Cerebral Blood Flow Velocity and Periventricular White Matter Hyperintensities in Type 2 Diabetes

TCD and MRI studies were conducted in the SAFE (Syncope and Falls in the Elderly) Laboratory and at the Magnetic Resonance Imaging Center at the Beth Israel Deaconess Medical Center using a General Electric 3.0 Tesla VHI scanner. All subjects were recruited consecutively and provided informed consent approved by the institutional review boards at the Beth Israel Deaconess Medical Center and the Joslin Diabetes Center. Demographic and clinical characteristics are summarized in Table 1. The diabetic group consisted of 28 patients (16 men and 12 women) with type 2 diabetes (duration 2.8 ± 11.5 years; means ± SD). The control group consisted of 22 healthy subjects (12 men and 10 women) who were normotensive, had normal HbA_1c (A1C) levels, and were not treated for any systemic disease. All subjects were screened with a medical history, physical examination, and standard battery of autonomic tests (12). Laboratory chemistries included routine glucose, A1C, lipid and renal panels, differential white blood cell count (WBC), and urine chemistry panel. Soluble intracellular adhesion molecule-1 (sICAM-1), soluble vascular adhesion molecule, endothelin-1, and human interleukin-6 were measured from venous blood samples using the quantitative sandwich enzyme immunoassay technique (R & D Systems, Minneapolis, MN). C-reactive protein (CRP) was measured using high-sensitivity CRP assay Immulite-1000 (Diagnostic Product, Los Angeles, CA). Diabetic retinopathy was diagnosed in 10 diabetic patients with the Joslin Vision Network video-digital retinal imaging system, which uses a nonmydriatic retinal fundus camera that has been optimized for low–light level imaging without pupil dilatation (13) and has been validated against clinical examination and standard retinal imaging (14).

Subjects with a history of stroke, clinically important cardiac disease, arrhythmia, significant nephropathy, kidney or liver transplant, renal or congestive heart failure, uncontrolled hypertension, carotid artery stenosis >50% by medical history and magnetic resonance angiography, and neurological or other systemic disorders were excluded. Diabetic subjects were treated with insulin (9), oral glucose-control agents (12), or diet (7). In 10 hypertensive diabetic subjects (10), antihypertensive medications were discontinued for 3 days before the study. Medications affecting autonomic function and stimulants were not allowed before the study, but hypoglycemic agents and anticoagulants were allowed. Studies using the 3.0 Tesla MRI were completed in 17 control and 24 diabetic subjects. Subjects with metal implants, pacemakers, arterial stents, and claustrophobia were excluded.

TCD studies were conducted at least 2 h after the last meal according to the following protocol. The subjects rested supine for 10 min with continuous monitoring of cardiovascular, cerebrovascular, and respiratory signals in order to establish a baseline. The subjects hyperventilated to reduce CO₂ to 25 mmHg for 3 min. Then the subjects breathed a mixture of 5% CO₂ and 95% air from a rebreathing bag to increase CO₂ up to 45 mmHg for 3 min, followed by a 5-min rest.

The middle cerebral arteries (MCAs), right (MCAr) and left (MCAl), were insonated from the temporal windows with 2-MHz probes using a TCD system (MultiDop X4; Neuroscan, Sterling, VA). Each probe was positioned to record the maximal BFVs and stabilized using a 3-D positioning system. Assumptions that MCA diameter (15) does not change must be made in order to relate BFV to blood flow. The electrocardiogram was measured from a modified lead II or III using a Spacelab monitor (SpaceLab Medical, Issaquah, WA). Beat-to-beat blood pressure was recorded from a finger with a Finapres device (Ohmeda Monitoring Systems, Englewood, CO) that reliably tracks intraarterial blood pressure when controlled for finger position and temperature (16) and was verified by arterial tonometry. Respiration and end-tidal CO₂ were measured from a mask using an infrared end-tidal volume monitor (Datex Ohmeda, Madison, WI).

