Left Ventricular Mass Increases With Deteriorating Glucose Tolerance, Especially in Women: Independence of Increased Arterial Stiffness or Decreased Flow-Mediated Dilation: The Hoorn Study

For the present investigation, we used data from the 2000 Hoorn Study follow-up examination (19) and the Hoorn Screening Study (24). Briefly, the Hoorn Study is a study of glucose metabolism in the general population (n = 2,484) that started in 1989 (25). In 2000, a follow-up examination was carried out among all surviving participants. We invited all those who had diabetes, as determined by an oral glucose tolerance test or who were treated for diabetes at the 1996 follow-up (n = 176). We also invited random samples of individuals with normal glucose metabolism (NGM) (n = 705) and IGM (n = 193). Of 1,074 individuals thus invited, 648 (60%) participated. Additionally, we invited 217 individuals with type 2 diabetes from the Hoorn Screening Study (24), of whom 188 (87%) participated. Data on 14 individuals were missing due to logistical problems. The study population (n = 822) thus consisted of three groups: 290 with NGM, 187 with IGM, and 345 with type 2 diabetes. The study was approved by the local ethics committee. All participants gave their written informed consent.

Among the 455 nonparticipants (53% women), 13% were complete nonresponders. The remaining nonparticipants gave various reasons not to participate: lack of interest (30%), comorbidity (23%), age (7%), unwillingness to travel (6%), participation too time consuming (6%), and miscellaneous reasons (15%).

An experienced research technician unaware of the participants’ clinical status or GTS obtained an echocardiogram in each participant, according to a standardized protocol, with the use of a single ultrasound scanner (HP SONOS 5500; Hewlett Packard, Andover, MA). M-mode recordings were digitally stored and read according to the guidelines of the American Society of Echocardiography (26).

Left ventricular end-diastolic diameter (EDD), posterior wall thickness (PWT), and interventricular septum thickness (IVS) were measured at end diastole. LVM was calculated as 0.8(1.04) ([EDD + IVS + PWT]³ − EDD³) + 0.6 (in grams) and relative wall thickness (RWT) as (IVS + PWT)/EDD (27) Each echocardiogram was inspected afterward by a senior cardiologist blinded to the participants’ clinical status or GTS in order to monitor the quality of both recordings and readings. Left ventricular geometric patterns were classified according to Heesen et al. (28).

Carotid and femoral arterial stiffness indexes were determined by ultrasonography, using previously described techniques (19) and calculated as follows. Distensibility coefficient = (2ΔD × D + ΔD²)/(ΔP × D²) (in 10⁻³ · kPa⁻¹); compliance coefficient = π (2D × ΔD + ΔD²)/(4 × ΔP) (in mm² · kPa⁻¹); and Young’s elastic modulus = D/(intima-media thickness × distensibility coefficient) (in kPa). ΔD is distension, D is diameter, and ΔP is local pulse pressure. Local pulse pressures were determined with the use of distension waveform calibration (29). The distensibility coefficient reflects the arterial elastic properties, whereas the compliance coefficient reflects the arterial buffering capacity and Young’s elastic modulus indicates the intrinsic elastic wall properties (30).

Systemic arterial compliance (ml/mmHg) was determined according to the exponential decay method (31) and the augmentation index from pressure wave analysis (32), both with the use of tonometry-derived aortic pressure waveforms (Sphygmocor, Moreton-in-Marsh, U.K.), as previously described (33). Systemic arterial compliance and the augmentation index represent the overall buffering capacity of the arterial system and the overall stiffness of the arterial system, respectively, taking into account the contribution of reflected (backward) pulse waves to aortic blood pressure.

The carotid-femoral transit time (in ms) was determined with the use of the ultrasonically recorded carotid and femoral distension wave forms (33) and represents the average stiffness over the aortic segment.

We measured brachial artery endothelium-dependent vasodilation and the presence of atherosclerosis (both as defined below) as proxies for coronary artery endothelium-dependent vasodilation (34,35).

