As the burden of end-stage renal disease (ESRD) in the United States (US) continues to increase [1], it remains important to understand the role of risk factors such as body mass index (BMI), which change during the course of life and which may interact with race to modify ESRD risk. Previous studies have demonstrated strong and significant associations between higher BMI in middle aged adults (and adolescents) and ESRD, which may be explained in part by associations of obesity with inflammation, diabetes and high blood pressure, all of which are risk factors for adverse renal outcomes [2–9]. Given these strong associations, the rising prevalence of early obesity [10] is likely to impact the already substantial health burden associated with ESRD.

A substantially higher incidence of ESRD among blacks than whites has been consistently reported, including among participants of the Southern Community Cohort Study (SCCS), an ongoing prospective cohort study of black and white adults in the southeastern US [11]. The association persists after adjustment for socioeconomic variables and known ESRD risk factors [12–17]. However, to our knowledge, only one study [8], which focused on midlife BMI, has investigated whether there may be a differential relationship between BMI and ESRD in blacks and whites. For any given value of BMI, whites tend to have higher percent body fat than blacks [18], and adiposity has been demonstrated to have different metabolic effects in whites and blacks [19–22].

Therefore, we investigated whether the association between BMI in both early adulthood and midlife and ESRD is stronger among whites compared to blacks, using a nested case–control study within the SCCS.

Among the 2,528 SCCS participants included in this study, 10.4 % were white and 89.6 % were black, and 45.2 % were men. The mean age at enrollment was 53.7 years. The BMI at enrollment and at age 21 were significantly correlated (person correlation coefficient, r = 0.38, p < 0.0001), with mean BMI at enrollment (31.2 ± 7.7 kg/m²) substantially higher than mean BMI at age 21 (23.9 ± 5.5 kg/m²).

The results of the stratified analyses (Table 2) show a similar pattern of increasing odds for ESRD with increasing BMI in both blacks and whites. However, the association was substantially stronger in whites. In blacks, the adjusted ORs were 1.41 (95 % CI: 1.09, 1.83) among overweight persons and 2.56 (95 % CI: 1.88, 3.47) among obese individuals. The corresponding ORs in whites were 2.13 (95 % CI: 0.92, 4.93) and 7.46 (95 % CI: 2.90, 19.21). The p-value for interaction in adjusted models revealed a marginally significant effect modification by race of the association between BMI at age 21 and ESRD (p = 0.05).

The plot of the predicted probabilities of incident ESRD vs BMI at age 21 showed a monotonic increase in the probability of ESRD after a BMI cutoff ≈ 25 kg/m² in both whites and blacks but the slope of the curve for whites appeared greater (Fig. 1). In sensitivity analyses excluding persons with BMI <18.5 kg/m² and >35 kg/m², the curves for both blacks and whites still showed monotonic increases in the probability of ESRD with increasing BMI at age 21 and the curve for whites was steeper than that of blacks after a BMI of ≈ 27 kg/m² (Additional file 1) i.e. the difference in slopes appeared greater in the region were we have the most data: BMI = 18.5–35 kg/m².

In further analyses subdividing the obese groups, among whites, odds of ESRD increased with increasing class of obesity, and those with class III obesity had a significant greater than 3-fold increased odds (OR: 3.31; 95 % CI: 1.08, 10.12) of ESRD. In contrast, among blacks, those with class III obesity did not have significantly increased odds (OR: 1.09; 95 % CI: 0.76, 1.55) of ESRD (Table 3).

