Abstract
ABSTRACT. Although 11 million people in the United States have chronic renal insufficiency, little is known about ethnic/racial disparities for early-onset renal impairment. This study sought to determine whether there is an independent association between race/ethnicity and early-onset renal impairment and to identify other risk factors that might account for observed disparities. All Coronary Artery Risk Development in Young Adults subjects in which serum creatinine was measured at the year 15 examination were identified (n = 3554), excluding those who were pregnant at year 15. Potential risk factors at study entry (ages 18 to 30 yr, 1985 to 1986) included age, weight, gender, race/ethnicity, glucose, uric acid, and systolic BP. Renal impairment was defined as creatinine ≥1.5 mg/dl for men and ≥1.2 mg/dl for women at year 15 (ages 33 to 45 yr). Fifty-two (2.7%) women and 39 (2.4%) men had renal impairment at the year 15 examination. In bivariate analyses, the odds of renal impairment among black women was estimated to be 2.4-fold that of white women, and among black men, the odds of renal impairment were 9.0-fold that of white men. In multivariate analysis, the odds of an elevated creatinine among black women compared with white women reduced to a nonsignificant 1.5-fold, whereas among men, the odds of an elevated creatinine among blacks was 11.4-fold that of whites. Although adjustment for baseline glucose levels accounted for much of the association between ethnicity and elevated creatinine among women, adjustment for weight, systolic BP, uric acid, glucose, and socioeconomic status did not account for the association between ethnicity and renal impairment among men. The data suggest that there are ethnic differences in the development of early-onset renal dysfunction. Among women, these differences are modest and largely accounted for by differences in glucose levels early in adult life. Differences in race/ethnicity related risk of early-onset renal impairment are particularly large among men and are not accounted for by the metabolic or socioeconomic factors evaluated. E-mail: cos@u.washington.edu
Renal disease is an important public health problem, disproportionately affecting certain ethnic groups. In particular, black race/ethnicity is associated with more rapid progression of renal disease (1,2⇓) than is white race/ethnicity. Although blacks account for 10% of the general population, they account for >30% of the ESRD population (2–8⇓⇓⇓⇓⇓⇓). The reasons for these disparities are not well understood but may be due to a greater risk of diabetes and hypertension, lower socioeconomic status, and poorer access to health care among blacks. Little is known about the contribution of race/ethnicity to the development of early-onset renal impairment (9) or about the factors that might be responsible for observed associations.
The Coronary Artery Risk Development in Young Adults (CARDIA) Study is a longitudinal study of coronary artery disease risk development in a biethnic population of young adults. CARDIA provides a unique opportunity to identify prospectively risk factors present in early life that predict later development of renal disease. The primary objective of this study was to determine whether there is an independent association between race/ethnicity and early-onset renal impairment among both men and women and to identify other risk factors that might account for observed disparities.
Materials and Methods
Setting and Subjects
Details of the CARDIA study design are described elsewhere (10,11⇓). Briefly, subjects ages 18 to 30 yr were recruited by clinical centers from the populations of four geographic locations (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA) by a stratified random sampling scheme to ensure balance across gender and ethnicity (blacks and whites only). Baseline examinations were performed on 5115 (51%) of eligible subjects contacted in 1985 to 1986. To date, the CARDIA clinical centers have completed five subsequent examinations, after 2, 5, 7, 10 (79% retention), and 15 (73% retention) years of follow-up. For the purposes of this study, we identified all CARDIA subjects in whom serum creatinine was measured at the year 15 examination (n = 3554).
Subject Characteristics
Sociodemographic history was obtained by questionnaire. Age, gender, race/ethnicity, cigarette smoking status (current, past, never), and illicit drug use (yes, no) were determined from the baseline examination. Years of education (<12, 12 to 16, and >16 yr) and employment status (full time, less than full time) were determined from the year 10 examination because baseline information from subjects whose ages ranged between 18 and 30 yr would have inconsistently reflected the impact of these variables at follow-up.
