Measurement of Serum N-Glycans in the Assessment of Early Vascular Aging (Arterial Stiffness) in Adults With Type 1 Diabetes

There is evidence that increased arterial stiffness (AS) in the elastic large arteries (e.g., aorta) should be considered as a sign of early vascular aging (1,2). The gold standard for measuring AS is the carotid-femoral pulse wave velocity, also known as aortic pulse wave velocity (aPWV). The higher the aPWV, the greater the AS (3). The availability of reference values for aPWV allows the estimation of the biological age of a given individual, which is greater than chronological age in a background of early vascular aging (3). Two large meta-analysis studies have demonstrated that aPWV is an independent predictor for cardiovascular disease (CVD) events and all-cause mortality (4) and significantly improves the predictions of the Framingham Risk Score (5). More recently, AS has been recognized as having a potential role in the development of microvascular complications (e.g., nephropathy, cognitive impairment) (3).

Patients with T1D experience a significant decrease in their life expectancy, and CVD is usually the main cause of death (6). They also have a higher aPWV than age- and sex-matched healthy individuals, which is associated with a greater risk for retinopathy, nephropathy, and CVD events (7). Notably, a recent study reported that an increase in aPWV over ∼6 years in adults with T1D was associated with progression of albuminuria, a decline in estimated glomerular filtration rate (eGFR), development of CVD events, and all-cause mortality (8). Overall, these findings support the concept that the arterial wall per se should be considered as a target organ in adults with T1D (9).

Inflammation plays a critical role in the development of both aging and AS (3,10). Indeed, the term inflammaging has been coined to describe the low-grade inflammatory response that occurs with aging (11). Along this line, we reported an association between AS and a low-grade inflammation score based on a combination of several “classical” circulating inflammation-linked biomarkers (e.g., hs-CRP) in adults with T1D (12).

Most acute phase proteins released from the liver during an inflammatory response are enzymatically glycosylated and contain N-glycan residues that can be quantified by ¹H nuclear magnetic resonance (NMR) spectroscopy (13). Accumulating evidence has indicated that N-acetylglucosamine/N-acetylgalactosamine (GlycA) and N-acetylneuraminic acid (GlycB) are potentially useful biomarkers of low-grade inflammation in several conditions (13). Specifically, high circulating levels of GlycA have been associated with the development of cardiovascular events, cardiovascular mortality, all-cause mortality, and even cancer mortality (13). Accordingly, higher circulating levels of GlycA have been associated with lower life expectancy (14). In this context, the current study explores the potential association between vascular aging and serum GlycA and GlycB in adults with T1D and the utility of such an association in the assessment of vascular aging in clinical practice.

Eighty-four patients (aged 35–65 years) with T1D of at least a 10-year duration, and 68 healthy control subjects were included in the current study. None of the subjects had established CVD (coronary artery disease, stroke, or peripheral artery disease). Exclusion criteria were the following: 1) chronic kidney disease (eGFR by Chronic Kidney Disease Epidemiology Collaboration equation [15] <60 mL/min/1.73 m²), 2) any other acute/chronic condition associated with an inflammatory response (e.g., acute or chronic inflammatory or infectious diseases), 3) use of anti-inflammatory drugs in the previous 6 months, 4) malignant disease in the previous 5 years, 5) hospitalization in the previous 2 months, 6) arrhythmia (except atrial premature complex), and 7) pregnancy. Subjects with T1D were consecutively recruited from our outpatient clinic. The healthy control group was recruited from hospital staff members and their relatives and friends. All nonmenopausal women were evaluated in the follicular phase of their menstrual cycle.

The study protocol was approved by the hospital ethics committee (Parc Taulí Research Ethics Committee, reference no. 2013563; date of approval 18 June 2013) and was conducted in accordance with the Declaration of Helsinki. All subjects gave written informed consent before participating in the study.

