Relationship Between Levels of Advanced Glycation End Products and Their Soluble Receptor and Adverse Outcomes in Adults With Type 2 Diabetes

Diabetes is associated with an increased risk of adverse renal and cardiovascular outcomes, which are only partially reduced by intensive glycemic control. This may be because of the long-lasting (legacy) effects of poor glycemic control over many years (1). Potential mediators of this “metabolic karma” include the long-lasting posttranslational molecular modifications induced by hyperglycemia, known as advanced glycation end products (AGEs) (2). Experimental data have linked AGEs to the development and progression of diabetes complications, including nephropathy, retinopathy, neuropathy, and cardiovascular disease (3–5). Posttranslational modification of functional groups on vulnerable proteins, lipids, and DNA targets has the potential to alter their structure, stability, and/or function (6,7). AGEs also stimulate pathogenic pathways after activation of the receptor for AGEs (RAGE) (8), which may represent the major mechanism by which AGEs lead to diabetes complications.

One potential marker for the expression of RAGE and activation of the AGE/RAGE axis is circulating soluble RAGE (sRAGE), a C-truncated isoform largely produced by proteolytic cleavage of the membrane-bound form via the action of metalloprotein sheddases (9). Tissue RAGE and circulating sRAGE are both increased in individuals with diabetes, paralleling increased levels of circulating AGEs and other RAGE ligands, including S100a8/9 and high-mobility group protein box-1, whose actions lead not only to RAGE activation but also to the autoinduction of RAGE expression (10–12). We and others have previously shown that sRAGE is independently associated with the development of cardiovascular complications and mortality in adults with type 1 diabetes (13,14). Similar findings have been reported in patients in type 2 diabetes without prior cardiovascular disease from the Collaborative Atorvastatin Diabetes Study (CARDS), although microvascular outcomes were not described in that report (15). In this current case-cohort study, we explore the association between levels of circulating AGEs, sRAGE, and adverse outcomes (death, cardiovascular complications, and new or worsening nephropathy) in adults with type 2 diabetes who participated in the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial.

The ADVANCE study has been described in detail previously (16–18). Participants with type 2 diabetes at increased risk of cardiovascular events were recruited from 20 countries in Asia, Australasia, Europe, and North America. Participants were also required to be aged ≥55 years and have a history of cardiovascular disease or one or more additional cardiovascular risk factor. The study made two randomized comparisons: a double-blind assessment of the efficacy of a fixed combination of perindopril and indapamide (2 mg/0.625 mg for 3 months, increasing to 4 mg/1.25 mg if tolerated) versus placebo, and an open-label evaluation of an intensive glucose-lowering regimen using modified release gliclazide, with a target glycated hemoglobin (HbA_1c) of ≤6.5%, versus standard guideline-based glycemic control. The study randomized 11,140 participants, and the median duration of follow-up was 5 years.

Nonfasting blood samples were taken at baseline, anticoagulated with EDTA, and stored centrally at −80°C for a median of 7.8 years before analyses. Stored plasma samples were available from all countries involved in ADVANCE, except China and India, giving a total base population of 7,376 patients. A case-cohort study population was constructed from a random sample of 3,500 individuals taken from these 7,376 participants plus 697 additional individuals who had a cardiovascular event, a microvascular complication, or died during follow-up.

Baseline data included demographic and clinical information, including age at diagnosis, presence and severity of diabetes complications, antidiabetic therapy, and other regular medications. Weight, height, urinary albumin-to-creatinine ratio (ACR), serum creatinine, fasting lipid levels, and HbA_1c were also measured. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation (19). Levels of baseline high-sensitivity cardiac troponin T (hs-cTnT), N-terminal pro-B type natriuretic peptide (NT-proBNP), and hs-CRP were also measured in stored samples (20–22).

sRAGE was measured by ELISA (Quantikine; R&D Inc., Minneapolis, MN). Intra-assay coefficients of variability were <7%, and between-assay coefficients of variability were <9%, as previously described (10). Circulating AGEs were estimated by the level of AGE-associated autofluorescence (excitation, 370 nm; emission, 440 nm) in EDTA-treated plasma samples, adjusted for protein quantity.

The study end points were all-cause death, major cardiovascular events (cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke), and new or worsening nephropathy (defined as the development of macroalbuminuria, doubling of serum creatinine to a level of at least 200 µmol/L, need for renal replacement therapy, or death due to renal disease). All outcomes were validated by an independent adjudication committee.