Anthropometric assessments from the baseline and year 10 examinations were used in analyses. Body weight was measured with participants wearing light clothing using a calibrated balance beam scale. Skinfold measurements were obtained as a surrogate measure of body fat and obtained at standard sites using duplicate measurements and the Harpenden calipers (Quinton, Seattle, WA). Averages were used to obtain the sum of skinfolds (11). Waist circumference was measured laterally midway between the iliac crest and the lowest lateral portion of the rib cage and anteriorly midway between the xiphoid process of the sternum and the umbilicus.
BP measurements obtained from the baseline and year 10 examinations were used for this analysis. Three BP were obtained with a random zero sphygmomanometer three times after a 5-min rest, with 1-min intervals between measures; the average of the last two measurements was used in analyses.
Dietary history, physical activity, and fitness were assessed at the baseline examination. Centrally trained and certified interviewers obtained a detailed, quantitative diet history (12). Dietary data were processed with the University of Minnesota Nutrition Coordinating Center Tape 10 nutrient database (13). Physical activity was assessed by questionnaire, and a score was calculated by summing the intensity weighted frequencies of 13 activities reported in the previous year (14). Physical fitness was assessed by calculating the total duration of exercise in seconds obtained from a symptom-limited, graded treadmill exercise test with a modified Balke protocol (15).
When available, we used laboratory data from the baseline examination including plasma cholesterol, fasting insulin, uric acid, glucose, serum calcium, phosphate, and total protein. Glucose from the year 10 examination was used to calculate change in glucose. Comparable measures of baseline and year 15 serum creatinine were not available; therefore, the current analysis investigates predictors of year 15 creatinine only. Fibrinogen (16) and Lp(a) (16) from the year 5 examination and C-reactive protein from the year 15 examination were used in the analysis. Albumin/creatinine excretion was assessed at the years 10 and 15 examinations using a single untimed urine sample with creatinine adjusted for race and gender as described previously (17). Microalbuminuria was defined as a ratio of albumin to creatinine ≥25 mg/g after adjustment for gender and race (A/kC, where k = 0.68*0.88 in black men, 0.88 in black women, 0.68 in white men, and 1 in white women (17)). Diabetes was defined as an 8-h fasting glucose level of >126 mg/dl and/or use of antidiabetic medication. At each examination, blood samples were drawn the morning after an overnight fast (at least 8 h) using EDTA-containing tubes. Plasma and serum were stored at −70° before assay.
Follow-up and Classification of Events
The primary outcome assessed was serum creatinine at the year 15 examination. Although GFR is the gold standard measure of renal function, there are no direct measures of GFR in CARDIA. We chose not to use formulas to estimate GFR, such as the Modification of Diet in Renal Disease or Cockcroft-Gault formulas (18,19⇓), for several reasons. First, because age, gender, race, and weight were predictors of interest and also used in the formulas (20), the use of these formulas would potentially induce a relationship between renal function and the covariates of interest. Second, the use of the formulas assumes an interaction between the components of the formula that may or may not exist in this population. Modeling the interaction with the available data is a more accurate way to determine the functional form of the relationship. Finally, these formulas have not been validated in young adults. We used cut points for creatinine to provide less biased estimates of renal function because several very high creatinine values among black individuals with diabetes weighted the results when creatinine was modeled as a continuous variable. Because there are not standard creatinine values that define renal disease, clinically meaningful cut points that minimized false positives were chosen. Sensitivity analyses using different cut points were also performed. Different cut points were used for men and women because serum creatinine level is dependent on muscle mass and muscle mass is lower among women than men.