Electrocardiogram, blood pressure, BFV, respiration, and CO₂ analog signals were continuously acquired at 500 Hz using Labview 6.0 NIDAQ (National Instruments, Austin, TX). Heartbeat intervals were extracted using a peak detection algorithm, and occasional extrasystoles were removed using linear interpolation. Systolic, diastolic, and mean BFVs were detected from the envelope of the arterial flow waveforms for each heartbeat interval. Beat-to-beat values were averaged over baseline, hyperventilation, and CO₂ rebreathing. A 30-s average was also calculated for the BFV minimum during hyperventilation and maximum during CO₂ rebreathing. CO₂ reactivity was calculated as a slope of the linear regression of mean BFV and CO₂ between hyperventilation and CO₂ rebreathing. Vasodilatation and vasoconstriction reserves were calculated as the percent increment of mean BFV from baseline to the CO₂ rebreathing maximum and to the hyperventilation minimum. Cerebrovascular resistance (CVR) was calculated as the mean blood pressure divided by the mean BFV.

Anatomical images (T1-weighted inversion recovery fast gradient-recalled echo [IR-FGR], fluid-attenuation inversion recovery [FLAIR], and dual T2-weighted fast spin echo) and 3-D magnetic resonance time-of-flight angiography were acquired using a General Electric 3.0 Tesla VHI scanner with quadrature head coil.

Periventricular WMHs present as hyperintense areas with >30% increase in signal intensity on T2-weighted images compared with adjacent white matter. Punctuate lesions are well-defined areas of >2 mm with high signal characteristics (10). FLAIR images were scored using a scale of 0–3 (0, no lesions; 1, focal; 2, beginning confluence; and 3, diffuse involvement of the entire region). Periventricular WMHs and punctuate lesions were graded on all slices in the anterior, middle, and posterior cerebral artery distributions and quantified as a sum, mean, and maximum grade. High-resolution T2-weighted FLAIR images were segmented using the thresholding of hyperintense pixels and a region growing method that allowed an accurate WMH detection without expertise. The brain tissue volumes were computed from the IR-FGR image using the Brain Extraction Tools algorithm (17) based on the expectation-maximization segmentation method (18,19) and was validated using a phantom model (20). The IR-FGR segmented image was registered on the FLAIR image (21) to compute the whole brain and gray and white matter volumes with the same resolution as WMHs and to normalize WMHs for the total brain volume. The graders who processed and scored images were masked to the subject and group assignments.

Descriptive statistics were used to summarize all variables. Physiological measurements were compared between the control and diabetic groups and among conditions (baseline, hyperventilation, and CO₂ rebreathing) using multivariate ANOVA with multiple measure adjustments and Wilk’s λ post hoc tests. One-way ANOVA and Fisher’s exact test were used for nonrepeated variables. The χ² test was used to compare WMH distribution. Stepwise linear and logistic regression models were used to determine the relationships between WMHs and the average resting mean BFVs in both MCAs. The effects of group (control vs. diabetic), BMI, baseline systolic blood pressure, A1C, lipids, inflammation markers, and age and their interactions with other variables were also investigated. The effects of WMHs on CVR and CO₂ reactivity were evaluated using the same approach.

Table 1 compares the demographic, clinical, and biochemical characteristics between the control and diabetic groups. The diabetic group had significantly higher BMI, glucose, A1C, and triglycerides but had lower total, LDL, and HDL cholesterol. Diabetes was not controlled in 18 subjects (A1C 8.1 ± 1.2%). Inflammatory markers such as sICAM-1 (P < 0.05) and WBC (P < 0.003) were elevated in the diabetic group, but soluble vascular adhesion molecule-1, endothelin-1, CRP, and interleukin-6 levels were not different.