All individuals underwent an ultrasound examination according to guidelines of the International Brachial Artery Reactivity Task Force (36). Briefly, baseline arterial diameter and peak flow velocity were determined. A blood pressure cuff was placed on the forearm, inflated, and kept constant at supra-systolic pressure to induce forearm ischemia. After 5 min, the cuff was released, resulting in an increase in arterial blood flow. This increase in blood flow increased shear stress, which served as the stimulus for flow-mediated dilation. After cuff release, maximum peak flow velocity was measured within 15 s and arterial diameter at 45, 90, 180, and 300 s. The maximum diameter in any of these four measurements was used in statistical analysis. After 15 min of rest to reestablish arterial baseline conditions, endothelium-independent, nitroglycerin-mediated dilation was determined as follows: baseline arterial diameter and peak flow velocity were redetermined. Nitroglycerin (400 μg, Nitrolingual Spray; Pohl-Boskamp, Hohenlockstedt, Germany) was then sublingually administered; after 5 min, arterial diameter and peak flow velocity were again determined.

Health status, medical history, medication use, and smoking habits were assessed by a questionnaire (25). Systolic and diastolic pressure; pulse pressure; hypertension; glucose; HbA_1c; serum total, HDL, and LDL cholesterol; serum triglycerides; BMI; waist-to-hip ratio; microalbuminuria; and ankle-brachial pressure index were determined as described elsewhere (24,25). Insulin resistance was calculated according to the homeostasis model assessment (37). Resting electrocardiograms were coded according to the Minnesota Code. Prior CVD was defined as Minnesota Code 1.1–1.3, 4.1–4.3, 5.1–5.3, or 7.1 on the electrocardiogram, coronary bypass operation or angioplasty, and/or peripheral arterial bypass or amputation. The presence of prior CVD, peripheral arterial disease as defined by an ankle-brachial pressure index <0.9, and/or a carotid intima-media thickness >1.05 mm (i.e., the 90th percentile of the distribution in individuals with NGM) was considered to indicate the presence of atherosclerosis.

Reproducibility was assessed in 10 individuals (5 men; age 58.2 ± 9.5 years) who were examined twice at 2-week intervals. The intraobserver intersession coefficients of variation ([SD of the difference/√2]/pooled mean) were distensibility coefficients 7.0% (carotid), 11.3% (femoral), and 12.8% (brachial); compliance coefficients 6.3% (carotid), 13.1% (femoral) and 13.9% (brachial); carotid elastic modulus 11.6%; carotid-femoral transit time 4.4%; aortic augmentation index 3.6%; systemic arterial compliance 8.3%; and endothelium-dependent and -independent dilation, expressed as absolute diameter, 14.7 and 10.3%.

All analyses were carried out with SPSS (SPSS, Chicago, IL). We used ANCOVA with linear contrast to investigate trends in left ventricular mean values across the categories of glucose tolerance. All statistically significant trends were tested whether they deviated from linearity. The associations between glucose tolerance and the measurements of cardiac structure and geometry were examined separately for men and women because, out of five previous population-based studies (5,7,9,11,12), three reported strong sex-specific differences (5,9,11), particularly so for LVM.

We first analyzed the associations without any adjustments (crude model) and then with adjustments for potential confounders (adjusted models). Because left ventricular structure is known to be affected by age, height, BMI, and blood pressure (38), these variables were considered first in the adjusted models. Arterial stiffness indexes, FMD, and the presence of atherosclerosis were then separately added into the models, as were estimates of hyperglycemia/-insulinemia and insulin resistance.

Individuals with impaired fasting glucose (n = 63) and impaired glucose tolerance (n = 116) did not significantly differ from each other with regard to any of the analyses and were therefore combined. P values <0.05 were considered statistically significant.

In 42 of the 822 participants, LVM could not be determined due to either a high BMI (n = 33; BMI of those with qualitatively satisfactory examinations versus those without, 27.5 ± 3.8 vs. 37.9 ± 8.9 kg/m², P < 0.001) or a poor transthoracic window (n = 9). In the remaining 780 individuals, carotid and femoral stiffness indexes and FMD could be determined in 739, 648, and 636 individuals, respectively. The main reason for missing arterial data was poor definition of the arterial wall due to obesity (BMI of those with qualitatively satisfactory examinations versus those without, 26.9 ± 3.3 vs. 30.5 ± 4.5 kg/m², P < 0.001).

Table 1 shows the characteristics of the entire study population. Both IGM and type 2 diabetes were characterized by a worsening cardiovascular risk factor profile compared with NGM.