Statistical Analyses
Logistic regression was used to estimate the association between race/ethnicity and year 15 creatinine after adjustment for possible confounders. Because of the different cut points used to define renal dysfunction across gender and the intent to assess interactions by race, separate logistic regression models were fit for men and women. Covariates included in each multiple regression model were chosen on the basis of their potential for confounding the association between race/ethnicity and creatinine or on the basis of their potential to predict change independently in renal function. Multiplicative interactions between the primary predictors listed above and race/ethnicity were evaluated to identify possible effect modification by race. Residual diagnostics were analyzed to investigate potential outlying data points and to determine the appropriate functional form of adjustment covariates. Although outlying data points were present in the analysis, the measurements were not found to be highly influential and could not be attributed to obvious data entry error and hence were not excluded.
Results
Characteristics of the study sample, stratified by gender and creatinine level measured at the year 15 examination, are presented in Table 1. Fifty-two (2.7%) women had a serum creatinine of 1.2 mg/dl or higher, and 39 men (2.4%) had a serum creatinine of 1.5 mg/dl or higher at the year 15 examination. Both men and women with high creatinine levels were more likely to be black. The prevalence of elevated creatinine among white women was 1.6% (16 of 987) and 3.7% (36 of 963) among black women, whereas the prevalence of elevated creatinine among white men was 0.6% (5 of 899) and 4.8% (34 of 705) among black men. Both men and women with high creatinine were more likely to be diabetic and have microalbuminuria at year 10. Among women, high creatinine was also associated with greater baseline weight, systolic BP (SBP), and C-reactive protein levels at the year 15 examination. Among men, high creatinine was also associated with greater change in SBP between baseline and year 10, level of education at the year 10 examination, and higher baseline uric acid levels. In most cases in which a variable was statistically significant in only one gender, the estimated direction of association was the same in the other gender.
Table 1. Patient characteristics by gender and serum creatinine level at year 15: The CARDIA study, 1985–2001
The association between race and creatinine was found to differ significantly by gender (P = 0.005). Therefore, we undertook separate analyses for men and women. Table 2 presents univariate logistic regression estimates modeling the association between selected risk factors and the odds of elevated year 15 serum creatinine among women and men. The unadjusted odds of renal impairment among black women was estimated to be 2.4-fold that of white women. Important risk factors for renal impairment for women included baseline glucose, weight, and SBP and change in SBP between baseline and the year 10 examination. The unadjusted odds of renal impairment among black men were 9.0-fold that of white men. In addition to ethnicity, important risk factors from bivariate analyses included baseline glucose and uric acid and change in SBP and glucose between baseline and year 10 examinations.
Table 2. Gender-stratified univariate logistic regression estimates modeling the probability of elevated creatinine at year 15: the CARDIA study, 1985–2001a
The multivariate logistic regression estimates modeling the association between selected risk factors and the odds of elevated year 15 serum creatinine are presented in Table 3. After adjustment for age, weight and change in weight, SBP and change in SBP, glucose and change in glucose, and uric acid, the risk of elevated creatinine among black women compared with white women reduced to 1.5-fold and was no longer significant. Baseline glucose and change in SBP were the only significant predictors of elevated creatinine among women after adjustment for other risk factors. After adjustment, the risk of increased serum creatinine remained higher among black men when compared with white men (odds ratio, 11.4). Adjustment for age, weight, SBP, uric acid, and glucose did not account for the association between ethnicity and renal impairment among men. In addition to ethnicity, baseline glucose, uric acid, and change in SBP and glucose between baseline and year 10 examinations were independently associated with elevated serum creatinine among men.
Table 3. Gender-stratified adjusted logistic regression estimates modeling the probability of elevated creatinine at year 15: the CARDIA study, 1985–2001a
Further analyses sought to identify additional explanatory variables that might account for the association between black ethnicity and high creatinine among men by adding additional variables to models presented in Table 3. Adjustment for year 10 education and employment status did not alter the relationship between race/ethnicity and high creatinine. Adjustment for baseline serum calcium, phosphorous, and total protein; year 5 fibrinogen and Lp(a); or year 15 C-reactive protein levels also failed to account for the association between race/ethnicity and elevated creatinine. The addition of baseline triceps skinfold thickness, waist circumference, total exercise intensity score, and dietary protein intake only minimally reduced the association between black ethnicity and high creatinine.
