Blood Pressure and Vascular Dysfunction Underlie Elevated Cerebral Blood Flow in Systemic Lupus Erythematosus

Subjects

This study was approved by the University of New Mexico Institutional Review Board and conformed with the ethical standards of the Declaration of Helsinki. All subjects signed an informed consent document. The diagnosis of SLE was established in each subject using the American College of Rheumatology criteria^18,19. A cohort of 42 SLE subjects (40 women, mean age 37.3 ± 12.8 years) and 19 age- and sex-matched healthy controls (17 women, mean age 31.9 ± 10.1 years) were studied. Twenty-six SLE subjects (62%) were positive for antiphospholipid antibody, 10 (24%) were positive for antiribisomal P antibodies, and 26 (65%) were positive for anti-dsDNA antibodies. Eighteen SLE subjects (43%) were diagnosed with active NPSLE using the neurological SLE Disease Activity Index (Neuro-SLEDAI) criteria²⁰ as follows: seizures (2 patients), psychosis (3 patients), organic brain syndrome (5 patients), visual disturbance (1 patient), cranial neuropathy (2 patients), lupus headache (7 patients), and stroke (4 patients). Patients with past but not active NPSLE were excluded from the study. The 2 groups were similar in all demographic variables (Table 1). More extensive clinical data and findings on this cohort can be found in Table 2 and a previous report¹¹.

Clinical and laboratory evaluations

Patients and controls underwent clinical and laboratory evaluations including specific measures of inflammation, coagulation, and fibrinolysis. Group differences in several of these tests are reported in Table 1. As part of the clinical evaluation, systolic (SBP) and diastolic (DBP) blood pressure measurements were obtained in the supine position from the brachial artery using an automatic sphygmomanometer. The mean arterial pressure (MAP) was derived as 2/3 DBP + 1/3 SBP. Mean disease duration, activity, severity, therapy, and serology of the patients with SLE, including antiphospholipid antibodies, are reported in Tables 1 and 2 and in our previous report¹¹.

Magnetic resonance imaging

Anatomical MRI, DSC-MRI, and MR angiography to assess brain injury, cerebral perfusion, and cerebral atherosclerosis were performed on all subjects within 7 days of clinical assessment using a 1.5-Tesla Siemens Sonata scanner with an 8-channel head coil. The total examination time of the MRI protocol was roughly 1 h. Scout images were used to prescribe a series of whole-head T1-weighted, T2-weighted, and fluid-attenuated inversion recovery (FLAIR) images aligned with the interhemispheric midline and parallel to the anterior commissure-posterior commissure line. T1-weighted images [3-D fast low-angle shot sequence, TR/TE = 12 ms/4.76 ms, flip angle 20°, field of view (FOV) 256 mm × 256 mm, resolution 1 mm × 1 mm, 128 slices, slice thickness 1.5 mm] and T2-weighted images (turbo spin-echo sequence, TR/TE = 9040 ms/64 ms, turbo factor 5, FOV 220 mm × 220 mm, resolution 1.1 mm × 1.1 mm, 128 slices, slice thickness 1.5 mm) were acquired for anatomical segmentation of normal-appearing gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). The FLAIR image (variable flip, TR/TE/TI = 6000 ms/358 ms/2100 ms, averages = 2, slice thickness 1.5 mm, FOV 220 mm × 220 mm, matrix size 192 × 192) was used to manually segment lesions. The segmented GM, WM, CSF, and lesion maps were used to identify these regions in the processed perfusion maps. The DSC-MRI images^21,22 were acquired with a perfusion-weighted echo planar imaging sequence (TR/TE = 1430 ms/46 ms, flip 90°, 20 slices, time course of 50 sequential acquisitions, FOV 200 mm × 200 mm, matrix size 128 × 128). Magnevist^® contrast agent (gadopentetate dimeglumine) was injected into an antecubital vein at the standard dose of 0.1 ml/kg body weight using a power injector at 5 ml/s, starting 15 s after the start of scan acquisition, followed by 20 ml saline at the same rate. Raw DSC-MRI data were analyzed using the Penguin software (NordicImagingLab, Bergen, Norway) and the CBF maps constructed with Penguin were coregistered to the T1-weighted images to identify distinct brain regions within GM and WM as described¹¹. Briefly, pixel intensity changes due to contrast (gadopentetate) passage over the time series of images are converted into pixel-by-pixel tissue contrast concentration curves, C(t). With user supervision, the software selects candidate arterial pixels to construct an arterial input function, AIF(t), based on the higher intensities and narrow time courses of the concentration changes. We attempted to locate and use pixels associated with the middle cerebral artery consistently for the construction of the AIF(t). A gamma variate function is fit to the raw C(t) and AIF(t) curves to capture just the first pass of the contrast bolus and used to represent C(t) and AIF(t) in further analyses. Calculation of CBF involves an equation relating C(t) to CBF and the convolution of AIF(t) and the residue function R(t), the fraction of contrast agent in the vasculature at any time: C(t) = CBF × [AIF(t) ⊗ R(t)], where ⊗ denotes the convolution operation. CBF maps constructed with Penguin were coregistered to the T1-weighted images to identify distinct brain regions within GM and WM. This procedure involved automatically segmenting the T1-weighted image into GM, WM, and CSF and coregistering the lesion map that was manually traced from the FLAIR image to the T1-weighted image. The T1-weighted image was warped into a standard brain template for identifying the major cortical and subcortical brain regions. Mean GM and WM CBF values for 4 cerebral regions (occipital, parietal, temporal, and frontal) and 4 subcortical regions (caudate, putamen, thalamus, and globus pallidus) for each hemisphere were determined. In order to resolve CBF in normal-appearing tissue from CBF in lesions, the mean regional cerebral or subcortical values of CBF were calculated without the lesion pixels included. This permitted comparison of normal-appearing tissue between groups (those with and those without lesions) without the confound of the lower lesion CBF values reducing the overall mean values for the region.