In women, LVM increased significantly with deteriorating glucose tolerance (P for trend <0.001) (Table 2). Similar trends could be observed for EDD, PWT, and IVS. RWT did not show any significant trend (P for trend = 0.358). The proportion of women with a normal geometric pattern decreased significantly with deteriorating glucose tolerance (P for trend ≤0.002), whereas the proportions of women with either eccentric or concentric hypertrophy increased significantly (P for trend = 0.042 and 0.039, respectively). No statistically significant trend was observed for concentric remodeling. In men, neither the measures of left ventricular structure nor the distribution of geometric patterns showed any significant trends with deteriorating glucose tolerance (Table 2).

In women, after adjustment for age, height, BMI and mean arterial pressure, glucose tolerance remained statistically significantly associated with increased LVM (P for trend <0.018) (Table 3, model 2). Additional adjustment for carotid distensibility, FMD, and the presence of atherosclerosis did not materially change the results (P for trend = 0.030, 0.043, and 0.028, respectively, models 3–5), even though carotid distensibility, FMD, and the presence of atherosclerosis were significantly associated with GTS (Table 1) and, after adjustment for age and mean arterial pressure, LVM (data not shown). A similar pattern could be observed for the association between glucose tolerance and end diastolic diameter (models 1–5).

Different effects were observed for PWT, IVS, and RWT. The associations between glucose tolerance and both PWT and IVS were not statistically significant after adjustment for age, height, BMI, and mean arterial pressure (Table 3, models 1 and 2), whereas no significant associations, either crude or adjusted, could be observed between GTS and RWT (RWT, models 1–5). Additional investigations to explore the individual impact of the variables age, height, BMI, and mean arterial pressure revealed that the associations of GTS with PWT and left ventricular wall thickness were rendered insignificant after adjustment for mean arterial pressure, whereas the association with IVS thickness was rendered insignificant after adjustment for BMI (detailed data not shown).

In men, neither crude (P for trend, all ≥0.096) nor adjusted (P for trend, all ≥0.218) analyses showed statistically significant associations between GTS and ventricular structure. Nevertheless, the P value for interaction (sex times GTS) was not significant (P = 0.11).

Results were similar when BMI was replaced by waist-to-hip ratio and when mean arterial pressure was replaced by systolic, diastolic, or pulse pressure (detailed data not shown). If we replaced IVS or PWT by left ventricular wall thickness, i.e., the sum of both, the crude association with glucose tolerance lost significance after adjustment for the variables of model 2 (detailed data not shown). Results were also similar when carotid distensibility was replaced by other carotid or femoral stiffness indexes or by systemic arterial compliance, augmentation index, or the carotid-femoral transit time (data not shown).

Our study population was too small to further explore any underlying mechanisms between glucose tolerance and ventricular geometry.

To estimate the contribution of hyperglycemia/-insulinemia or insulin resistance to the increase in LVM associated with IGM and type 2 diabetes, we compared the LVM analyses (adjusted for age, height, BMI, and mean arterial pressure) with those additionally adjusted for HbA_1c (or fasting or postload glucose), insulin, and insulin resistance. This showed that in women, at most 7% of the association between glucose tolerance and LVM could be explained by indexes of hyperglycemia/-insulinemia or insulin resistance.

Additional adjustment for brachial stiffness indexes, heart rate, lipid profile, the use of lipid-lowering or antihypertensive medication including ACE inhibitors, smoking, or (micro-)albuminuria did not materially alter our results (detailed data not shown). Results were also not materially altered if we excluded individuals with left ventricular wall motion abnormalities (n = 55) (data not shown). Finally, if we replaced LVM mass (in grams) by LVM index, i.e., LVM divided by body surface area (in grams per meters squared), our results were again not altered (data not shown).

The present population-based study on the association between glucose tolerance and cardiac structure had three main findings. First, glucose tolerance was independently associated with increased LVM in women, but not clearly so in men. Second, the association between glucose tolerance and LVM in women was not mediated by glucose intolerance-associated increased arterial stiffness and impaired endothelial function. Third, indexes of hyperglycemia/-insulinemia or insulin resistance explained at most 7% of the association between glucose tolerance and LVM.