Baseline characteristics were summarized by ordinal categories defined by the tertiles of sRAGE and AGE levels. Tests for linear trend were performed across the thirds of each biomarker to identify possible covariates associated with sRAGE and AGE using linear regression. Hazard ratios (HRs) for log-linear effects of sRAGE and AGE levels on each of the three studied outcomes were obtained from weighted Cox regression models for case-cohort analyses in the SAS package. sRAGE levels were log-transformed to remove the effects of skewness. Results were expressed per unit SD, rounded up to whole numbers (for log sRAGE, 0.927 log[pg/mL], rounded to 1.0; and for AGEs, 93.23 relative fluorescent units [RFU]/mg protein, rounded to 100).

Six models with different sets of potential confounding variables were fitted for each of the biomarker/outcome combinations:

• model 1, with age, sex, and randomized treatment;

• model 2, with, additionally, duration of diabetes, current smoking, systolic blood pressure, BMI, HbA_1c, plasma glucose, total and HDL cholesterol, triglycerides, and a history of macrovascular complications;

• model 3, with, additionally ACR and eGFR;

• model 4, which added the other biomarker (i.e., sRAGE or AGE levels) to model 2;

• model 5, which added hs-cTnT, NT-proBNP, hs-CRP, and use of β-blockers to model 3; and

• model 6, which added the other biomarker (i.e., sRAGE or AGE levels) to model 5.

Analyses were also performed stratified by age (>67 or ≤67 years), sex, a history of macrovascular complications, a history of microvascular complications, eGFR (<60 or ≥60 mL/min/1.73 m²), and HbA_1c (<7% or ≥7%) adjusted for the variables described in model 2, except a prior history of macrovascular complications. Log-likelihood ratio tests were conducted to test the nonlinearity effect of sRAGE and AGE level to outcomes by comparing a model with categorical biomarkers and a model with continuous biomarkers.

Because our data tended to have relatively high levels of sRAGE compared with earlier studies in diabetes (13–15), a sensitivity analysis was run repeating the main analyses of association in a restricted population including only those below the second tertile of sRAGE or AGE, as appropriate.

Discrimination was evaluated through C-statistics for 5-year risk, accounting for censoring, and compared between the full clinical model (with the covariates in model 3) and when adding each biomarker individually. In addition, the ability to reclassify the 5-year risk for new and worsening nephropathy was assessed by the integrated discrimination index (IDI) and net reclassification improvement (NRI) methods, using techniques suitable for survival data applied to the random subcohort (23,24).

Of the 4,197 participants in this study, blood samples from 434 (10%) were missing or unsuitable for analysis. The remaining 3,763 subjects in this cohort had a mean age of 66.9 years and a median duration of diabetes of 7.9 years (Tables 1 and 2); 92% were Caucasian. Of these participants, 942 (25%) had eGFR <60 mL/min/1.73 m² and 1,098 (29%) had micro- or macroalbuminuria; overall, 1,700 (45%) had eGFR <60 mL/min/1.73 m² or albuminuria. Circulating levels of sRAGE and AGEs were highest in patients with poorer renal function, higher urinary ACRs, dyslipidemia, or advanced age. Circulating concentrations of sRAGE and AGEs were associated with levels of hs-cTnT and NT-proBNP and with a history of heart failure. AGE levels, but not sRAGE levels, were higher in patients with poor glucose control or a prolonged duration of diabetes before enrolment.

During a median of 5 years of follow-up, 689 participants (18%) died, 683 (18%) suffered a major cardiovascular event, and 272 (7%) developed new or worsening nephropathy. Individuals with higher levels of sRAGE were more likely to experience each of these adverse outcomes, after adjusting for age, sex, and randomized treatment (Table 3). When the random subcohort was divided into thirds, participants with higher concentrations of sRAGE were at increased risk of all three outcomes (Supplementary Fig. 1). After adjusting for several other clinical risk factors, these associations were attenuated, largely through the adjustment for renal impairment, although the association between sRAGE and all-cause mortality and sRAGE with adverse renal events remained significant. By contrast, the association with cardiovascular events was no longer significant. In analyses restricted to the lowest two-thirds of sRAGE, associations with new or worsening nephropathy were largely preserved (Supplementary Table 1). Circulating levels of AGEs were associated only with an increased risk of new or worsening nephropathy. This relationship persisted even after adjustment for a wide range of potential confounding factors, including sRAGE levels (Table 3 and Supplementary Fig. 2). AGEs and sRAGE both improved the accuracy with which the 5-year risk of new or worsening nephropathy could be discriminated or correctly classified compared with only accounting for the other clinical and demographic measures considered (Table 4).

