The initial ESA dose is based on converting the prior ESA dose to the nearest available study rhEPO dose and is administered IV. The dose is adjusted thereafter in order to achieve the target range.

Oral placebo tablets will be taken from Week -4 up to randomization (Day 1).

Participants will receive supplemental iron therapy if ferritin is <=100 ng/mL or TSAT is <=20%. The investigator will choose the route of administration and dose of iron.

Time to first occurrence of adjudicated MACE or thromboembolic event (vascular access thrombosis, symptomatic deep vein thrombosis or symptomatic pulmonary embolism) was analyzed using a Cox proportional hazards regression model with with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period. This endpoint was adjusted for multiplicity using the Holm-Bonferonni method.

Time to first occurrence of adjudicated MACE or hospitalization for heart failure was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period. This endpoint was adjusted for multiplicity using the Holm-Bonferonni method.

Average monthly IV iron dose (milligrams) per participant from Day 1 to Week 52 was determined by calculating the total IV iron dose per participant from treatment start date + 1 to the earliest of (Week 52 visit date, first blood (red blood cell [RBC] or whole blood) transfusion date, and treatment stop date + 1 day) which corresponds to the time while the participant was on randomized treatment and before receiving a blood transfusion. This total IV iron dose was divided by (the number of days from treatment start date + 1 to the earliest of (Week 52 visit date, first blood transfusion date (RBC or whole blood), and treatment stop date +1) / 30.4375 days). This endpoint was adjusted for multiplicity using the Holm-Bonferonni method.

Time to first occurrence of adjudicated all-cause mortality was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the vital status for follow-up time period.

Time to first occurrence of adjudicated CV mortality was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Time to first occurrence of adjudicated MI (fatal and non-fatal) was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Time to first occurrence of adjudicated stroke (fatal and non-fatal) was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Number of participants with adjudicated MACE or hospitalization for heart failure (recurrent events analysis) is presented, categorized by number of occurrences of adjudicated MACE or hospitalization for heart failure per participant.

Time to first occurrence of adjudicated CV mortality or non-fatal MI was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

All-cause hospitalization events were hospital admissions recorded on the Hospitalization electronic case report form (eCRF) with a hospitalization duration >=24 hours. Time to first occurrence of all-cause hospitalization was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

All-cause hospital re-admissions within 30days are defined as hospital admissions recorded on hospitalization eCRF with hospitalization duration of >=24 hours and admission date within 30days following previous discharge date of all-cause hospitalization event, where previous hospitalization was >=24 hours. Time to first occurrence of all-cause hospital re-admission within 30days was analyzed using Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date)+1. Incidence rate per 100person years calculated as(100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event+cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Time to first occurrence of adjudicated MACE or hospitalization for heart failure or thromboembolic events were analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Time to first occurrence of adjudicated hospitalization for heart failure was analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Time to first occurrence of adjudicated thromboembolic events were analyzed using a Cox proportional hazards regression model with treatment group, dialysis type and region as covariates. Time to the first occurrence was computed as (event date minus randomization date) + 1. The incidence rate per 100 person years calculated as (100*number of participants with at least 1 event)/first event person-years) is presented along with 95% CI. First event person years=(cumulative total time to first event for participants who have the event + cumulative total of censored time for participants without the event)/365.25, based on the CV follow-up time period.

Blood samples were collected from participants for hemoglobin measurements. Change from Baseline was defined as post-Baseline value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date. Analysis was performed using mixed model repeated measures (MMRM) model fitted from Baseline up to Week 52, excluding values collected during the stabilization period, with factors for treatment, time, dialysis type, region, Baseline hemoglobin and Baseline hemoglobin by time and treatment by time interactions.

Mean Hgb during the evaluation period was defined as the mean of all evaluable Hgb values during the evaluation period (Week 28 to Week 52) including any evaluable unscheduled Hgb values that were taken during this time period. Hemoglobin responders were defined as participants with a mean Hgb during the evaluation period that falls within the Hgb analysis range of 10-11.5 g/dL.

