Background
Chronic kidney disease (CKD) is a significant health problem which is associated with increased morbidity and mortality, making its prevention a public health priority. Thirteen percent of the adult population of the United States have reduced kidney function or albuminuria [
1]. Early identification of persons at greater risk of developing CKD is critical in prevention and management strategies. Traditionally, hypertension and diabetes are the most commonly known key risk factors for CKD. Others include advanced age, low high-density lipoprotein cholesterol (HDL) and metabolic syndrome [
2]. Previous studies have shown that traditional factors alone are inadequate to explain CKD risks and improve risk stratification for CKD or progression of CKD [
3‐
5]. Established CKD risk factors explain only 34% of renal disease progression among whites and 44% for African Americans after adjusting for sociodemographic, lifestyle and clinical factors [
3,
6]. In clinical settings, CKD prognosis largely depends on traditional markers such as estimated glomerular filtration rate (eGFR) and albuminuria, however these biomarkers only offer modest risk prediction particularly in people with preserved levels of renal function [
7] and are subject to intra-individual variability over time when hydration and medication use are involved. Additionally, albuminuria and eGFR can have a variable relationship, an example being the development of CKD (eGFR < 60 mL/min/1.73 m
2) without albuminuria [
8,
9]. Several other pathways may be involved in CKD development including inflammation and endothelial function [
3,
10]. Studies in cardiovascular disease (CVD) and metabolic syndrome, have benefited from the use of circulating biomarkers in risk prediction [
11‐
13]. Unlike biomarkers in CVD, the list of prognostic biomarkers in CKD is in continuous growth and the concept of a multi-marker approach has been proposed as single biomarkers are unable to fully describe changes in renal function [
14]. While multi-marker approach to predict CKD has been reported in whites [
10], the predictive value of models incorporating multiple biomarkers in CKD prediction among African Americans is not well studied.
In a community-based sample of African Americans enrolled in the Jackson Heart Study (JHS), we sought to identify biomarkers of interest and evaluate their incremental predictive value from a multi-marker panel representing physiological pathways implicated in kidney diseases: adiposity (adiponectin and leptin); adrenal (aldosterone and cortisol); endothelial function (endothelin and homocysteine); inflammation (C - reactive protein, [CRP]); natriuretic (B-type Natriuretic Peptide [BNP]) and renin angiotensin (plasma renin activity, and renin mass). We conducted tests on model improvement using both the C-statistic and the newer measures of net reclassification index (NRI) and integrated discrimination index (IDI) [
15,
16].
Discussion
We investigated the relation of a multi-marker panel with the development of incident CKD and RKFD in a large community-based sample of African Americans. We observed that a panel consisting of nine circulating biomarkers (adiponectin, aldosterone, BNP, hsCRP, endothelin, homocysteine, leptin, PRA, ARM) representing several distinct biologic pathways was associated with development of CKD and RKFD. We identified a smaller subset of biomarkers representing adiposity (adiponectin, leptin); and RAS (aldosterone) pathways that were also associated with these outcomes. Plasma adiponectin and leptin were both associated with development of CKD while plasma aldosterone had a protective effects against both CKD development and RKFD. The addition of biomarkers only marginally improved model discrimination and reclassification compared to the model with traditional risk factors as demonstrated by the small change in C-statistic and reclassification indices. In secondary analyses stratified by obesity status, selected biomarkers were significantly associated with incident CKD and RKFD in non-obese participants only, suggesting modification by obesity status.
In our study, adiponectin and leptin, two of the key cytokines secreted by adipocytes, predicted the development of incident CKD in a multivariable adjusted model. These findings are consistent with previous studies that investigated their association with CKD. In a case-control study among Chinese and Indian adults, patients with CKD had higher levels of leptin and adiponectin compared to controls [
33]. Similar findings were also found in a patient population comprised of 60% African Americans in the greater New Orleans, Louisiana region, after adjusting for race and other risk factors associated with kidney disease [
34]. To the contrary, other studies have also reported no difference in adiponectin levels [
35‐
37]. The link between adipokines and changes in glomerular filtration rate has been reported previously [
38]. Through endothelial dysfunction, oxidative stress and changes in immune response and inflammation, the adipokines are involved in kidney damage [
39]. While serum aldosterone was reported to have weak but significant association with lower eGFR in Framingham Offspring Study, an inverse association where aldosterone appear to be protective was observed in this present study. Aldosterone’s conflicting results are also reported in the Ohasama Study where the authors attributed the lack of association of aldosterone with eGFR to a high salt-intake resulting from high sodium dietary conditions [
40,
41].
