Central Pulse Pressure and Its Hemodynamic Determinants in Middle-Aged Adults With Impaired Fasting Glucose and Diabetes: The Asklepios study

The Asklepios study recruited a representative cohort of apparently healthy, community-dwelling male and female volunteers aged 35–55 years, sampled from the twinned Belgian communities of Erpe-Mere and Nieuwerkerken as previously described (9). Subjects with clinically evident atherosclerosis/atherothrombosis, malignant tumors, cardiac disease, atrial fibrillation, renal disease, hepatic insufficiency, previous organ transplant, life expectancy <5 years, or pregnancy in the preceding 6 months were excluded. All analyses are based on data acquired during the baseline study visit, which occurred between October 2002 and October 2004. Central hemodynamic data derived from aortic pressure-flow analyses were available from 2,368 subjects (1,223 women and 1,145 men), which constituted the population for this study. None of the participants had a history of cardiovascular disease or stroke. All subjects provided informed consent. This study was approved by the ethical committee of the Ghent University Hospital and the University of Pennsylvania Institutional Review Board. IFG and diabetes were defined according to current American Diabetes Association criteria (10).

Brachial blood pressure was measured with a validated oscillometric device (Omron HEM-907 device; Omron, Matsuka, Japan). Hypertension was defined as a brachial systolic blood pressure (SBP) ≥140 mmHg, diastolic blood pressure (DBP) ≥90 mmHg, or antihypertensive medication use. Echocardiographic examinations were performed using a Vivid-7 ultrasound platform (GE Healthcare Vingmed Ultrasound, Trondheim, Norway) (9). Pulsed-wave Doppler measurements of flow velocities in the left ventricular outflow tract were performed and recorded, placing the Doppler sample immediately proximal to the aortic valve leaflets within the centerline of the left ventricular outflow tract. We computed left ventricular outflow tract cross-sectional area from its radius measured in the parasternal long axis view (area = pir²). Flow recordings were performed along with carotid artery applanation tonometry recordings using a Millar pen-type high-fidelity tonometer (SPT 301; Millar Instruments, Houston, TX) and dedicated hardware and software for acquisition of the arterial pulse (11). Carotid (central) pressure waveforms were calibrated according to brachial mean and diastolic pressure (11) because in contrast to systolic pressure, mean and diastolic pressure exhibit little variation in the arterial tree (12,13).

Pressure and Doppler flow velocity recordings were processed offline using custom-designed software written in Matlab (The Mathworks, Natick, MA) as previously described (11). Proximal aortic Zc was computed in the time domain as the slope of the early systolic pressure-flow relation using the method described by Mitchell and colleagues (11,14). Reflection magnitude was computed using wave-separation analysis (11,12). In this method, after separation of the pressure waveform into its forward and backward components, reflection magnitude is computed as the ratio of the amplitudes of the backward/forward components (11,12). Reflected wave transit time was also computed from wave-separation analysis (15). TAC was calculated with the pulse pressure method (11). Cardiac output was computed as the product of stroke volume and heart rate. Systemic vascular resistance (SVR) was computed as mean arterial pressure/cardiac output.

We measured aortic diameter as the distance between the internal borders (intima) of the aortic walls and in the proximal ascending aorta 1 to 2 cm distal to the sinotubular junction. Aortic cross-sectional area was computed as pir². Aortic root PWV was calculated from the water hammer equation: PWV = (Zc × A)/ρ, where Zc is aortic characteristic impedance, A is ascending aortic cross-sectional area, and ρ is blood density, assumed to be 1.06 g/cm³. The product of aortic elastic modulus and wall thickness (Eh) was then computed using a rearranged Moens-Korteweg equation as follows: Eh = PWV² × ρ × D, where D is measured ascending aortic diameter. Proximal aortic diameter could be reliably measured in 1,927 subjects (1,654 subjects with normal FG, 240 subjects with IFG, and 33 subjects with type 2 diabetes).

Carotid Zc was measured as the slope of the early systolic carotid pressure-flow relation in a carotid pressure-flow loop as previously described (16). Carotid PWV and Eh were computed using carotid Zc and diameter as described above for the ascending aorta. Carotid PWV was computed using the Moens-Korteweg equation as described above for the ascending aorta.

Aortic root dimension was compared between the groups after allometric normalization for body surface area (BSA). To assess the normal allometric relation between aortic root area and BSA, we first selected a healthy reference sample as previously described (17). We used the following general allometric equation: y = ax^b + ε, where x is BSA, a and b are parameters, and ε is a random additive error term. The allometric model also included a sex term to satisfy the group-difference principle (17,18). The allometric power to normalize aortic area for BSA was estimated using nonlinear regression, in which an iterative technique is applied to estimate model parameters while maximizing data fit (17).

