OS was defined as the time from the date of randomization to the date of death due to any cause. OS (in months) was calculated as (date of death - first randomization date +1)/30.4. For participants still alive at the time of the analysis, the OS time was censored on the last date they were known to be alive. All participants were followed up for survival at least every 3 months after discontinuing study treatment until at least two years after randomization of the last participant.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

PFS was defined as time from randomization to first documented objective tumor progression or to death due to any cause, whichever occurred first. PFS (in months) was calculated as (first event date - first randomization date +1)/30.4. Tumor progression was determined from oncologic assessment data (where data meet the criteria for progressive disease [PD]), or from adverse event (AE) data (where the outcome was death). As per response evaluation criteria in solid tumors (RECIST) 1.1, progression was defined as greater than or equal to (>=) 20% increase in sum of longest dimensions of target lesions or appearance of one or more new target lesions and unequivocal progression of existing non-target lesions, or appearance of 1 new non-target lesions. Participants discontinuing study treatment without documented evidence of PD were to be followed up at least every 8 weeks after discontinuing study treatment until disease progression, or initiation of another anticancer treatment.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

ORR was defined as the percentage of participants with confirmed complete response (CR) or confirmed partial response (PR) according to the RECIST 1.1. CR was defined as the disappearance of all target lesions. Any pathological lymph nodes (whether target or non-target) must decrease to normal (short axis <10 millimetres [mm]). PR was defined as a 30% decrease in the sum of the longest dimensions of the target lesions taking as a reference the baseline sum longest dimensions.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

TTP was defined as the time from randomization to first documentation of objective tumor progression. If tumor progression data included more than 1 date, the first date was used. TTP (in months) was calculated as (first event date - first randomization date +1)/30.4.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

DR was defined as the time from the first documentation of objective tumor response (CR or PR) that was subsequently confirmed to the first documentation of PD or to death due to any cause, whichever occurs first. If tumor progression data included more than 1 date, the first date was to be used. DR (in months) was to be calculated as (the end date for DR - first CR or PR that was subsequently confirmed +1)/30.4.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received. DR was calculated for the subgroup of FAS participants with objective response.

CBR was defined as the percentage of participants with confirmed CR or confirmed PR or a best response of stable disease >=8 weeks according to RECIST 1.1 criteria, relative to all randomized participants who had baseline measurable disease. Confirmed responses were defined as those that persisted on repeat imaging study >=4 weeks after the initial documentation of response. Participants who did not have on study radiographic tumor re-evaluation or who died, progressed, or dropped out for any reason prior to reaching a CR, PR, or stable disease were counted as non-responders in the assessment of CBR. A participant who initially met the criteria for a PR and then subsequently became a confirmed CR was to be assigned a best response of CR.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing. In the below table, 4 participants in Child-Pugh A and 1 participant in Child-Pugh B were not reported due to nonestimable half-life.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing. In the below table, 4 participants in Child-Pugh A and 1 participant in Child-Pugh B were not reported due to nonestimable half-life.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing. In the below table, 4 participants in Child-Pugh A and 1 participant in Child-Pugh B were not reported due to nonestimable half-life.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Axitinib samples were to be collected from all participants on Cycle 1 Day 15 at the following time points: pre-dose, 1, 2, 3, 4, 6 and 8 hours after axitinib dosing. The PK parameter, Vz/F has been presented in this outcome measure. In the below table, 4 participants in Child-Pugh A and 1 participant in Child-Pugh B were not reported due to nonestimable half-life.

The PK concentration set included all participants who were treated and had at least 1 measured concentration on at least 1 day of PK assessment. The PK parameter analysis set included all participants treated who had at least 1 estimated PK parameter of primary interest.

Plasma soluble proteins interleukin-6 (IL-6), E-Selectin, interleukin-8 (IL-8), hepatocyte growth factor (HGF), matrix metalloproteinase-2 (MMP-2), stem cell factor (SCF), angiopoietin-2 (Ang-2), vascular endothelial growth factor-A (VEGF-A), vascular endothelial growth factor-C (VEGF-C), soluble vascular endothelial growth factor receptor 2 (sVEGFR2), soluble vascular endothelial growth factor receptor 3 (sVEGFR3), stromal cell-derived factor-1 (SDF1), neutrophil gelatinase-associated lipocalin (NGAL), migration inhibitory factor (MIF), c-MET, regulated upon activation normal T cell expressed and presumably secreted (RANTES), and monocyte chemotactic protein-3 (MCP-3) were only measured in randomized participants.

The soluble protein analysis set included all participants in the safety analysis set who had a Baseline soluble protein assessment. Safety analysis population included all randomized participants who received at least 1 dose of study drug with treatment assignments designated according to actual study treatment received.

A 5 millilitres (mL) whole blood sample was collected from all randomized participants to evaluate the miRNA transcripts.

The miRNA analysis set included all participants in the safety analysis set who had a baseline miRNA assessment. Safety analysis population included all randomized participants who received at least 1 dose of study drug with treatment assignments designated according to actual study treatment received.