All participants underwent a standardized clinical history and physical examination. The following information was recorded using a predefined standardized form: age, sex, diabetes duration, family history of premature CVD (defined as clinical CVD occurring before the age of 55 years in male and 65 years in female first-degree relatives), physical activity (International Physical Activity Questionnaire) (16), active smoking, alcohol intake, insulin dose, and use of any other medication. Body weight, height, and waist and hip circumferences were also registered. Systolic and diastolic blood pressure were measured, and mean arterial pressure (MAP) was calculated as 1/3 × systolic blood pressure + 2/3 × diastolic blood pressure. Venous blood samples were drawn after an overnight fast, and aliquots of plasma and serum were stored at −80°C until processing. Complete blood counts, fasting plasma glucose, glycated hemoglobin (HbA_1c), creatinine, and conventional lipid profile were determined. Hypertension was defined as systolic/diastolic blood pressure >140/90 mmHg and/or treatment with antihypertensive drugs. Dyslipidemia was defined as measured total cholesterol >200 mg/dL, triglycerides >150 mg/dL, HDL cholesterol <40 mg/dL, LDL cholesterol >130 mg/dL, and/or drug treatment for dyslipidemia.

Plasma glucose, serum creatinine, total serum cholesterol, triglycerides, and HDL cholesterol were measured using standard enzymatic methods. LDL cholesterol was calculated using the Friedewald equation (17). HbA_1c was determined by high-performance liquid chromatography (Menarini Diagnostics, Firenze, Italy).

To estimate insulin resistance, we used the equation proposed by Williams et al. (18) for subjects with T1D, which was subsequently adapted for the use of HbA_1c instead of HbA₁ by Kilpatrick et al. (19) in the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) cohort. It yields an estimated R_d based on glycemic control, waist-to-hip ratio (WHR), and blood pressure (estimated R_d = 24.31 − 12.22 × WHR − 3.29 × hypertension [0 = no, 1 = yes] − 0.57 × HbA_1c).

Peripheral polyneuropathy was assessed through a previously described two-step protocol combining the 15-item Michigan Neuropathy Screening Instrument questionnaire and a physical examination (20). Evaluation of the presence and degree of retinopathy was performed by the same ophthalmologist. Subjects were classified into the following three groups according to the degree of retinopathy: no retinopathy, nonproliferative retinopathy, or proliferative retinopathy. Nephropathy was assessed by the measurement of the urinary albumin-to-creatinine ratio (ACR) in three morning urinary samples. Subjects with two of three urinary ACR values >30 mg/g (21) or previously treated with ACE inhibitors or angiotensin-receptor blockers (for microalbuminuria or macroalbuminuria) were considered to have diabetic nephropathy.

Skin autofluorescence was assessed using an AGE Reader (DiagnOptics BV, Groningen, the Netherlands) as previously described (22). Briefly, with the subjects seated, three measurements per subject were taken at room temperature on the forearm in three different positions at ∼5–10 cm below the elbow fold. The AGE Reader illuminates ∼1 cm² of the skin surface (guarded against surrounding light) with an excitation light source of 300–420 nm. This technique provides results in arbitrary units and has been previously validated for T1D (23). The coefficient of variation of the three repeated measurements was 4.2.

Glycoprotein analyses of serum were performed with the Glycoscale test developed by our group (Biosfer Teslab), a ¹H-NMR spectroscopy–based method (24). Briefly, ¹H-NMR spectra were recorded at 310 K on a Bruker Avance III 600 spectrometer operating at a proton frequency of 600.20 MHz (14.1 T). The glycoprotein profiling method was performed in the region between 2.15 and 1.90 ppm of chemical shift. Using ¹H-NMR, the line shape method was developed to characterize the two peaks associated with glycoproteins (GlycA and GlycB) and the following derived variables: areas of GlycA and GlycB and the shape factors of these two peaks (height/width [H/W] ratio of the peaks). The area of these functions gives information about the concentration of acetyl groups of N-acetylglucosamine and N-acetylgalactosamine (for GlycA) and N-acetylneuraminic acid (for GlycB). The H/W ratio provides information on its flexibility and the aggregation of the molecules generating the signal. Higher ratios are considered a marker of inflammation. The ¹H-NMR spectroscopy technique showed an intra- and interassay coefficient of variation <5% for the different glycoproteins assayed. The reproducibility was proven for samples frozen at −80°C for at least 5 years.

The aPWV was measured according to international consensus recommendations (25). The method has been previously described in detail (12). In brief, aPWV was determined by sequential applanation tonometry using a Millar tonometer (SPC-301; Millar Instruments, Houston, TX) at the carotid and femoral arteries, gated to three-lead electrocardiography using the SphygmoCor system (ATCOR, West Ryde, New South Wales, Australia). Those aPWV recordings not satisfying the automatic quality controls specified by the SphygmoCor software were rejected. The mean of two aPWV measurements was taken for each subject for all calculations. Data were available for all the subjects included in the study.

