Racial/Ethnic Influences on CV and Renal Events in Diabetes
TREAT was a prospective, randomized, double-blind, placebo-controlled trial that enrolled 4,038 patients with type 2 diabetes mellitus, estimated glomerular filtration rate (eGFR) of 20 to 60 mL/min per 1.73 m (calculated using the 4-variable Modification of Diet in Renal Disease Study formula), and iron-replete anemia. Patients were randomized to darbepoetin alfa or placebo and followed up for mean of 2.4 years with no significant difference in coprimary end points but a 2-fold increase in stroke with darbepoetin alfa.
At the time of enrollment, patients were asked by study staff to define their race into 1 of 9 categories: Aborigine, American Indian or Alaskan, Asian, black or African American, Hispanic or Latino, Japanese, Native Hawaiian or Pacific Islander, white or Caucasian, or other. Patients were grouped into "black," "Hispanic," and "white" with these terminologies used for simplicity for the purpose of this analysis. The 115 (2.9%) of patients who were in any of the other 6 categories were excluded from the primary analysis; however, they were added to the white patient designation for a sensitivity analysis (online Appendix A). The term race is used to represent race and ethnicity for simplicity.
Blood and urine samples collected at baseline on all patients were analyzed for creatinine (used for eGFR calculation) every 2 to 4 weeks and other key variables detailed in online Appendix B. All patients provided written informed consent for participation in the primary trial.
End points were adjudicated by a central clinical end points committee. End-stage renal disease was defined as (a) the initiation of dialysis that persisted for >30 days, (b) the initiation of dialysis with death within 30 days, (c) physician recommendation to initiate dialysis with documented patient refusal, or (d) kidney transplantation. Death events were classified as CV and non-CV with subcategories or as unknown when insufficient information was provided. Definitions of all end points were previously reported.
Baseline characteristics were stratified according to racial classification: black, Hispanic, and white. Categorical variables were compared across racial categories using Fisher exact or χ tests, and continuous variables were compared using Kruskal-Wallis tests. Mixed-effects models were fitted that incorporated all observations available before the time of ESRD (or end of study visit) for systolic blood pressure, serum creatinine, eGFR, and log-transformed urinary protein-to-creatinine ratio. For each observation for each patient, the number of days before ESRD at which the observation was obtained was modeled via restricted cubic splines to allow for a flexible, potentially nonlinear relationship over time. Random-effect intercept terms at the patient level were incorporated to allow for within-patient correlation.
The univariate association between race and ESRD and between race and each component of the composite of death or nonfatal CV events was analyzed using Cox proportional hazards regression models, and incidence rates were calculated for each outcome grouped by race. Descriptive statistics were performed to assess number of patients who had a nonfatal CV event before ESRD. Four progressive multivariable models were performed with (a) ESRD and (b) death and nonfatal CV events as the outcomes to identify the independent association between race and these outcomes. Contents of the models are detailed in online Appendix B. Model 1 adjusted for age and sex. Model 2 added clinical variables and factors known to be predictive for developing ESRD in the TREAT renal model. Model 3 further adjusted for renal and interim time-updated factors. Model 4 further adjusted for interim nonfatal CV events. To further eliminate potential confounding, a sensitivity analysis was performed in which all patients were censored at the time of a nonfatal CV event.
TREAT was sponsored by Amgen. No additional funding was received to support this work. The academic leadership generated the concept for this manuscript. The primary data were at Brigham and Women's Hospital, and all authors had full access to the data for analysis. The first and senior authors wrote the manuscript with critical input from all other coauthors.
Methods
Study Population and Design
TREAT was a prospective, randomized, double-blind, placebo-controlled trial that enrolled 4,038 patients with type 2 diabetes mellitus, estimated glomerular filtration rate (eGFR) of 20 to 60 mL/min per 1.73 m (calculated using the 4-variable Modification of Diet in Renal Disease Study formula), and iron-replete anemia. Patients were randomized to darbepoetin alfa or placebo and followed up for mean of 2.4 years with no significant difference in coprimary end points but a 2-fold increase in stroke with darbepoetin alfa.
At the time of enrollment, patients were asked by study staff to define their race into 1 of 9 categories: Aborigine, American Indian or Alaskan, Asian, black or African American, Hispanic or Latino, Japanese, Native Hawaiian or Pacific Islander, white or Caucasian, or other. Patients were grouped into "black," "Hispanic," and "white" with these terminologies used for simplicity for the purpose of this analysis. The 115 (2.9%) of patients who were in any of the other 6 categories were excluded from the primary analysis; however, they were added to the white patient designation for a sensitivity analysis (online Appendix A). The term race is used to represent race and ethnicity for simplicity.
Blood and urine samples collected at baseline on all patients were analyzed for creatinine (used for eGFR calculation) every 2 to 4 weeks and other key variables detailed in online Appendix B. All patients provided written informed consent for participation in the primary trial.
Outcomes
End points were adjudicated by a central clinical end points committee. End-stage renal disease was defined as (a) the initiation of dialysis that persisted for >30 days, (b) the initiation of dialysis with death within 30 days, (c) physician recommendation to initiate dialysis with documented patient refusal, or (d) kidney transplantation. Death events were classified as CV and non-CV with subcategories or as unknown when insufficient information was provided. Definitions of all end points were previously reported.
Statistical Analysis
Baseline characteristics were stratified according to racial classification: black, Hispanic, and white. Categorical variables were compared across racial categories using Fisher exact or χ tests, and continuous variables were compared using Kruskal-Wallis tests. Mixed-effects models were fitted that incorporated all observations available before the time of ESRD (or end of study visit) for systolic blood pressure, serum creatinine, eGFR, and log-transformed urinary protein-to-creatinine ratio. For each observation for each patient, the number of days before ESRD at which the observation was obtained was modeled via restricted cubic splines to allow for a flexible, potentially nonlinear relationship over time. Random-effect intercept terms at the patient level were incorporated to allow for within-patient correlation.
The univariate association between race and ESRD and between race and each component of the composite of death or nonfatal CV events was analyzed using Cox proportional hazards regression models, and incidence rates were calculated for each outcome grouped by race. Descriptive statistics were performed to assess number of patients who had a nonfatal CV event before ESRD. Four progressive multivariable models were performed with (a) ESRD and (b) death and nonfatal CV events as the outcomes to identify the independent association between race and these outcomes. Contents of the models are detailed in online Appendix B. Model 1 adjusted for age and sex. Model 2 added clinical variables and factors known to be predictive for developing ESRD in the TREAT renal model. Model 3 further adjusted for renal and interim time-updated factors. Model 4 further adjusted for interim nonfatal CV events. To further eliminate potential confounding, a sensitivity analysis was performed in which all patients were censored at the time of a nonfatal CV event.
TREAT was sponsored by Amgen. No additional funding was received to support this work. The academic leadership generated the concept for this manuscript. The primary data were at Brigham and Women's Hospital, and all authors had full access to the data for analysis. The first and senior authors wrote the manuscript with critical input from all other coauthors.