Supplementary Table 1 summarizes the definitions of key hemodynamic indices and measures of arterial stiffness used in this study.

Hemodynamic indices were compared between subjects with normal FG, IFG, and type 2 diabetes using ANOVA. Eh was log-transformed to improve normality for statistical models. Age- and sex-adjusted comparisons were performed with ANCOVA. For comparisons of pulsatile hemodynamic indices, further adjustments for MAP were performed, because MAP determines the operating wall stiffness (and arterial compliance) for any given wall-material properties (12). Because the relationship between MAP and stiffness is nonlinear, we included both a MAP and a MAP-squared term in these models. MAP and MAP-squared were mean-centered to minimize colinearity. Similarly, all hemodynamic indices that depend on volume flow (stroke volume, cardiac output, SVR, Zc, and TAC) are markedly influenced by body size and were therefore normalized for BSA, as estimated with the Gehan method (19) according to previous analyses that derived appropriate allometric powers that linearize the normal relationship between BSA and hemodynamic indices in this population (17). Therefore, for stroke volume, cardiac output, SVR, and TAC, a BSA term was introduced to all linear statistical models as a covariate, whereas for aortic Zc, BSA was raised to the power of −0.64 and introduced in the linear models (17). All probability values are two-tailed. Statistical significance was defined as α < 0.05. Statistical analyses were performed using SPSS for Mac OS v19 (SPSS Inc., Chicago, IL).

Baseline characteristics of study subjects with normal FG, IFG, and type 2 diabetes are shown in Table 1. Age, BMI, waist circumference, serum fasting triglycerides, serum high-sensitivity C-reactive protein, the prevalence of hypertension, and antihypertensive medication use were highest in subjects with type 2 diabetes and lowest in those with normal FG, whereas HDL cholesterol was highest is subjects with normal FG and lowest in subjects with type 2 diabetes. Men constituted a higher proportion of the IFG or type 2 diabetes group compared with the normal FG group. Heart rate was significantly higher among subjects with type 2 diabetes compared with either those with IFG or those with normal FG. There were no significant differences in cardiac index or stroke volume index between the groups. Subjects with IFG demonstrated significantly greater SVR compared with subjects with normal FG.

Thirteen of the 37 subjects with type 2 diabetes had a previous diagnosis (i.e., known type 2 diabetes). Among these subjects, mean diabetes duration was 4.8 years (interquartile range 2.3–6.3 years), 11.8% were treated with diet only, 52.9% were treated with a sulfonylurea, 70.6% were treated with metformin, and 5.9% were treated with insulin. The proportions of subjects treated with one, two, or three or more drugs were 52.9, 29.4, and 5.9%, respectively, and median HbA_1c was 6.7% (interquartile range 5.15–8.1%).

Brachial and central SBP, PP, DBP, and MAP were significantly lower in subjects with normal FG compared with subjects with IFG and type 2 diabetes, without significant differences between the latter two groups (Table 2). After adjustment for age and sex, the differences in brachial or carotid PP disappeared, whereas the differences in SBP, MAP, and DBP persisted. Similarly, after further adjustment for MAP, no differences in brachial or central PP were observed between the groups.

Cross-sectional proximal ascending aortic areas were significantly greater in subjects with IFG or type 2 diabetes compared with those with normal FG in unadjusted analyses (Supplementary Table 2). The allometric power to normalize aortic cross-sectional area for BSA was 0.50 (95% CI 0.14–0.86). Aortic area was directly associated with increasing age (P < 0.0001) and male sex (P < 0.0001). After adjustment for age and sex, allometrically normalized aortic areas were not significantly different between subjects with normal FG, IFG, and type 2 diabetes.

In unadjusted analyses, subjects with type 2 diabetes demonstrated a greater carotid-femoral PWV, aortic root PWV, and aortic root Eh and a lower TAC index compared with subjects with normal FG (Table 3). After adjustment for age and sex, subjects with type 2 diabetes demonstrated a higher aortic root Zc index, aortic root PWV, aortic root Eh, and carotid-femoral PWV and a lower TAC index and reflection magnitude compared with the other groups, without significant differences between subjects with IFG and normal FG in any of these hemodynamic indices. Similar differences were observed after further adjustment for MAP. There were no significant differences in reflected wave-transit time between the groups. These trends were not appreciably changed with further adjustment for BMI or for serum HDL cholesterol, LDL cholesterol, or triglycerides (data not shown).