All data were tested for normality using the Shapiro-Wilk test. Data are presented as a percentage, mean (SD) for normally distributed quantitative variables, or median (interquartile range [IQR]) for nonnormally distributed quantitative variables. Nonnormally distributed quantitative variables were log₁₀ transformed. Differences between groups were analyzed using the χ² test for comparisons of proportions and the unpaired t test or the Mann-Whitney U test for comparisons of normally and nonnormally distributed quantitative variables, as needed. Spearman rank correlation coefficients were used for the analysis of the relationships among the advance glycation end products (AGEs), ¹H-NMR–assessed glycoprotein profile, classical cardiovascular risk factors, and aPWV. To assess the potential independent relationships between AS and ¹H-NMR–assessed glycoprotein profile, multivariate linear regression analyses were performed (stepwise backward procedure). Variables initially included in the linear regression analyses were selected based on univariate correlation results (P < 0.20) or whether they were variables known or likely to be associated with aPWV. Finally, to evaluate the potential utility of serum N-glycans for detecting subjects with aPWV >10 m/s, we developed several logistic regression models, one for each N-glycan–related variable (GlycA concentrations, GlycB concentrations, H/W ratio of GlycA, and H/W ratio of GlycB). To test the discrimination value of GlycA and GlycB (concentrations and H/W ratios), C-statistics from several regression logistic models with an aPWV >10 m/s as the dependent variable were developed. The cutoff value for aPWV was chosen according to previous results showing that the initial stages of retinopathy appear with lower mean levels of aPWV (10 m/s) than those associated with the initial stages of nephropathy (11.0 m/s) or the first CVD event (12.2 m/s) (7). The C-statistic, also known as the receiver operating characteristic (ROC) area under the curve (AUC), is an overall measure of goodness of fit for binary outcomes. Thus, it represents the probability that a randomly selected subject who experienced the outcome (aPWV >10 m/s) will have a higher predicted probability of having the outcome occur than a randomly selected subject who did not experience the outcome (aPWV ≤10 m/s). ROC curves were constructed to represent C-statistic values. The best cutoff values for both GlycA and GlycB (concentrations or H/W ratios) were selected based on the Youden index calculation. Two-tailed P < 0.05 was considered statistically significant. The calculations and figures were done using Stata 13.1 for Macintosh (StataCorp, College Station, TX) and GraphPad Prism (GraphPad Software, San Diego, CA) statistical software.

A total of 84 adults with T1D and 68 healthy control subjects were included in the study. The clinical characteristics are shown in Table 1. When compared with healthy individuals, subjects in the T1D group were older and had a higher prevalence of arterial hypertension and dyslipidemia. They also had a higher BMI, WHR, and systolic blood pressure; worse glycemic control; and higher values of aPWV. By contrast, the T1D group had a better conventional lipid profile, with a lower concentration of total cholesterol, LDL cholesterol, and triglycerides and a higher concentration of HDL cholesterol, likely because of the greater number of individuals on statins in this group and the administration of insulin therapy.

As shown in Fig. 1, subjects with T1D compared with healthy control subjects had significantly higher serum concentrations of GlycA (median 850.4 μmol/L [IQR 781.3–916.1] vs. 652.4 μmol/L [581.5–727.1], P < 0.001) and GlycB (386.1 μmol/L [353.2–426.3] vs. 310.0 μmol/L [280.5–331.9], P < 0.001). They also presented with higher H/W ratios of GlycA (median 16.5 [IQR 14.9–18.1] vs. 15.0 [13.7–16.7], P < 0.001) and GlycB (5.0 [4.6–5.5] vs. 4.0 [3.4–4.3], P < 0.001). These differences were maintained after adjusting for age, sex, classical cardiovascular risk factors (smoking habit, hypertension, dyslipidemia, and BMI), and the presence of T1D (or HbA_1c levels).

The relationship between ¹H-NMR–assessed glycoprotein profile and other variables evaluated in the whole population are shown as a heat map in Supplementary Fig. 1. Glycoproteins positively correlated with age, smoking habit, hypertension, dyslipidemia, BMI, WHR, systolic and diastolic blood pressure, MAP, fasting plasma glucose, HbA_1c, urinary ACR, triglycerides, insulin resistance assessed as the estimated R_d, and skin autofluorescent AGEs. Of note, the skin accumulation of autofluorescent AGEs was positively associated with concentrations of GlycA (r = 0.462, P < 0.001), GlycB (r = 0.342, P < 0.001), and the H/W ratios of GlycA (r = 0.317, P < 0.001) and GlycB (r = 0.425, P < 0.001). No significant correlations between circulating GlycA and GlycB and microvascular complications were found.