In analyses restricted to cases (and their matched controls) with more than 12 months of follow-up, we observed non-significant increased odds for ESRD in whites in the overweight, and class I and II obesity categories (compared to the normal weight), and a significant 4-fold (OR: 3.99; 95 % CI: 1.11, 14.35) increased risk for those in the class III obesity category. In contrast, there was no increased risk among blacks with class III obesity (OR: 1.02; 95 % CI: 0.70, 1.52). In additional analyses restricted to cases with more than 2 or more than 3 years of follow up, the associations with ESRD in the highest BMI categories were more pronounced, but again the point estimates were substantially higher in whites compared to blacks. In particular, in the analyses restricted to those with more than 3 years of follow-up, there was a non-significant 5-fold (OR: 5.04; 95 % CI: 0.32, 79.49) increased odds of ESRD among whites with class III obesity, as compared to a corresponding non-significant 45 % (OR: 1.45; 95 % CI: 0.84, 2.51) increased odds of ESRD among blacks (data not shown).

In this case–control study nested within the SCCS cohort of low-income blacks and whites, we found that the association between elevated BMI (both in early adulthood and midlife) and ESRD differed by race. Among both black and whites, being overweight or obese in early adulthood (age 21) was associated with a higher risk of ESRD, but the magnitude of the association was considerably stronger in whites compared to blacks. Being obese in midlife (mean age at enrollment was 53.7 years) was not associated with increased risk of ESRD among blacks, while among whites a significantly increased risk of ESRD was observed only among those in the highest category of obesity (BMI ≥ 40).

A few studies have previously reported increased risk of ESRD with higher BMI in early adulthood or adolescence. Using data from over 1.2 million adolescents who underwent fitness evaluation for the Israeli military, Vivante and colleagues reported hazard ratios 3.00 (95 % CI: 2.50–3.60) and 6.89 (95 % CI: 5.52–8.59) for ESRD among those who were overweight or obese, respectively, over a 25-year follow-up period [9]. A retrospective Swedish study of 926 cases and 998 controls observed a 3-fold increased odds (OR 3.1, 95 % CI: 2.1–4.8) of chronic renal insufficiency among persons who were overweight at age 20 [5]. Silverwood et al. also reported an association between younger age when first overweight and other overt renal outcomes using data from the UK Medical Research Council National Survey [26]. In particular, those who first became overweight in early adulthood (26 or 36 years) had increased risk of developing stage 3 chronic kidney disease (CKD) in later life [26]. Similarly, children who were overweight in early life (2–20 years) had increased odds of developing CKD compared to those who were never overweight [27]. These prior studies involved predominantly white populations, and our findings for BMI at age 21 among whites are consistent with those results. However, ours is the first study to investigate this association among blacks, who are at highest risk for ESRD, and to specifically test for effect modification by race, with substantially more pronounced associations among whites compared to blacks.

For BMI in middle-age, a number of studies in various populations, primarily Asian and Caucasian, have reported greater than two-fold increased risks of ESRD in overweight or obese individuals [5, 6]. To our knowledge, the study by Hsu and colleagues using data for 320,252 adults (18 % black, 69 % white) and 1,427 ESRD cases from Kaiser Permanente is the only study that investigated potential heterogeneity of the BMI-ESRD relationship in middle-aged blacks and whites. They reported that among blacks, when compared to persons with normal weight, individuals with obesity class II and III had a risk ratio of 5.5 (95 % CI: 4.1–7.5) and 7.2 (95 % CI: 5.0–10.4), respectively; among whites, corresponding risk ratios were 7.2 (95 % CI: 5.2–10.0) and 8.0 (95 % CI: 5.0–12.8) [8]. A major difference between the two studies is the duration of follow up, being relatively short (median ~ 2.6 years) in our investigation compared to that of the Kaiser study (~26 years). Thus, with longer follow-up, it is possible that our results may be more similar, as suggested by our findings of a strong link between ESRD and BMI at age 21 and by our sensitivity analyses hinting at stronger associations of ESRD with middle age BMI among those with longer follow up. Also, differences in socioeconomic status between blacks and whites are likely to be more pronounced in the Kaiser population than in our study population, and thus residual confounding may contribute to the elevated ESRD risk associated with BMI among blacks.