After adjustment for age, sex, and MAP, no significant differences were found between the groups in carotid Zc, carotid diameter, carotid PWV, or carotid Eh (Supplementary Table 3). Similarly, reflection magnitude measured at the carotid site was not significantly different between the groups.

We comprehensively characterized, for the first time, central PP and its hemodynamic determinants in subjects with IFG and type 2 diabetes. We demonstrate that IFG and type 2 diabetes are associated with an increase in MAP and central PP. However, in contrast to type 2 diabetes, after adjustment for age, sex, and MAP, IFG was not associated with intrinsic arterial stiffening or abnormalities in pulsatile arterial hemodynamic function. Subjects with type 2 diabetes, in contrast, demonstrated an increase in proximal aortic Zc (which resulted from an increase in the elastance-thickness product, rather than an abnormal aortic diameter), an increase in carotid-femoral PWV, and a decrease in TAC. In contrast, carotid size, Zc, distensibility, and elastance-thickness product were not abnormal in type 2 diabetes or IFG. Subjects with type 2 diabetes also demonstrated paradoxically lower wave reflections. Our findings are important because they demonstrate that 1) large artery stiffening, rather than muscular artery stiffness, wave reflections, or aortic geometric remodeling, is the key abnormality underlying the higher central pulsatile load type 2 diabetes; 2) large artery stiffening is only apparent in established type 2 diabetes rather than IFG, although subjects with IFG had greater operant PP due to their higher MAP; and 3) type 2 diabetes is associated with reduced reflection magnitude, which suggests increased penetration of pulsatility to the microvasculature, which may promote target organ damage.

Type 2 diabetes was associated with an increased aortic Zc, due to an increased aortic root elastance-thickness product, but not a reduced ascending aortic cross-sectional area. Aortic root stiffness and hemodynamic function have not been previously examined in type 2 diabetes. This segment exerts an important influence on central PP and is not assessed with usual measurements of carotid-femoral PWV. Our findings are consistent with those of a previous study that reported greater aortic Zc among 17 individuals with type 1 diabetes, compared with nondiabetic control subjects (8). Also, consistent with our results, most (20–23), but not all (24,25), previous studies failed to show a significant association between type 2 diabetes and aortic dimension. We note that this is the first study to perform allometric normalization of aortic dimension for body size, an important methodologic aspect. We found that ascending aortic area relates to the square root of BSA (allometric power 0.5), statistically rejecting a linear relationship, thus precluding linear ratiometric adjustments. It follows that linear adjustment for BSA will overcorrect ascending aortic area in subjects with high BSA and undercorrect it in those with low BSA. Since type 2 diabetes is associated with obesity and thus a larger BSA, this may explain the slightly lower linearly adjusted aortic size values in type 2 diabetes in a previous study (25).

We also found a decrease in TAC among patients with diabetes. Arterial compliance is directly proportional to arterial size and distensibility. In turn, for any given relative geometry (wall-to-thickness ratio), distensibility is directly related to the elastic modulus of the wall material. The TAC is provided mainly by large arteries, although smaller arteries do contribute (12,13,26). We were unable to detect abnormalities in aortic diameter or carotid stiffness or diameter, suggesting that the decrease in TAC seen in type 2 diabetes is purely the result of aortic wall stiffening. Although we only assessed the aortic root, other studies have similarly failed to show smaller distal aortic segments in type 2 diabetes (21,27,28).

Potential mechanisms of arterial wall stiffening in type 2 diabetes include excessive extracellular matrix deposition of collagen or enhanced collagen cross-linking by advanced glycation end products (29,30), aortic wall calcification (31,32), endothelial dysfunction, chronic low-grade inflammation, increased oxidative stress, and increased sympathetic tone (33). Given the prognostic value of arterial stiffness in type 2 diabetes (34), further studies are required to better understand its mechanistic determinants in order to design therapies to reduce it.

The role of large artery stiffness in prediabetic states is also a subject of great interest. Glycemic spikes have been associated with a higher PP in patients with impaired glucose tolerance and essential hypertension (35), whereas PP and arterial stiffness have been associated with insulin resistance (36). In a recent study among high-risk hypertensive subjects (37), PP measured at baseline was a significant predictor of new-onset diabetes. Therefore, it was suggested that the relationship between large artery stiffness and type 2 diabetes is bidirectional. Microvascular alterations in association with increased arterial stiffness leading to impaired tissue perfusion and impaired insulin-mediated changes in muscle perfusion and glucose metabolism have been suggested as potential mechanisms by which increased arterial stiffness may promote the development of type 2 diabetes (33). Our findings indicate that greater large artery stiffness is not yet present in subjects with IFG, whereas prominent stiffening leading to a wide range of hemodynamic abnormalities are present in those with type 2 diabetes. This may indicate that aortic stiffening occurs only after type 2 diabetes has established or that subjects with IFG who demonstrate stiff large arteries preferentially develop to type 2 diabetes. Indeed, both mechanisms may be at play, and longitudinal studies are required to assess this issue.

