Introduction

Identifying biomarkers predictive of estimated GFR (eGFR) decline could point to pathways that may help identify new therapeutic agents. With the recognition that type 2 diabetes itself may be a chronic inflammatory state, the role of potentially modifiable lipid and inflammatory biomarkers has been of particular interest.

Notably, higher baseline C-reactive protein (CRP) was associated with eGFR decline in people with hyperlipidaemia and a history of myocardial infarction [1] and in older community-dwelling adults [2]. No relation between CRP and eGFR decline was identified, however, in an analysis of the Modification of Diet in Renal Disease study [3], but mean duration of follow-up was only 2.2 years. Soluble tumour necrosis factor receptor 2 (sTNFR-2) has also been associated with eGFR decline [1]. These investigations included less than 15% diabetic patients, however.

We previously reported that dyslipidaemia, as well as higher fibrinogen, vascular cell adhesion molecule 1 and sTNFR-2, were independently associated with lower eGFR in a cross-sectional study of men with type 2 diabetes [4]. In this study we expanded our investigations by examining the relation between lipid and inflammatory biomarkers with respect to the 11-year eGFR decline in women with type 2 diabetes.

Methods

This study was approved by the Partners’ Healthcare Brigham and Women’s Hospital Human Research Committee Institutional Review Board.

Participants

The Nurses’ Health Study (NHS) was initiated in 1976 with the enrolment of 121,700 US nurses aged 30–55 years. Questionnaires are mailed biennially; return of questionnaires constitutes informed consent under the study’s institutional review board protocol. Between 1989 and 1990, 32,826 participants provided blood samples that were shipped on ice by overnight delivery and stored at −130°C as previously described [5]. In the year 2000, many of these participants submitted a second blood sample under the same shipment and storage conditions.

We mailed a diabetes supplementary questionnaire (DSQ) to all women reporting diabetes on biennial questionnaires. We used the National Diabetes Data Group criteria [6] to define diabetes self-reported up to the 1996 biennial questionnaire. Self-reported diagnosis using the DSQ was 98% accurate in a validation study with medical records review [7].

We included women who reported being diagnosed with type 2 diabetes up to the year 2000 (because glucose intolerance can precede the diagnosis of type 2 diabetes by ≥10 years [8]) and who had plasma samples from both 1989 and 2000. We excluded women with a serum creatinine concentration ≤44 mmol/l (felt to be implausible) (n = 12) and those for whom data on plasma creatinine were missing (n = 19). After applying these criteria, 516 women had eGFR change data available for analysis (Fig. 1).

Fig. 1
figure 1

Biomarkers and type 2 diabetes participant selection in the Nurses’ Health Study. DM, diabetes mellitus

Full size image

Biomarker analyses

Biomarker assays were performed at Children’s Hospital, Boston, MA, USA, or in the laboratory of Dr C. Mantzoros at Beth Israel Deaconess Hospital, Boston, MA, USA, using standard methods and assays (see Electronic supplementary material). On blinded quality control samples that were run simultaneously with participant samples, biomarker measurements had CVs of less than 15% with the exception of E-selectin (20%), adiponectin (27.4%) and sTNFR-2 (16.7%).

Measurements of kidney function decline

Plasma creatinine was analysed using a modified kinetic Jaffe reaction (CV 10%). In 2007, repeat creatinine assays of 20 NHS plasma samples (with a range of 53 to 124 mmol/l) initially measured in the year 2000 revealed a mean recalibration coefficient (new value/original value) of 0.97 and confirmed that creatinine was stable under our storage conditions.

Glomerular filtration was estimated by the abbreviated MDRD (Modification of Diet in Renal Disease) equation [9]. An eGFR decline of ≥25% between 1989 and 2000 was the primary outcome. An eGFR decline of ≥25% over 11 years was determined a priori and had been used in previous analyses of renal function decline in NHS participants [10].

Statistical analysis

Spearman correlations were calculated for eGFR and individual biomarkers at baseline. Multivariable logistic regression modelling was applied to analyse associations between biomarkers and an eGFR decline of ≥25%. Animal protein intake, alcohol intake and baseline cardiovascular disease were also considered for the adjusted models but were removed as they did not change results. Analyses were performed with SAS software, version 9.1 (SAS Institute, Cary, NC, USA).

Results

Characteristics of the study cohort are summarised in Table 1. In 1989, significant modest cross-sectional inverse correlations were noted between eGFR and leptin (r = 0.16, p < 0.001), adiponectin (r = −0.13, p = 0.003) and sTNFR-2 (r = −0.28, p < 0.001). CRP was not associated with eGFR (r = 0.03, p = 0.57). The women who experienced an eGFR decline of ≥25% between 1989 and 2000 had higher rates of hypertension, cardiovascular disease, angiotensin-converting enzyme inhibitor (ACE-I)/angiotensin-type 2 receptor blocker medication use, and lower physical activity. In 1989, only 35 women (6.8%) had eGFR <60 ml min−1 1.73 m−2. In 2000, 139 women (27%) had eGFR <60 ml min−1 1.73 m−2.