Statistical analysis

The primary goal of the statistical analysis was to identify factors most associated with the presence of elevated CBF in patients with SLE. Since a large number of clinical and laboratory variables were considered, the statistical strategy incorporated 3 levels of analysis with progressively increasing complexity and focus. At the first level, group differences in the various variables obtained for both controls and SLE subjects were examined with the Student t test and correlations between regional CBF and these measures were explored with univariate Pearson correlation analyses. In order for a clinical or laboratory variable to be a candidate for further analysis, the Pearson correlation with CBF was required to be significant (p ≤ 0.05) in at least 1 of the 16 GM or WM brain regions in which CBF was measured. To discount significant correlations that may have depended on outliers, a Spearman correlation analysis was also performed. Hence, both Pearson and Spearman correlations needed to be significant in order for a variable to be considered in further analyses. Using these criteria, 6 out of 36 variables collected from both SLE and control subjects qualified for further analyses (Table 3). The mean values and significance of group differences in the variables are given in Table 1.

In the second stage of analysis, each of these candidate factors was added as a covariate in a repeated measures (RM) analysis of covariance (ANCOVA) model. The use of RM ANCOVA follows the approach taken in our initial report on differences in CBF and CBV only, and is motivated by the fact that CBF varies substantially from region to region in the brain¹¹. This regional variability would obscure small group differences in CBF in conventional t tests. In the current RM ANCOVA model, 8 brain regions were entered as a repeated factor (combined normal-appearing GM and WM CBF in occipital, parietal, temporal, frontal, caudate, putamen, thalamus, and globus pallidus regions), the 2 groups (SLE, controls) as a grouping factor, and the presence of lesions as a covariate. The brain regions used in computation of CBF excluded the lesions, but the presence of lesions still was included in the model as an important confounder. If the effects of group (SLE vs control) were rendered nonsignificant by addition of the covariate, then the factor would be considered a potential mediator between CBF and group. A common interpretation of such an effect is that the covariate “explains” (“intervenes” between) the group effect on the dependent variable. In each analysis, if the group-by-covariate interaction term were significant, then an interaction plot of CBF versus the covariate by group was examined to determine how the relationship between CBF and covariate differed in SLE and control groups. In the interaction plot for our model, we used the least-square means of the CBF values for each group while controlling for the lesion variable.

In the final stage of analysis, both univariate and stepwise multivariate regression models (forward/backward selection with p value ≤ 0.05 for variables to enter and to stay in the model) were used to directly explore the relationships between explanatory covariates identified in the various RM ANCOVA models and various physiologic measures in the study.