Studies on single biomarkers have reported on the relation of CRP and aldosterone with kidney function. Works by Fox and colleagues as well as Shankar and others showed that CRP is associated with prevalent CKD but not with the development of CKD [
42,
43]. While previous studies both clinical and observational have demonstrated CRP’s pathogenic role in renal damage [
44,
45], in the current analysis, CRP was not associated with the development of either CKD or RKFD. Hannemann and colleagues found an inverse association of plasma aldosterone concentration with eGFR in the general population [
46]. In the present study, participants with medium level quartile and higher aldosterone level had 46% (
P = 0.004) and 38% (
P-value = 0.009) less likely to develop CKD or experience RKFD, respectively. When stratified by obesity status, biomarkers were associated with development of CKD and RKFD in non-obese, particularly for leptin and adiponectin. Biomarkers linkages to the development of CKD in the absence of obesity has been reported before even though the mechanism is poorly understood [
38,
47].
Few community-based studies have evaluated kidney disease biomarkers to assess their usefulness in stratifying disease risk [
3]. We undertook this study to address this gap in CKD literature. Data from the Framingham Heart Study (FHS) followed a multi-marker approach to predict incident CKD and microalbuminuria. A panel of seven biomarkers (C-reactive protein, aldosterone, renin, BNP, plasminogen-activator inhibitor type 1, fibrinogen, and homocysteine) was associated with the development of CKD with homocysteine and aldosterone retained as significant markers in the backward elimination model [
10]. Our data extends these findings to a large community-based sample of African Americans in Mississippi. Unlike FHS where homocysteine and aldosterone were retained as significant markers for incident CKD prediction, in JHS adiponectin and leptin were the significant markers.
When comparing indices of model improvement the biomarker model was associated with the same change in C-statistic (ΔC = 0.01) for prediction of incident CKD and RKFD. Researchers generally consider a change in C-statistic of at least 0.05 as indicative of a predictor with clinical significance [
48]. While C statistics has been criticized for being insensitive to small changes in predictive accuracy [
49], it was preferred here to permit easy comparison with findings in the literature, which often used the metric [
50]. We also computed the NRI and IDI indices to complement C-statistic. Consistent with results based on C-statistics for the biomarker model, relative IDI had small but significant incremental predictive ability that was also higher than that reported in FHS. NRI though statistically non-significant was higher than that reported in FHS (JHS NRI = 16.1%,
P = 0.08; FHS NRI = 6.9%,
P = 0.0004). Though the metrics of model improvement are study/cohort specific, suffice it to say that they hold promise for CKD prediction in African Americans as is in white populations. The utility of biomarkers in improving disease prediction is highly successful in the area of cardiovascular medicine [
50‐
52]; however, the yield has been relatively small in women and the elderly [
53,
54]. With exception of Velagaleti et al. report on prediction of heart failure (ΔC =0.02) [
55], most CVD research has reported lower incremental benefit compared to current analyses. This may be because CVD risk factors are well characterized and the existence of multiple risk-algorithms aid prediction, something which is lacking in CKD research.
Strengths and limitations
This study has some strengths and limitations. The analyses had a large sample size and a well-documented spectrum of biomarkers. We also adjusted for many CKD factors so the independent association between multi-marker panel and CKD development could be assessed. Some limitations require mentioning. Our sample was primarily African Americans, limiting generalizability to other ethnicities. JHS was designed to investigate CVD risk factors, thus the biomarkers collected were not specifically for CKD, although CVD is a potent risk factor for CKD and CKD progression. Finally, with the study being observational in nature, it is possible that the CKD being detected might have developed at an earlier date.
Acknowledgements
The authors thank the participants and data collection staff of the Jackson Heart Study.