Table 1 Demographic and clinical characteristics of participants of the Nurses’ Health Study with type 2 diabetes in year 2000
Full size table

Those in the highest quartile of sTNFR-2 in 1989 had significantly higher odds for an eGFR decline of ≥25% between 1989 and 2000 (multivariate OR 5.81; 95% CI 2.90–11.65) (Table 2 and Fig. 2). The association between sTNFR-2 and eGFR decline was stronger in women with a BMI of ≥30 (OR 16.76; 95% CI 4.69–59.9) than in women with a BMI of <30 (OR 2.78; 95% CI 1.12–6.89) (p for interaction = 0.02). No other significant associations between biomarkers and eGFR decline were present after multivariable adjustment. Restricting the analyses to the 369 women diagnosed with diabetes in 1990 or before did not change the results meaningfully.

Table 2 Logistic regression models for odds ratios for an eGFR decline of ≥25% between 1989 and 2000 for quartile 4 vs quartile 1 of lipid and inflammatory biomarkers
Full size table
Fig. 2
figure 2

Odds ratios with 95% CIs for univariate age-adjusted (UA) and multivariable-adjusted (MA) associations of quartiles of sTNFR-2 with an eGFR decline of ≥25% between 1989 and 2000. Quartile 1 is the referent, 939–2,058 pg/ml; quartile 2, 2,062–2,446 pg/ml; quartile 3, 2,450–2,975 pg/ml; quartile 4, 2,979–5,758 pg/ml

Full size image

Discussion

After multivariable adjustment, sTNFR-2 was a strong predictor of eGFR decline and varied with obesity status. These data support the theory that sTNFR-2, our surrogate marker for the volatile cytokine TNF-α, promotes tubular epithelial apoptosis [11] and progressive renal tubulointerstitial fibrosis [12]. Tubulointerstitial fibrosis is one of the strongest predictors of subsequent kidney disease progression [13], which ultimately manifests as GFR decline. Our findings are consistent with the Cholesterol and Recurrent Events (CARE) trial, in which significantly faster eGFR decline was associated with higher levels of sTNFR-2 [1]. Unlike the CARE study, however, we did not see an association between CRP and eGFR decline. This may be attributable to the difference in populations (type 2 diabetic vs mostly non-diabetic individuals with a history of myocardial infarction).

Obesity is an independent predictor of chronic kidney disease development [14]; we therefore stratified by obesity status and found that obesity modified the associations between sTNFR-2 and eGFR decline. TNF-α is elevated in obese compared with non-obese women and weight loss decreases these cytokine levels [15]. Because adipocytes are now recognised as important regulators of metabolism and inflammation through their secretion of cytokines, chemokines and hormone-like proteins, we hypothesise that the larger mass of adipose tissue in obese humans may intensify the detrimental effects of elevated sTNFR-2 on kidney function.

We confirmed some of our previous findings from a cross-sectional analysis of 732 men with type 2 diabetes in the Health Professionals Follow-Up Study [4]. The strongest correlation with eGFR was seen with sTNFR-2 (r = −0.39, p < 0.001 for men, r = −0.28, p < 0.001 for women), whereas CRP and eGFR were weakly or not correlated (r = −0.09, p = 0.02 for men and r = 0.03, p = 0.57 for women). No significant relations were noted between dyslipidaemia and eGFR in NHS diabetic women, whereas in the diabetic men eGFR was significantly and inversely associated with total cholesterol, triacylglycerols, non-HDL, lipoprotein(a) and apolipoprotein B. Therefore, it appears that associations between lipids and eGFR may vary by sex.

In contrast to our findings, an analysis of over 4,000 participants with type 2 diabetes in the UK Prospective Diabetes Study reported that higher baseline triacylglycerols and LDL-cholesterol were independently associated with incident renal impairment (creatinine clearance <60 ml/min or doubling of plasma creatinine) [16]. The larger number of male and female participants in the UK Prospective Diabetes Study analysis allowed the examination of incident renal impairment, and the modelling approach used lipids as continuous variables rather than in quantile categories; thus, the results may not be directly comparable to the current study.

Limitations of this study include the relatively small number of women with an eGFR decline of ≥25% (thus limiting power for this study) and the high CVs for some of the biomarkers, which resulted in wide confidence intervals; however, the high CVs would be likely to bias the results towards the null, whereas an association with eGFR decline was observed. The participants were mostly older white women so results may not necessarily be generalisable to other populations. Although it is possible that some women with type 2 diabetes who were lost to follow-up would not have been included in these analyses, this would probably bias the results towards the null. Despite these limitations, this study provides important new additional information on biomarkers and kidney function in diabetes, which has not been studied extensively in humans.

We conclude that a higher level of sTNFR-2 is associated with subsequent eGFR decline in women with type 2 diabetes; the relation of sTNFR-2 with eGFR decline appeared stronger in those with a BMI of ≥30. Progressive kidney disease in type 2 diabetes currently represents a major public health problem worldwide and further investigation into its pathogenesis, especially inflammatory pathways, may open new avenues for potential therapeutics. Therefore, future research may include studying whether TNF-α blockade is beneficial in slowing eGFR decline in type 2 diabetes.