing value on or prior to the first dose of study therapy. Biopsies were also collected on treatment.

All response evaluable participants who were treated per the protocol

The assessment of safety was based on frequency of deaths and adverse events (AEs). AEs were graded for severity according to the NCI CTCAE version 4.0.

All participants who received treatment per the protocol

The assessment of safety was based on frequency of serious adverse events (SAEs) and adverse events (AEs) leading to discontinuation of study drug. AEs were graded for severity according to the NCI CTCAE version 4.0.

All participants who received treatment per the protocol

Abnormalities in hepatic parameters measured included those in aspartate aminotransferase (AST), alanine aminotransferase (ALT)and total bilirubin, with respect to upper limit of normal (ULN)

Participants who were treated per the protocol and who had at least one on-treatment measurement of the corresponding laboratory parameter

Abnormalities in thyroid parameters measured included those in thyroid stimulating hormone (TSH) levels with respect to upper limit of normal (ULN) and lower limit of normal (LLN)

Participants who were treated per the protocol and who had at least one on-treatment measurement of TSH

The objective response rate (ORR) was defined as the percentage of participants with a best overall response (BOR) of a complete response (CR) or partial response (PR) divided by the number of randomized participants in the population of interest. The BOR was defined as the participant's best response designation, over the study as a whole, recorded between the date of first study drug administration and the date of objectively documented progression per RECIST 1.1, with subsequent confirmation, or date of subsequent anti-cancer therapy, whichever occurred first in the study.

All participants who received treatment per the protocol

Median duration of response (mDOR) was calculated for subjects with BOR of CR or PR only, and is defined as time between the date of first documented objective response and the date of the first subsequent disease progression or death, whichever occurred first, if death occurred within 100 days after last dose of study medication.

All participants who received treatment per the protocol

Median time to response (mTTR) for a participant with a BOR of CR or PR is defined as the time from the first dosing date to the date of the first documented objective response (CR or PR).

All participants who received treatment per the protocol

Progression free survival (PFS) for a participant was defined as the time from the date of first dose of study medication to the date of the first documented disease progression, or death due to any cause, whichever occurred first, if death occurred within 100 days after last dose of study medication.

All participants who received treatment per the protocol

Time Frame: Part 1: Day 1, Day 15, Day 43 of cycle 1, Day 1 of cycle 2, Day 15 of cycle 3, every 16 weeks after cycle 3 up to 2 years, follow-up visit 1 (40-60 days after last treatment), and follow-up visit 2 (101-120 days since last treatment) Part 2, 3 and 4: Weeks 1, 3, 4, 7, 9, 10, 13, 25, 53, 79, 95 follow-up visit 1 (40-60 days after last treatment), and follow-up visit 2 (101-120 days since last treatment)

A subset of all participants who were treated per the protocol who had a baseline and at least 1 post-baseline ADA assessment for nivolumab and ipilimumab separately

For immunohistochemistry (IHC) measurements, to explore the PD-L1 expression as a potential predictive marker of clinical activity, PD-L1 expression status were derived from percent of tumor cells exhibiting cell surface staining for PD-L1 at baseline and/or in archived biopsy samples using verified and/or validated assays. In the case of multiple specimens, a subject would be identified as PD-L1 expression levels >= x%, where x% can be 10%, 5%, and/or 1% in any of the baseline and/or archived specimens. The association between PD-L1 expression status and/or level and clinical efficacy measures was assessed. The objective response rate (ORR) was defined as the number of subjects with a best overall response (BOR) of a complete response (CR) or partial response (PR) divided by the number of randomized subjects in the population of interest (all response-evaluable participants).

All participants who received treatment per the protocol

For immunohistochemistry (IHC) measurements, to explore the PD-L1 expression as a potential predictive marker of clinical activity, PD-L1 expression status were derived from percent of tumor cells exhibiting cell surface staining for PD-L1 at baseline and/or in archived biopsy samples using verified and/or validated assays. In the case of multiple specimens, a subject would be identified as PD-L1 expression levels >= x%, where x% can be 10%, 5%, and/or 1% in any of the baseline and/or archived specimens. Median duration of response (mDOR) was calculated for all response-evaluable participants with best overall response of CR or PR only, and is defined as time between the date of first documented objective response and the date of the first subsequent disease progression or death, whichever occurred first, if death occurred within 100 days after last dose of study medication.

All response evaluable participants who were treated per the protocol

For immunohistochemistry (IHC) measurements, to explore the PD-L1 expression as a potential predictive marker of clinical activity, PD-L1 expression status were derived from percent of tumor cells exhibiting cell surface staining for PD-L1 at baseline and/or in archived biopsy samples using verified and/or validated assays. In the case of multiple specimens, a subject would be identified as PD-L1 expression levels >= x%, where x% can be 10%, 5%, and/or 1% in any of the baseline and/or archived specimens. The association between PD-L1 expression status and/or level and clinical efficacy measures was assessed. The progression free survival rate (PFSR) for a subject was defined as the time from the date of first dose of study medication to the date of the first documented disease progression, or death due to any cause, whichever occurred first, if death occurred within 100 days after last dose of study medication.

All response evaluable participants who were treated per the protocol

The percentage of participants surviving to time t, where t is a specific length of time, eg, 12 months, which was determined by the available data for final analysis and was documented in the DPP. The percentage was calculated by the product-limit method (Kaplan-Meier estimate), which takes into account censored data. The overall survival rate (OSR) for a participant was defined as the time from the date of first dose of study medication to the date of death for any cause. A participant who had not died was censored at last known date alive.

All participants who were treated per the protocol

The Progression Free Survival Rate (PFSR) is defined as the probability of a participant remaining progression free or survival to time t, where t is equal to the specified timepoints. This probability will be calculated by the product limit method (Kaplan-Meier estimates) which takes into account censored data.

Medians and 95% confidence interval are estimated using the Kaplan-Meier method. If there is an insufficient number of events, the median and confidence intervals cannot be calculated.

All participants treated per the protocol