FACT-Hep consists of 27-item FACT-G, and 18-item Hepatobiliary Subscale. FACT-Hep questionnaire uses 5-point Likert rating scale, range '0'-not at all to '4'. FACT-Hep total score ranges from 0 to 180, where highest score represents maximum achievable quality of life. Domains of FACT-G include Physical Well-Being (PWB), Social/Family Well-Being (SWB), Emotional Well-Being (EWB) and Functional Well-Being (FWB). Hepatobiliary disease specific items include: swelling or cramps, losing weight, gastrointestinal (GI)-related questions, lack of energy, side effects, pain, fatigue, usual activities, jaundice, fevers, itching, taste of food and chills. Eight of the items (pain, back pain, stomach pain/discomfort, lack of energy, fatigue, nausea, weight loss and jaundice) make up FACT-Hepatobiliary Symptom Index (FHSI-8), and are considered to be symptoms specific to hepatobiliary cancer. Table below included mixed effect model estimated average based on all observed values/time points.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

FACT-G is core questionnaire of Functional Assessment of Chronic Illness Therapy (FACIT) measurement system to evaluate QoL in cancer population. FACT-G consisted of 27 questions grouped in 4 domains of general Health-Related QoL (HRQoL): PWB, SWB, EWB and FWB; each ranging from 0 (not at all) to 4 (very much). FACT-G ranged between 0 and 108. Since questions could be reversed coded, as appropriate, before calculating FACT-G, 0 and 108 could be considered worst and best health states. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

The FACT-Hep includes the FACT-G and a hepatobiliary module. The hepatobiliary disease specific items include: swelling or cramps, losing weight, GI related questions, lack of energy, side effects, pain, fatigue, usual activities, jaundice, fevers, itching, taste of food and chills. Eight of the items (pain, back pain, stomach pain/discomfort, lack of energy, fatigue, nausea, weight loss, and jaundice) make up the FHSI-8, and are considered to be symptoms specific to hepatobiliary cancer. FHSI-8 total score ranges from 0 to 32 where "0" is a severely symptomatic participant and the highest score indicates an asymptomatic participant. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

FACT-G is core questionnaire of Functional Assessment of Chronic Illness Therapy (FACIT) measurement system to evaluate QoL in cancer population. FACT-G consisted of 27 questions grouped in 4 domains of general HRQoL: PWB, SWB, EWB and FWB. Each of the individual subscale, except EWB has 7 items and each integer scored 0 to 4 making a maximum possible score of 28 (range 0 to 28). EWB has 6 items and each integer scored 0 to 4 making a maximum possible score of 24 (range 0 to 24). For all the 4 scales, higher values correspond to better health. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

This subscale consists of 18 items rated on a scale from '0' - not at all to '4' - very much regarding how much each item was present in the last 7 days. FACT-Hep-CS18 total score ranges from 0 to 72. The higher score reflects better QoL or fewer symptoms. The 18 items of this scale are associated with hepatocellular carcinoma. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

The trial outcome index is defined to be the sum (PWB+FWB+HepCS), making it 32 items altogether. Each ranges from '0' - not at all to '4' - very much regarding how much each item was present in the last 7 days. FACT Hep -TOI total score ranges from 0 to 128, where the highest score represents a maximum achievable quality of life. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

TTD analysis was performed for FHSI-8. Time to deterioration was defined as the time between date of randomization and date of the event.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

EQ-5D: participant rated questionnaire to assess health-related quality of life in terms of a single utility score. Health State Profile component assesses level of current health for 5 domains: mobility, self-care, usual activities, pain and discomfort, and anxiety and depression; 1 indicates better health state (no problems); 3 indicates worst health state. Scoring formula developed by EuroQol Group assigns a utility value for each domain in the profile. Score is transformed and results in a total score range -0.594 to 1.000; higher score indicates a better health state. The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

EQ-5D VAS in rates the participant's overall health status using values from 0 (worst imaginable) to 100 (best imaginable). The below table included the model estimated average based on all the observed values/time points. The mixed effect model was used.

FAS included all randomized participants, and participants were classified according to the randomized treatment arm regardless of what treatment, if any, was received.

Number of Child-Pugh Class B (score 7) participants with DLT was evaluated during Cycle 1 of treatment in the non-randomized portion of the study.

Participants with Child-Pugh Class B, score 7 are only included in this analysis.

An AE was an untoward medical occurrence in a participant who received study treatment without regard to possibility of causal relationship. Serious adverse event (SAE) was an AE resulting in any of the following outcomes: death, initial or prolonged inpatient hospitalization, life-threatening experience, persistent or significant disability/incapacity, congenital anomaly. Treatment-emergent AEs were those with initial onset or that worsen in severity after the first dose of study medication. The grade of an AE was determined according to Common Terminology Criteria for Adverse Events (CTCAE) Version 3.0.

The safety analysis population included participants who received at least 1 dose of study medication with treatment assignments designated according to actual study treatment received.

Treatment-related AE was any untoward medical occurrence in a participant with causal relationship to the study drug. An SAE was an AE resulting in any of the following outcomes or deemed significant for any other reason: death; initial or prolonged inpatient hospitalization; life-threatening experience (immediate risk of dying); persistent or significant disability/incapacity; congenital anomaly. The grade of an AE was determined according to CTCAE Version 3.0.

The safety analysis population included participants who received at least 1 dose of study medication with treatment assignments designated according to actual study treatment received.

An AE was an untoward medical occurrence in a participant who received study treatment without regard to possibility of causal relationship. SAE was an AE resulting in any of the following outcomes: death, initial or prolonged inpatient hospitalization, life-threatening experience, persistent or significant disability/incapacity, congenital anomaly. Treatment-emergent AEs were those with initial onset or that worsen in severity after the first dose of study medication. The grade of an AE was determined according to CTCAE Version 3.0.

The safety analysis population included all randomized participants who received at least 1 dose of study medication with treatment assignments designated according to actual study treatment received.

Treatment-related AE was any untoward medical occurrence in a participant with causal relationship to the study drug. An SAE was an AE resulting in any of the following outcomes or deemed significant for any other reason: death; initial or prolonged inpatient hospitalization; life-threatening experience (immediate risk of dying); persistent or significant disability/incapacity; congenital anomaly. The grade of an AE was determined according to CTCAE Version 3.0.

The safety analysis population included all randomized participants who received at least 1 dose of study medication with treatment assignments designated according to actual study treatment received.