Percentage of days for which a participant's Hgb was within the analysis range of 10-11.5 g/dL (both inclusive) during the evaluation period (Week 28 to Week 52), including any unscheduled evaluable Hgb values that were taken during this time period was calculated. Percentage of time in the analysis range during evaluation period is calculated as time in range during the evaluation period / [Earlier of (Date of the last evaluable Hgb value, Week 52 visit date) - Later of (Date of the first evaluable Hgb value that between Week 16 and Week 52 inclusive, Week 28 visit date)].

Percentage of days for which a participant's Hgb was within the analysis range of 10-11.5 g/dL (both inclusive) during the evaluation period (Week 28 to Week 52), including any unscheduled evaluable Hgb values that were taken during this time period was calculated. Percentage of time in the analysis range during evaluation period is calculated as time in range during the evaluation period / [Earlier of (Date of the last evaluable Hgb value, Week 52 visit date) - Later of (Date of the first evaluable Hgb value that between Week 16 and Week 52 inclusive, Week 28 visit date)].

Percentage of days for which a participant's Hgb was within the analysis range of 10-11.5 g/dL (both inclusive) during the maintenance period (Week 28 to end of study), including any unscheduled evaluable Hgb values that were taken during this time period was calculated. Percentage of time in the analysis range during maintenance period is calculated as time in range during the maintenance period / [Earlier of (Date of the last evaluable Hgb value, End of study date)- Later of (Date of the first evaluable Hgb value that is on or after week 16, Week 28 visit date)].

Percentage of days for which a participant's Hgb was within the analysis range of 10-11.5 g/dL (both inclusive) during the maintenance period (Week 28 to end of study), including any unscheduled evaluable Hgb values that were taken during this time period was calculated. Percentage of time in the analysis range during maintenance period is calculated as time in range during the maintenance period / [Earlier of (Date of the last evaluable Hgb value, End of study date)- Later of (Date of the first evaluable Hgb value that is on or after week 16, Week 28 visit date)].

SBP, DBP and MAP were measured in a semi-supine or seated position in the dialysis chair after at least a 5-minutes of rest. MAP is the average BP in an individual's arteries during a single cardiac cycle. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date. Analysis was performed using MMRM model with treatment group + time + dialysis type + region + Baseline value + Baseline value*time + treatment group*time, using an unstructured covariance matrix. Data for post-dialysis BP measurements have been presented.

SBP, DBP and MAP were measured in a semi-supine or seated position in the dialysis chair after at least a 5-minutes of rest. MAP is an average BP in an individual's arteries during a single cardiac cycle. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date. This analysis was carried out by using ANCOVA model with terms for treatment group, dialysis type, region and Baseline value. Data for post-dialysis BP measurements have been presented.

BP exacerbation was defined (based on post-dialysis) as: SBP >= 25 millimeter of mercury (mmHg) increased from Baseline or SBP >=180 mmHg; DBP >=15 mmHg increased from Baseline or DBP >=110 mmHg. The BP exacerbation events per 100 participant years was estimated using the negative binomial model with treatment, dialysis type and region as covariates and the logarithm of time on-treatment as an offset variable. Data for post-dialysis BP measurements have been presented.

BP exacerbation was defined as: SBP >= 25 mmHg increased from Baseline or SBP >=180 mmHg; DBP >=15 mmHg increased from Baseline or DBP >=110 mmHg. Number of participants with at least one BP exacerbation event is presented.

Percentage of participants permanently stopping randomized treatment due to meeting rescue criteria has been presented.