As shown in Fig. 1, diabetes had significant effects on mean BFV and CVR in MCAr and MCAl during supine baseline, hyperventilation, and CO₂ rebreathing. Across all conditions, mean BFVs in both MCAs were lower in the diabetic compared with the control group (P < 0.0001; Fig. 1A and B). For each condition, BFVs were lower in the diabetic group during baseline (MCAr and MCAl, respectively: systolic BFV P = 0.006 and 0.05, diastolic BFV P = 0.04 and 0.005, and mean BFV P = 0.005 and 0.01), hyperventilation minimum (MCAr systolic P = 0.02, MCAl diastolic P = 0.02, and mean P = 0.04), and CO₂ rebreathing maximum (MCAr and MCAl, respectively: systolic P = 0.003 and 0.007, diastolic P = 0.01 and 0.002, and mean P = 0.004 and 0.006) (Fig. 1A and B). CVR was significantly higher in the diabetic group in all conditions in both MCAs (Fig. 1C and D). Heart rate (P = 0.03) and systolic blood pressure (P = 0.03) were higher in the diabetic group during baseline. CVR was positively associated with WMHs normalized for brain volume (P < 0.0001) and with A1C and BMI (P < 0.0004) (R² = 0.83, P < 0.0001). Mean BFVs in both MCAs, blood pressure, and CVR were not different between hypertensive and normotensive diabetic subjects in all conditions. There was a trend suggesting that autonomic neuropathy may affect mean BFV values (P = 0.052).

CO₂ reactivity was reduced in the diabetic group (P = 0.05) (Table 1) and was negatively associated with glucose (P = 0.01), diabetic retinopathy (P = 0.02), and normalized WMH volume (R² = 0.54, P = 0.03).

Figure 2 is an example of WMH segmentation on axial FLAIR slices at the level of the ventricles for a control (Fig. 2A–C) and a diabetic (Fig. 2C–E) subject. The distribution of periventricular WMHs differed between the diabetic and control groups (P < 0.0001) and among the frontal, temporal, and parieto-occipital regions (P < 0.0001). Mean WMH grade in the frontal area was greater in the diabetic than in the control group (mean 0.09 ± 0.3 vs. 1.8 ± 0.4, P = 0.01) and was borderline in the parieto-occipital area in the diabetic group (P < 0.07). Periventricular and punctuate WMH volume (5.9 ± 5.5 vs. 6.7 ± 5.4 cm³) and WMH volume normalized for brain volume (0.7 ± 0.7 vs. 0.9 ± 0.7%) and for white matter volume (2.1 ± 1.9 vs. 2.5 ± 2.0%) were not different between the diabetic versus control group, respectively. Hypertensive diabetic subjects had a greater periventricular WMH grade (P = 0.02) and a greater number of subcortical punctuate WMHs (left temporal P = 0.01, basal ganglia P = 0.02) than normotensive diabetic subjects, but WMH volume was not different. Magnetic resonance angiography was normal. Brain volumes and diameters of MCAs and internal carotid artery diameters were comparable.

The total periventricular WMH grade was associated with reduced mean BFV in both groups (P = 0.0002; Fig. 3A). The total WMH volume (P = 0.001; Fig. 3B) and percentage of WMH volume normalized for brain volume (P = 0.0006) were negatively associated with mean BFV. We found excellent correlations between the total WMH grade (sum of all continuous and punctuate WMHs on the visual rating scale) and the WMH volume measured on FLAIR images (P < 0.0001; Fig. 3C), thus validating visual WMH rating using the quantitative volumetric WMH measures.

We assessed an independent relationship between WMH and mean BFV during baseline and determined contributions of type 2 diabetes and other risk factors. Mean baseline BFV was negatively associated with periventricular WMH (P < 0.0001) or normalized WMH volume (P < 0.0001) and uncontrolled diabetes, as indicated by A1C (P = 0.01), WBC (P = 0.05), and sICAM-1 (P = 0.03), and was positively associated with baseline systolic blood pressure (P = 0.004) (whole model R² = 0.86, P < 0.0001). This model was controlled for the effects of age and BMI. CRP was negatively correlated with mean BFV (P = 0.008) but positively associated with age (P = 0.01), BMI (P = 0.003), and WBC (P = 0.01).