We assessed the potential relationship of AS with serum N-glycans. In univariate analyses, aPWV was positively associated with the concentrations of GlycA (r = 0.550, P < 0.001) and GlycB (r = 0.423, P < 0.001), and with the H/W ratios of GlycA (r = 0.453, P < 0.001) and GlycB (r = 0.515, P < 0.001) (Fig. 2). In multivariate analyses, the GlycA concentration (β = 0.212, P < 0.001) and the H/W ratios of GlycA (β = 0.150, P = 0.009) and GlycB (β = 0.155, P = 0.011) were independently associated with aPWV after adjustment for age, sex, classical cardiovascular risk factors (smoking arterial hypertension, dyslipidemia, and BMI), and T1D. The C-statistic for detecting individuals with aPWV >10 m/s was 0.866 (95% CI 0.794–0.937) for GlycA concentrations, 0.705 (0.588–0.821) for GlycB, 0.790 (0.691–0.888) for H/W ratio of GlycA, and 0.862 (0.78–0.943) for H/W ratio of GlycB (Fig. 3). Cutoff values were 877.9 μmol/L (sensitivity 0.82, specificity 0.83) for GlycA concentrations, 354.9 μmol/L (sensitivity 0.82, specificity 0.58) for GlycB concentrations, 16.97 (sensitivity 0.76, specificity 0.71) for H/W ratio of GlycA, and 5.018 (sensitivity 0.88, specificity 0.79) for H/W ratio of GlycB.

Finally, because subjects with T1D were significantly older than control subjects and age might increase several variables included in the analyses, a sensitivity analysis in a subsample of 21 subjects with T1D and 21 age- and sex-matched control subjects was performed (Supplementary Material). The results of this sensitivity analysis were similar to those reported above for the whole population (Supplementary Tables 1–4), reinforcing the main results of the current study.

In the current study, we describe an independent association between vascular aging (assessed as aPWV) and the serum inflammation-related N-glycans GlycA and GlycB. Of potential clinical interest, we provide cutoff values for GlycA concentration and H/W ratio of GlycB for identifying adults with T1D and no CVD but at higher risk for developing both macro- and microvascular complications because of their greater vascular age. These results could be of great utility in clinical practice to better phenotype adults with T1D and could represent a step forward in the development of precision medicine for T1D. We previously reported an association between AS (vascular aging) and a combination of classical circulating biomarkers of low-grade inflammation (e.g., interleukin-6, hs-CRP) in adults with T1D (12). However, the association was not strong enough to support its use in everyday clinical practice for discriminating subjects according to their vascular aging, likely because it performed poorer than GlycA and GlycB when capturing low-grade inflammation (13).

The association between vascular aging (AS) and circulating GlycA and GlycB we identify here fits well with previous studies reporting an association between CVD or conditions associated with high cardiovascular risk and all-cause mortality with N-glycans, especially with GlycA, in apparently healthy populations, in people at higher cardiovascular risk (13), and, more recently, in adults with T1D and subclinical carotid atherosclerosis (26). GlycA has been previously measured in apparently healthy individuals from large population studies such as the Women’s Health Study (27,28) and the Multi-Ethnic Study of Atherosclerosis (MESA) (29,30). Both studies reported that elevated baseline circulating glycoprotein N-acetyl methyl groups were associated with longitudinal risk of CVD incidence and mortality, supporting the concept of GlycA as an important predictor of CVD risk independently of age, ethnicity, smoking, systolic blood pressure, hypertension medications, cholesterol treatment, postmenopausal status, hormone use, BMI, and diabetes (27). In the MESA study, lower levels of GlycA were associated with better cardiovascular health as assessed with the Life’s Simple 7 score (30), and GlycA levels have been shown to predict the progression of calcification of the descending thoracic aorta (31), a major pathogenic mechanism involved in AS (3). Other studies have examined N-glycans in high-risk populations for CVD. For example, glycoprotein acetyls were found to be strongly related to myocardial infarction, ischemic stroke, and intracerebral hemorrhage (32), and AS might play a pathogenic role in all these events (3). Finally, it has recently been reported in a cohort of adults with T1D (>10 years of T1D duration, no previous cardiovascular events) that the presence of carotid atherosclerosis (carotid plaques assessed by B-mode ultrasound imaging) was independently associated with higher circulating concentrations of GlycA and higher H/W ratios of GlycA and GlycB (26). These results further support our findings because the relationship between AS (also called arteriosclerosis) and atherosclerosis would be bidirectional (9).