There was no evidence of nonlinearity (P > 0.05) between sRAGE or AGE levels and the three study outcomes. No significant interactions were found in subgroup analyses for either biomarker when considering the outcomes of death or cardiovascular events (Supplementary Fig. 3A and B). A similar stratified analysis showed the relationship between sRAGE levels and new or worsening nephropathy was most evident in patients with preserved renal function (Supplementary Fig. 3C).

Activation of RAGE by AGEs and other ligands has been suggested to be an important mediator of vascular complications in diabetes (4,25,26). The expression of RAGE is increased at sites of vascular injury, including in the diabetic kidney (4). Circulating levels of sRAGE are increased in individuals with diabetes, especially those with microvascular and/or macrovascular complications of their disease (13–15). In this report we also demonstrate that sRAGE levels are also independently associated with new-onset or worsening renal disease and all-cause mortality in individuals with type 2 diabetes from the ADVANCE trial. The current results are consistent with our previous data in patients with type 1 diabetes (13) and in patients with type 2 diabetes enrolled in CARDS (15), where sRAGE levels were positively associated with all-cause mortality. However, this is the first report confirming an independent additional association between sRAGE and progressive renal disease specifically in patients with type 2 diabetes. Taken together, these results highlight the potential importance of the AGE/RAGE axis in the pathogenesis of diabetes complications as well as their outcomes.

It is important to note that studies of general population cohorts have reported that lower sRAGE levels may be associated with poor health outcomes (27,28). For example, in the Atherosclerosis Risk in Communities (ARIC) study in adults from the general population with normal renal function, lower levels of sRAGE were paradoxically associated with an increased risk of diabetes, coronary heart disease, and mortality during 18 years of follow-up (26). It is possible to speculate that in populations (like ADVANCE) with a high prevalence of renal impairment, reduced renal clearance of sRAGE serves to dramatically uncover the induction of RAGE expression that might not otherwise be observed in healthy adults.

In a case-cohort study nested within the ARIC study, sRAGE levels were positively associated with development of chronic kidney disease and end-stage renal disease (29). However, these associations were not significant after adjustment for eGFR, which, as in the ADVANCE study, was the major determinant of circulating sRAGE levels. By contrast, the association between sRAGE and all-cause mortality and sRAGE with adverse renal events in the ADVANCE study remained significant after adjustment for renal impairment, possibly because of a higher event rate and,consequently, greater power to observe an effect. Unlike ARIC, ADVANCE had few (<2%) black participants, in whom sRAGE levels are known to be lower (15). There may be an effect of treatment, such that the aggressive management of risks in patients with type 2 diabetes included in the ADVANCE study effectively eliminated the dominant role of conventional risk factors in the pathogenesis of adverse outcomes, while glucose and lipids remain strong determinants of outcome in untreated cohorts. In addition, it may be that increased levels of RAGE ligands and elevated levels of RAGE (as occurs in diabetes) are both required for the pathogenic aspects of this pathway to become manifest. This would be consistent with our finding that AGEs and sRAGE were both independently associated with renal outcomes in our cohort.

Although a causal relationship cannot be established from the presented observational data, the finding that activation of the AGE/RAGE axis may be associated with progressive renal disease is consistent with prior experimental data. Genetic depletion of RAGE is able to attenuate complications associated with diabetes (30), and treatment with truncated RAGE also significantly reduces the renal damage in diabetic mice (25). Similarly, the potential importance of AGEs as downstream mediators of renal damage has been demonstrated in studies using chemically disparate inhibitors of AGE formation to retard the development of diabetic kidney disease (31). Previous studies have reported a strong association between the presence and severity of kidney damage and circulating levels of sRAGE in patients with type 2 diabetes (11,32–34). This is partly determined by the reduced clearance of sRAGE in patients with renal impairment. However, we show here for the first time that, independent of baseline renal function and other risk factors for progressive renal disease, sRAGE was significantly associated with new or worsening nephropathy in adults with type 2 diabetes. Moreover, the improvements in the NRI achieved when adding sRAGE and/or AGE were substantial compared with other recognized biomarkers, suggesting these indices may be of value when stratifying the risk of incident or worsening nephropathy in patients with diabetes as well as being a potential target for novel treatments.