Type 2 diabetes exerts complex effects on cerebral microvasculature that may alter cerebral blood flow regulation. We found a decrease of mean BFV and an increase of CVR in type 2 diabetic patients during baseline, hypocapnia, and hypercapnia. Baseline mean BFV was negatively associated with periventricular WMH grade and volume on T2-weighted images and with A1C and inflammation markers. WMHs were also linked with uncontrolled diabetes, elevated CVR, and impaired CO₂ reactivity. The relationship between WMH, uncontrolled diabetes and reduced BFV is of clinical relevance as a potential mechanism for cerebrovascular disease in elderly with type 2 diabetes.

Aging is associated with brain atrophy, changes in frontal subcortical white matter, and executive cognitive dysfunction (9). The CO₂ reactivity diminishes with age, uncontrolled diabetes, and risk factors for atherosclerosis (22). In community-living elderly people, blood flow in the WMHs was lower compared with normal-appearing white matter but flow augmentation to acetazolomide was preserved (23). Correlations among periventricular hyperintensities, demyelination, astrocytic gliosis, and dilatation of perivascular spaces support the theory of arteriosclerosis (8). Diabetic angiopathy is characterized by the vessel wall remodeling, media hypertrophy and increased stiffness (24) that may be enhanced by circulating vasoconstrictors and vascular inflammation. Diabetic subjects had impaired vasodilatation to hypercapnia, but vasoconstriction to hypocapnia was preserved. The MCA diameter was comparable between the groups, similarly to other human (25) and animal (24) studies.

Diabetes alters the glucose and insulin transfer across the blood-brain barrier (26,27), thus affecting regional metabolism and microcirculation (4). Chronic hyperglycemia, which further alters membrane permeability (26,27) and decreases regional blood flow, may lead to permanent cell damage (25). Therefore, diabetes seems to be associated with progressive metabolic disturbance in the cerebrovascular bed that may affect blood flow and accelerate the white matter degeneration. Elevated sICAM-1 and WBC levels in the diabetic group, as well as a negative correlation between BFV with CRP and inflammation markers, support the notion of an active arteriosclerotic process affecting the cerebrovascular bed. An elevation of tumor necrosis factor-α and sICAM-1 accompanied retinal neuronal cell death and blood-brain barrier breakdown induced by oxidative stress in experimental diabetes (28). Tumor necrosis factor-α, which depresses endothelium-dependent vasorelaxation, was increased in diabetic patients with microangiopathy, indicating a relationship between endothelium dysfunction and suppressed production of endothelium-derived nitric oxide (29,30). Furthermore, elevated plasma hemostatic and inflammation markers may reflect insulin resistance and endothelial dysfunction antecedent to diabetes (31,32).

This study addressed an important question about the relationship between WMHs on MRI and BFVs in older adults with type 2 diabetes. It provided further evidence that type 2 diabetes is associated with microvascular disease that may reduce cerebral blood flow in elderly people. Interventions to treat microvascular disease and to enhance cerebral blood flow may play an important role in preventing cerebrovascular complications of diabetes. Future prospective studies are needed to determine whether low cerebral blood flow is a cause or effect of white matter disease.

This study was supported by American Diabetes Association Grant 1-03-CR-23 to V. N., National Institutes of Health (NIH) Older American Independence Center Grant 2P60 AG08812, NIH Program project AG004390, and General Clinical Research Center (GCRC) Grant MO1-RR01032.

We acknowledge the contributions of C. Kantor, BA; R.P. Marquis, BS; A. Tolson, BA; H.-C. Li, PhD; M. Munshi, MD; and the GCRC staff.