Discussion
We found an elevated risk for renal impairment in blacks compared with whites, especially among men. The association between ethnicity and renal impairment in men was independent of other risk factors, such as weight, SBP, and glucose assessed 15 yr before creatinine measurement and the change in these risk factors over the initial 10 yr of follow-up. However, the association between ethnicity and renal impairment in women, which was found in bivariate analyses, was no longer significant after adjustment for baseline glucose.
A reduction in the relative risk of renal impairment after adjusting for baseline glucose among women suggests that a higher prevalence of diabetes or glucose intolerance may account at least partially for the differences in renal impairment among black women. The risk of diabetes is greater among blacks than whites, and, in the general population, the risk of renal dysfunction is 2 to 4 times higher among black individuals with diabetes than among white individuals with diabetes (21,22⇓). However, given the relative rarity of renal insufficiency in our generally healthy study sample, this study may have had limited power to detect the moderate associations of race/ethnicity with renal dysfunction noted among women.
The reason that black men may be at greater risk for the development of renal disease is not known. We were unable to demonstrate that disparities in a variety of factors, including education level, employment status, glycemia, or BP, accounted for this association. However, it is possible that unmeasured socioeconomic disparities or environmental factors may account for the association. Further studies are necessary to explore this possibility. Given the recent identification of candidate genes that mediate hypertension, formation of extracellular matrix proteins, and lipid metabolism (23–29⇓⇓⇓⇓⇓⇓), it is also possible that there may be some contribution of genetic factors to the observed ethnic differences in the development of renal impairment.
This study had certain limitations. In an ideal setting, baseline measurements of serum creatinine would have been available to identify subjects with preexisting renal disease at baseline. With this said, participants were only 18 to 30 yr old at baseline, and previous reports suggest that only 0.4 to 0.9% of people in this age range have estimated GFR consistent with significant renal disease (<60 ml/min) (30). In addition, recalibration of baseline creatinine values indicated that there were few subjects with renal dysfunction at baseline. Second, because different cut points were chosen for men and women to account for gender differences in serum creatinine, comparisons of risk of renal impairment between genders should be done with caution. Third, although cut points for creatinine to define renal impairment were chosen to reflect clinically significant renal disease, the ethnic differences identified were also present using lower or higher cut points for creatinine in a sensitivity analysis (data not shown). For example, for levels of serum creatinine >1.1 mg/dl in men and at 0.9 mg/dl in women, the prevalence of renal impairment remained lower for whites than for blacks. It is also possible that factors that were not assessed in this study, such as disparities in access to health care, control of hypertension, and treatment of comorbid conditions, may account for disparities in early-onset renal impairment observed in this study. Ethnic disparities in renal dysfunction may also, in part, be accounted for by the inability to completely account for differences in creatinine production between ethnic groups by adjustment for body fat and weight. Finally, there are a limited number of subjects with renal insufficiency even after 15 yr of follow-up. However, the large population of young adults followed for 15 yr provided a unique opportunity to identify early markers of early-onset renal dysfunction among a biracial cohort of young adults. In addition, clinical and biochemical evaluation was extensive and allowed us to assess a variety of variables that may be relevant to renal disease.
We conclude that there are ethnic differences in the development of early-onset renal dysfunction. Among women, these differences are modest and largely accounted for by differences in glucose levels at the baseline examination. Differences in race/ethnicity-related risk of early-onset renal impairment are particularly large among men and are not accounted for by the metabolic or socioeconomic factors evaluated in this analysis. Future studies should confirm and also identify the basis for the particularly large risk of early-onset renal dysfunction among black men.
Acknowledgments
This study was supported by a Veterans Administration Career Development Award. The data in this manuscript were presented at the American Heart Association Epidemiology Council Meeting 2002.
- © 2003 American Society of Nephrology