While the present study was not designed to explore pathophysiological mechanisms linking AS and inflammation, there are at least two potential, nonexclusive, pathophysiological mechanisms that might explain the association. First, the stiffening of the wall of large arteries (e.g., aorta) translates an excessive pulsatile pressure in the microcirculation (especially in organs with low resistance and high flow, such as kidney, retina, brain, or liver). As a consequence, tissue perfusion and capillary transit time are lower, decreasing the supply of oxygen and nutrients and promoting inflammation and oxidative stress. Indeed, a decrease in kidney, retinal, and brain perfusion beds and an increase in the prevalence of nonalcoholic liver disease and liver fibrosis have been reported as aPWV increases (3,33–35). In this case, the inflamed peripheral tissues would be the source of circulating N-glycans. Second, inflammation of the wall of large arteries driven, for example, by the accumulation of lipids in the intima or AGEs in the media, promotes the stiffening process of the arterial wall through several mechanisms, such as rupture of elastin fibers with subsequent substitution by collagen fibers (3). In this case, circulating N-glycans would be released from the inflamed vascular wall. Of note, this second mechanism promotes the stiffening of arterial walls, providing positive feedback to the first mechanism.

Thus far, the assessment of aPWV in patients with T1D has been restricted to its use as an investigational tool, and information is still scarce (9). Indeed, evidence to strongly support the key pathogenic role of vascular aging (aPWV) in the development of both macro- and microvascular complications in T1D is often overlooked. For example, in patients with T1D diagnosed during childhood and adolescence, the prevalence of AS (aPWV >90th percentile of reference values) was found to double the prevalence of nephropathy and retinopathy after a follow-up of 7.9 years (11.6, 5.8, and 5.6%, respectively) (36). These findings are consistent with studies reporting a decrease in kidney and retinal perfusion as aPWV increases (33,34). Similarly, the degree of urinary albumin excretion rate and retinopathy in adults with T1D increases as aPWV increases, and the presence of cardiovascular events is higher in patients with higher aPWV, wherein mean aPWV is 10.0 m/s when the first signs of retinopathy are detectable, 11.0 m/s when microalbuminuria begins, and 12.2 m/s when clinical CVD is clinically apparent (7). Finally, an increase in aPWV of 1 SD (3.3 m/s) over a 6.2-year follow-up in adults with T1D was associated with an increase in the risk of the progression of urinary albumin excretion (adjusted HR 1.69 [95% CI 1.10–2.32]), a decline in eGFR of >30% (HR 1.36 [95% CI 1.06–1.79]), the development of CVD events (HR 1.31 [95% CI 1.01–1.70]), and the risk of all-cause mortality (HR 1.38 [95% CI 1.00–1.85]) (8). In this context, the current study provides, for the first time to our knowledge, cutoff values of GlycA concentrations and for the H/W ratio of GlycB with high sensitivity and specificity and C-statistic values for detecting patients with aPWV >10 m/s who are at higher risk of both micro- and macrovascular complications.

As shown in Table 1, there were some differences between subjects with T1D and control subjects, particularly with regard to age and lipid profile, that deserve mention. First, subjects with T1D were older than the control subjects, and it is well known that both AS and inflammation increase with aging. However, the results remained significant after adjusting for age in the multivariate analysis. Second, subjects with T1D had lower levels of LDL cholesterol than control subjects likely because of the higher percentage of individuals with T1D taking statins. Subjects with T1D also had higher levels of HDL cholesterol, which might be related to insulin therapy. Along this line, it has been previously described that the conventional lipid profile in patients with T1D and good glycemic control is “supernormal,” which is characterized by lower triglycerides and LDL cholesterol levels and by HDL cholesterol levels within the upper normal range (37). This has mainly been attributed to the effects of exogenous insulin through several mechanisms, such an increasing the activity of lipoprotein lipase (38). At first glance, all these differences in the conventional lipid profile should have favored lower aPWV (and inflammation) in the subjects with T1D. Nevertheless, we previously demonstrated that despite having an apparently better conventional lipid profile than healthy control subjects, subjects with T1D show significant proatherogenic alterations in their ¹H-NMR–assessed lipoprotein profile associated with higher aPWV (39). Thus, these findings do not refute the potential role of disturbances in lipid metabolism in the pathogenesis of vascular disease in T1D. Finally, other factors that might increase (e.g., eGFR <60 mL/min/1.73 m², cardiovascular events) or decrease (T1D duration <10 years) AS and/or inflammation were exclusion criteria or were adjusted for in the statistical analyses (e.g., use of antihypertension drugs was included in the hypertension variable).