As ESRD and cardiovascular disease (CVD) share risk factors including obesity, one could expect similar racial patterns in BMI-ESRD and BMI-CVD relationships. Previous studies have reported stronger associations between BMI and cardiovascular outcomes in whites compared to blacks [28, 29]. In the SCCS cohort, it has previously been shown that whites with class III obesity had a greater than 2-fold increased risk of cardiovascular death compared to those of normal weight, while in blacks the increase in risk was modest (17–40 %) and non-significant [29]. Similarly, using SCCS data, Lipworth et al. found that in whites the HR for atrial fibrillation was 1.49 (95 % CI: 1.11, 2.01) in obese persons (compared to normal weight persons), while among blacks the corresponding HR was 0.90 (95 % CI: 0.69, 1.16) [30]. Our findings, which suggest a stronger BMI-ESRD association in whites compared to blacks, are thus in line with CVD results in the same cohort.

The underlying reasons for these differential relationships by race remain poorly understood. It is well-established that blacks and whites have different body compositions, whereby on average whites have higher visceral fat and percent body fat than blacks at equivalent BMI [18, 31]. Waist circumference may be a better proxy than BMI for visceral adiposity [32], which is central to the pathophysiology of cardiorenal endpoints. Waist circumference was measured for only about 10 % of SCCS participants, and in the current analysis was available for only 16 ESRD cases and 54 controls. Among whites, waist circumferences was higher (123.3 vs 101.1 cm) in cases (n = 4) compared to controls (n = 13), while amongst blacks it was similar (103.3 vs 107.7 cm) in cases (n = 12) and controls (n = 41). While caution is mandatory given the small numbers, these data lend some support to the hypothesis that there may be a differential association between obesity and ESRD by race. Waist circumference was previously shown to be more strongly associated with all-cause mortality in whites compared to blacks [20], raising the possibility that higher adiposity may have less detrimental health effects in blacks compared to whites. The existence of such differential effects and its potential link to differences in metabolic pathways related to inflammation and insulin resistance or gene-environment interactions remains to be explored. Previous studies in the SCCS have also revealed differences in mean levels of BMI-related biomarkers (including adiponectin, leptin, insulin-like growth factor 1, and C-reactive protein) between blacks and whites and in the magnitude of associations of these biomarkers with BMI [19–22], providing additional evidence that biological mechanisms associated with obesity may differ across racial groups.

The SCCS provides a unique opportunity to examine health disparities for a sizeable number of black participants, who have traditionally been under-represented in most studies focused on cardio-renal outcomes. In addition, both black and white participants had similar income and education levels thereby minimizing confounding by socioeconomic differences; and a high burden of obesity and its comorbidities (diabetes and hypertension). Linkage of the cohort with the USRDS ensured near complete and unbiased ascertainment of ESRD cases. The prospective nature of the nested case–control study design minimized reporting bias on the part of the subjects related to knowing their disease status. Limitations of the study included our reliance on self-reported height and weight for calculation of BMI at enrollment, but a prior validation study for a subset (n = 14,000) of SCCS participants for whom weight and height were abstracted from contemporaneous clinic records showed very high correlation (r >0.95) between self-reported and measured weight and height [29]. Data from the National Health and Nutrition Examination Survey also suggest that BMI based on self-report has good agreement with BMI from measured values [33]. Also, data for weight at age 21 is prone to recall bias which is likely non-differential as data collection was done at baseline before ascertainment of ESRD. Additionally, there may be some variation in BMI throughout early adulthood which a singular value of BMI at age 21 may not completely capture. The limited number of cases among whites curtailed the power of the analyses in this group so we advise some caution in interpreting these results.

In summary, we observed marked racial differences in the relationship between BMI, particularly BMI in early adulthood, and ESRD in a low-income population with a high burden of obesity. This should prompt more research into understanding the underlying metabolic and potential gene-BMI interactions over the lifespan that may explain these differences and which may add to the body of evidence needed to tailor strategies to combat the deleterious effects of obesity.