Abnormal interactions between large and small vessels may also be important in the progression of microvascular complications in established type 2 diabetes. Retinopathy and nephropathy have been shown to be associated with carotid-femoral PWV (but not with carotid-radial PWV, a measure of muscular artery stiffness) in type 2 diabetes (38). In contrast to the clearly higher aortic wall stiffness in type 2 diabetes, we did not find greater carotid Zc, abnormal carotid diameter, or increased elastance-thickness product in type 2 diabetes. Our data regarding carotid diameter are consistent with a previous study (39). This selective stiffening of the aorta in type 2 diabetes without stiffening of more distal arteries may promote more penetration of pulsatile energy into the microcirculation of the brain and kidneys and reduce partial reflections in proximal arterial bifurcations. This may explain the apparently paradoxical observation of a lower reflection magnitude in subjects with type 2 diabetes. In a previous study, we showed that type 2 diabetes is associated with a lower augmentation index, a rough index of wave reflections, in a large multiethnic sample (40). Our finding is novel and contrasts with the current thinking that links diabetes with increased wave reflections (33), although it should be taken conservatively, because antihypertensive medication use was more common in type 2 diabetes and may have impacted the properties of muscular arteries, leading to decreased wave reflections.

Our study has limitations. Oral glucose tolerance tests were not performed, and our definition of IFG was based on a single measurement of FPG rather than repeated measurements. It is possible that individuals with normal FG in our study may include some individuals with impaired glucose tolerance. Our findings are restricted to a middle-aged population without established cardiovascular disease, which, however, was suitable to assess early vascular changes in diabetes and IFG. We used carotid pressure as a surrogate for aortic pressure, a limitation inherent to the noninvasive nature of our study in a large general population sample. Our comparisons were also limited by the fact that more subjects in the IFG and type 2 diabetes groups were receiving antihypertensive medication. We note, however, that this therapy would bias the comparisons toward the null and that we found greater (not lower) levels of MAP and intrinsic arterial stiffness in type 2 diabetes. Antihypertensive therapy, however, may have affected our comparisons of reflection magnitude (which was lower in type 2 diabetes), and therefore, findings regarding reflection magnitude should be taken conservatively, as discussed above. Our study aimed to characterize prevalent abnormalities in pulsatile load in diabetes and IFG, but specific data regarding retinal or renal microvascular complications were not available.

In summary, we comprehensively characterized, for the first time, pulsatile arterial hemodynamic abnormalities in IFG and type 2 diabetes in a large community-based sample of middle-aged adults. We found that IFG is associated with a greater MAP but not with intrinsic large artery stiffening, whereas type 2 diabetes is associated with higher aortic root Zc, elastance-thickness product, PWV, carotid-femoral PWV and lower TAC, without significant differences in aortic root or carotid diameter, Zc, or Eh. Our findings, therefore, identify aortic wall stiffening, rather than muscular arterial abnormalities, aortic root geometric remodeling, or wave reflections, as the key abnormality that increases pulsatile load type 2 diabetes. Future longitudinal studies are required to assess the role of the hemodynamic macrovascular–microvascular cross-talk in the development of microvascular complications in type 2 diabetes and in the risk of progression of earlier stages of the metabolic disease process to frank type 2 diabetes.

This study was supported by Fonds voor Wetenschappelijk Onderzoek Vlaanderen Grant G.0.838.10 and the American Heart Association Research Award 0885031N. J.A.C. received significant (>$10,000) grants from the National Institutes of Health and the American Heart Association for research studies related to arterial hemodynamics and has received minor support (equipment loans) from Atcor Medical, Cardiodynamics, and APC Cardiovascular Ltd. No other potential conflicts of interest relevant to this article were reported.

J.A.C. was responsible for study design, data analysis, and manuscript preparation. P.S., T.C.G., M.L.D.B., C.M.V.d., D.D.B., and E.R.R. were responsible for parent cohort study design and study design and reviewed and edited the manuscript. Z.A.K. and U.K. were responsible for data analysis and reviewed and edited the manuscript. J.A.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.

Data from this study were presented at the Annual Scientific Session of the American College of Cardiology, San Francisco, California, 9–11 March 2013.