Our study provides information on the potential association between N-glycans and the other variables analyzed (Supplementary Fig. 1), which should be interpreted with caution. The highly significant positive correlation between skin accumulation of AGEs and changes in circulating GlycA and GlycB is not surprising, as both nonenzymatic glycosylation leading to the formation of AGEs and enzymatic glycosylation leading to the synthesis of N-glycans are increased under conditions of chronic excessive glucose. Although they are biochemically different processes, both are involved in the pathogenesis of the chronic complications of diabetes (40). However, the nonsignificant association between circulating GlycA and GlycB and the presence of microvascular complications could be attributable to the fact that the current study was not designed to address this issue (e.g., patients with T1D with an eGFR <60 mL/min/1.73 m² were excluded, which is a significant bias in evaluating this issue).

Our study has several limitations. First, the study’s cross-sectional design does not allow for the establishment of causal relationships among the variables evaluated. Second, the study included a limited number of adults with T1D with a disease duration of >10 years and no previous CVD events. Accordingly, our findings will need to be replicated in larger cohorts of more heterogeneous populations of adults with T1D, preferably in a prospective manner, before extrapolating results to any adult with T1D. Third, the technique used for detecting serum N-glycans does not allow for the specific identification of the elevated glycoproteins involved. Nonetheless, our objective was not to specifically identify the glycoproteins but, rather, to test the utility of the detection of serum N-glycans with ¹H-NMR spectroscopy in the assessment of vascular aging in adults with T1D. The main strength of our study is the novelty of the concept of searching for serum biomarkers of vascular aging in adults with T1D as a way to better phenotype those individuals at the highest risk for developing macro- and microvascular complications in the context of precision medicine.

In summary, this study establishes an association between changes in the serum inflammation-related N-glycans GlycA and GlycB and vascular aging assessed by aPWV in adults with T1D. It also provides cutoff values for GlycA concentrations and the H/W ratio of GlycB, with an exciting potential in clinical practice to categorize patients with T1D. Further studies are needed to confirm these results.

Acknowledgments. The authors thank all the individuals enrolled in this study and their physicians for their participation, Dr. Kenneth McCreath (freelance science editor and writer, Madrid, Spain) for editing the English version of the manuscript, and the Biobank-IISPV (B.0000853 + B.0000854), part of the Spanish National Biobanks Network (RD09/0076/0049), for its cooperation.

Funding. Financial support was provided by a grant to J.-M.G.-C. from the Ministerio de Ciencia e Innovación (project PI15/00567), cofinanced by the Instituto de Salud Carlos III (ISCIII) and by the European Regional Development Fund (ERDF). It was also supported by the CB07708/0012, CIBERDEM (Spanish Biomedical Research Center in Diabetes and Associated Metabolic Disorders), cofinanced by the ISCIII (Ministerio de Ciencia e Innovación) and the EDRF. CIBERDEM is an initiative of the ISCIII.

Duality of Interest. N.A. is a stock owner of Biosfer Teslab, the company that commercializes the glycoprotein profile described in this article. R.F.-M. is a Biosfer Teslab employee. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. G.L. performed the statistical analyses and contributed to the writing, discussion, and review of the manuscript. N.A. and R.F.-M. performed the ¹H-NMR spectroscopy analyses and contributed to the writing, discussion, and review of the manuscript. A.R., A.C., and L.A. recruited all subjects, performed their clinical evaluation (including the assessment of aPWV), and contributed to the discussion of the manuscript. O.G.-P. performed the data collection and contributed to the discussion of the manuscript. E.B. was in charge of the storage of blood samples and conventional analytical determinations and contributed to the discussion of the manuscript. S.F.-V. contributed to the discussion and review of the manuscript. X.C. contributed to the data analyses and discussion and review of the manuscript. J.V. contributed to the design of the study and the writing, discussion, and review of the manuscript. J.-M.G.-C. contributed to the design, statistical analyses, and writing, discussion, and review of the manuscript. J.-M.G.-C. 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.