Our study extends previous population-based investigations (5–7,9,11,12), as we are the first to examine the impact of arterial stiffness and endothelial function on the association between glucose tolerance and cardiac structure. In addition, to be able to sufficiently adjust for potential confounders, we extensively characterized our study population in terms of CVD risk factors, prior CVD, and glucose metabolism.

Our results are in general agreement with previous population-based studies on the association between glucose tolerance and LVM (5,7,9,11,12). However, it is currently unclear whether (5,9,11) or not (7,12) this association is stronger in women than in men. In our investigation, the statistical test for interaction between sex and glucose tolerance on LVM was P = 0.11. Therefore, the result of our interaction analysis neither excluded nor proved the existence of a sex-specific association.

We suggest that the independent association we report between glucose tolerance and EDD, as well as the absence of such an association with IVS or PWT, may be explained by the fact that individuals with type 2 diabetes are characterized by an expanded extracellular volume (39), which has been shown to increase left ventricular chamber size relatively more than myocardial wall thickness (40). However, it remains poorly understood why the left ventricle apparently remodels differentially over its regions (41), although local dispersion of ventricular pressure, ventricular flow patterns (42), and (sex-specific) nonhemodynamic factors are thought to play a role (43,44) in addition to specific, diabetes-associated alterations in cardiac structure (5,7,9,12). Taken together, these (5–7,9,12) and the present data show that the left ventricle is subject to maladaptive alterations with deteriorating glucose tolerance, especially in women. This sex-specific increase in LVM may explain, at least in part, the increased CVD risk in women with type 2 diabetes (45).

Our study showed that increased carotid and femoral arterial stiffness, or any of the other stiffness indexes, influenced the association between glucose tolerance and measurements of cardiac structure only minimally, despite the fact that deteriorating glucose tolerance was associated with increased arterial stiffness (19) and that increased arterial stiffness was associated with increased LVM. The association between glucose tolerance and LVM was also not explained by impairment of endothelial function, as estimated by brachial artery FMD and the presence of atherosclerosis, even though deteriorating glucose tolerance was associated with impaired endothelial function, which in turn was associated with increased LVM. Decreased endothelial synthesis or bioavailibility of nitric oxide may contribute to increases in LVM by loss of its orchestrating role in the synthesis of extracellular matrix components and by shifting the local myocardial homeostasis toward hypertrophy (22,23,46). Our results therefore suggest that other pathophysiological mechanisms may play a role, such as increased oxidative and carbonyl stress (47), chronic low-grade inflammation (48), and(or) autonomic neuropathy (49). We cannot exclude, however, that the methods we used are insufficiently precise estimates of coronary endothelial function, which, however, is difficult to directly measure in large-scale studies.

Hyperglycemia/-insulinemia and insulin resistance have been implicated in the development of increased LVM (15–18). In our study, the association between GTS and LVM could only be explained to a minor extent (<7%) by any of these variables. These observations are in line with several previous (population-based) studies (5,50–53) However, conflicting results have been published in relatively small, select, non-population-based investigations (15–18) and one population-based investigation among non-Caucascian individuals (14).

Our study had several limitations. First, it should be kept in mind that the cross-sectional nature of our data does not allow us to make strong causal inferences. Second, our results might have been influenced by the coexistence of CVD affecting left ventricular shape. However, it has been shown that LVM can be estimated accurately from M-mode echocardiography, even in the presence of CVD (27). Nevertheless, to circumvent this possibility, we adjusted for CVD in our analyses. Third, due to the rapid responsiveness of the body to hormonal and metabolic changes, determination of insulin resistance by homeostasis model assessment method is subject to biological variability, which might have weakened its association with LVM. However, recommendations (37) made to diminish this source of variation are not easily applicable in large population-based studies. Finally, our results were obtained in a relatively elderly Caucasian population, and therefore, it remains to be determined whether our results can be generalized to other ages and ethnicities (14).

In conclusion, our data expand the conceptual view of the pathogenesis of glucose tolerance-related changes in LVM, as we show that the increase in LVM in women is independent of increased arterial stiffness, impaired FMD, or the presence of atherosclerosis. In addition, we show that this increase in LVM is only minimally explained by indexes of hyperglycemia/-insulinemia or insulin resistance. Our data may explain, in part, the increased CVD risk seen in women with deteriorating glucose tolerance.

This study was supported by the Netherlands Heart Foundation (grant 98154).