Some small cross-sectional studies have previously described an association between circulating levels of sRAGE and atherosclerosis burden in type 2 diabetes (28,35,36). However, we found no independent association between levels of sRAGE and cardiovascular outcomes in our cohort after adjusting for other risk factors, including baseline renal function. This is perhaps surprising given the clear vasculoprotective effects of RAGE deletion observed in animal studies of atherosclerosis (30) and in our previous report of an association between sRAGE and cardiovascular outcomes in adults with type 1 diabetes (13). However, the greatest benefit of glucose lowering in the ADVANCE study was observed on renal outcomes, while the same cardiovascular outcomes not linked to sRAGE cardiovascular outcomes were not significantly modified by intensive glucose control (16).

In the ADVANCE study, AGE levels were higher in patients with poor glycemic control or prolonged duration of diabetes before intervention, consistent with the nonenzymatic chemistry of the Maillard reaction, which is dependent on both glucose concentration and time. However, sRAGE levels were not associated with HbA_1c or diabetes duration. This may be reconciled because RAGE is a multiligand receptor and may be induced by a range of other proinflammatory ligands (e.g., S100) and by inflammatory processes via the nuclear factor-κB binding element in its promoter. This may partly explain the association with renal injury and mortality observed in this study.

The strengths of the current study include its large size, international recruitment in a well-characterized trial population that was closely monitored and treated uniformly, rigorous a priori definition of outcomes, completeness of follow-up, and adjustment for major risk factors. Some limitations include the potential for recruitment bias in a randomized controlled trial and using only single measures of the baseline variables of interest. Our indirect measure of AGEs was also crude compared with the more quantitative assays based on mass spectroscopy and ELISA. However, there is no evidence that more specific measures offer any advantages, and even skin autofluorescence has been shown to be a good predictor of AGE burden and of the risk of adverse outcomes (37). That associations demonstrated in this study may be due to confounding by unmeasured factors or ones that are difficult to quantify is also possible. For example, sRAGE concentrations may be associated with a range of occult differences in inflammatory or metabolic pathways that may themselves affect clinical outcomes in individuals with diabetes.

In summary, this large multicenter study in adults with type 2 diabetes demonstrates that sRAGE is positively associated with new-onset or worsening renal disease and all-cause mortality, independent of and in addition to chronic kidney disease, the major predictor of adverse outcomes in this cohort. These data support the accumulating body of experimental data demonstrating that activation of the AGE/RAGE axis is a key mediator of microvascular damage in diabetes but also suggest that it may not be a direct mediator of macrovascular complications. Genetic studies would be useful to explore whether any effects are truly causal.

Funding. This work was supported by a National Health and Medical Research Council (NHMRC) of Australia project grant (632507). The Diabetes Australia Research Trust helped to fund the hs-cTnT, NT-proBNP, and hs-CRP assays. M.W. has a Principal Research Fellowship and B.N. a Senior Research Fellowship from the NHMRC. M.W., B.N., V.P., and J.C. hold an NHMRC Program Grant.

Duality of Interest. M.W. has received grant funding from Sanofi and is a consultant for Amgen. B.N. declares fees from Janssen; lecture fees (honoraria) and travel reimbursements for speaking for Abbott, Merck Sharp & Dohme, Novartis, Pfizer, Roche, and Servier; research support from AbbVie Inc., Dr. Reddy’s Laboratories, Janssen Pharmaceuticals, Inc. (Johnson & Johnson), Merck Schering-Plough, Roche, and Servier; all fees from commercial entities are paid to The George Institute for Global Health. M.M. declares boards and lectures for Abbott and Servier; boards for Intarcia Therapeutics, Inc.; lectures for Eli Lilly; boards, lectures for, and scientific support from Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, and Sanofi; and scientific support from Novartis. B.W. has received honoraria for lectures from Servier, Novartis, Boehringer Ingelheim, and Daiichi Sankyo. V.P. has received speaker’s fees from Servier. J.C. has received research grants from Servier, administered through The University of Sydney, as Chief Investigator for the ADVANCE trial and ADVANCE-ON posttrial follow-up study, as well as honoraria and travel support from Servier for speaking about these studies at scientific meetings. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.C.T. and G.S.H. drafted the paper, which was redrafted by M.W. M.C.T. and R.P. performed the laboratory tests on sRAGE and AGEs. M.W. and Q.L. were responsible for the statistics. B.N., M.M., B.W., V.P., M.E.C., S.Z., and J.C. commented on the original draft and assisted with modifications. M.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.