The SF-36 acute version 2 is a 36-item generic quality of life instrument designed to measure a participant's level of performance in the following 8 health domains: physical functioning, role-physical (role limitations caused by physical problems), social functioning, bodily pain, mental health, role-emotional (role limitations caused by emotional problems), vitality and general health. Each domain is scored from 0 (poorer health) to 100 (better health). The PCS is an average score derived from 4 domains (physical functioning, role-physical, bodily pain and general health) representing overall physical health. PCS ranges from 0 to 100; higher scores represent better health. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

The SF-36 acute version 2 is a 36-item generic quality of life instrument designed to measure a participant's level of performance in the following 8 health domains: physical functioning, role-physical (role limitations caused by physical problems), social functioning, bodily pain, mental health, role-emotional (role limitations caused by emotional problems), vitality and general health. Each domain is scored from 0 (poorer health) to 100 (better health). MCS is an average score derived from 4 domains (vitality, social functioning, role-emotional and mental health) representing overall mental health. MCS ranges from 0 to 100; higher scores represent better health. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

The SF-36 acute version 2 is a 36-item generic quality of life instrument designed to measure a participant's level of performance in the following 8 health domains: bodily pain, general health, mental health, role-emotional (role limitations caused by emotional problems), role-physical (role limitations caused by physical problems), social functioning (Social fun), physical functioning (Phy. fun) and vitality. Each domain is scored from 0 (poorer health) to 100 (better health). Each domain score ranges from 0 to 100, higher score indicates a better health state and better functioning. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

The SF-36 acute version 2 is a 36-item generic quality of life instrument designed to measure a participant's level of performance in the following 8 health domains: physical functioning, role-physical (role limitations caused by physical problems), social functioning, bodily pain, mental health, role-emotional (role limitations caused by emotional problems), vitality and general health. Each domain is scored from 0 (poorer health) to 100 (better health). Vitality score ranges from 0 to 100; higher scores represent better health. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

The SF-36 acute version 2 is a 36-item generic quality of life instrument designed to measure a participant's level of performance in the following 8 health domains: physical functioning, role-physical (role limitations caused by physical problems), social functioning, bodily pain, mental health, role-emotional (role limitations caused by emotional problems), vitality and general health. Each domain is scored from 0 (poorer health) to 100 (better health). Physical functioning score ranges from 0 to 100; higher scores represent better health. Change from Baseline was calculated as on-treatment visit value minus (-) Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

EQ-5D-5L is self-assessment questionnaire,consisting of 5 items covering 5 dimensions (mobility,self care,usual activities,pain/discomfort and anxiety/depression). Each dimension is measured by 5-point Likert scale (1=no problems, 2=slight problems, 3=moderate problems, 4=severe problems and 5=extreme problems). Responses for 5 dimensions together formed a 5-figure description of health state (e.g.11111 indicates no problems in all 5 dimensions). Each of these 5 figure health states were converted to a single index score by applying country-specific value set formula that attaches weights to dimensions and levels. Range for EQ-5D-5L index score is -0.594 (worst health) to 1 (full health state). Change from Baseline was calculated as on-treatment visit value-Baseline value. Baseline was latest non-missing pre-dose assessment on or before randomization date.

The EQ VAS records the respondent's self-rated health on a vertical VAS, ranging from 0 to 100, where 0 represents the worst health one can imagine and 100 represents the best health one can imagine. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

The PGI-S is a 1-item questionnaire designed to assess participant's impression of disease severity on a 5-point disease severity scale (0=absent, 1=mild, 2=moderate, 3=severe, or 4=very severe). A higher score indicated worse outcome. Change from Baseline was calculated as on-treatment visit value minus Baseline value. Baseline was defined as the latest non-missing pre-dose assessment on or before the randomization date.

Participants received placebo tablets orally once daily in run-in period from Week -4 up to randomization (Day 1) and subsequently received treatment with rhEPO. Participants on hemodialysis received epoetin alfa as intravenous (IV) injection once weekly or three-times weekly with total weekly dose levels ranging from 1500 to 60,000 Units. Participants on peritoneal dialysis received subcutaneous (SC) injection of darbepoetin alfa every 1, 2, or 4 weeks with 4-weekly total dose levels ranging from 20 to 400 microgram (mcg). Darbepoetin could be given by IV injection for peritoneal dialysis participants switching to hemodialysis. Study treatment was dose-titrated to achieve and maintain hemoglobin in the target range (10 to 11 g/dL) and administered until the required number of MACE events occurred, at approximately 45.1 months of randomized treatment.