Methods for the ARIC MRI study have been described previously.10, 14
Relevant details are described here. The study sample consisted of members of the Atherosclerosis Risk in Communities (ARIC) study cohort who participated in the ARIC Carotid MRI substudy in 2004–2005 (Year 18). The ARIC study is a population-based cohort study of cardiovascular disease incidence among African American and Caucasian adults (n=15,792)15
The ARIC MRI sampling strategy included approximately 1,200 participants with carotid artery wall thickness (maximum over 6 sites: left and right of the common, bifurcation, internal) that was at least >68th percentile as measured by carotid B-mode ultrasound on ARIC study visits 3 (1993–1995) and 4 (1996–1998). Intima-media thickness (IMT) cut-offs were 1.35, 1.00, 1.28, and 1.22 mm in Forsyth County, Jackson, Minneapolis, and Washington County, respectively, representing the 73rd, 69th, 73rd, and 68th percentiles of maximal IMT. A cohort random sample of approximately 800 participants whose carotid intima-media thickness was <68th percentile was also included. Ineligibility criteria for the carotid MRI substudy included standard contra-indications to the MRI exam or to the contrast agent, carotid revascularization on either side for the low CIMT group or on the selected side for imaging for the high CIMT group, and difficulties in understanding questions or in completing the informed consent.
Protocols for fasting glucose, blood pressure, height and weight measurements were identical at the baseline ARIC cohort examination and 18 years later at the Carotid MRI substudy examination. The study was approved by the institutional review committees of all participating centers, and all participants provided written informed consent.
Traditional cholesterol measurements
Twelve-hour fasting plasma lipid assays were performed on participants of the ARIC MRI substudy. Plasma samples were collected on ice using EDTA as the anticoagulant at the time of the MRI visit. Total cholesterol, triglycerides, and HDL-C were measured enzymatically and expressed in mg/dL units. LDL-C levels were calculated using the Friedewald equation.16
Non-HDL-C levels were calculated as total cholesterol minus HDL-C. Blind duplicate coefficient of variation for total cholesterol, HDL-C, and triglycerides were 2.0%, 3.0%, and 2.7%, respectively.
Apolipoproteins (ApoB and Apo A-1) were measured in frozen plasma using a commercially available immunoturbidimetric assay (OLYMPUS®, Olympus America Inc). The inter-assay coefficient of variation for Apo A-I and ApoB were 3.9% and 7.2%, respectively. Intra-assay coefficient of variation for Apo A-1 and ApoB were 1.5% and 1.7%, respectively. Reliability coefficients based on 120 blinded split QC samples were 0.93 and 0.95 for ApoA-I and ApoB, respectively.
Lipoprotein particle analysis
Frozen EDTA plasma samples were thawed, and a 200 μL aliquot withdrawn, refrozen, and shipped on dry ice to Liposcience, Inc. (Raleigh, NC) for NMR lipoprotein particle analysis. Using this technique, particle concentrations of lipoproteins of different sizes were calculated from the measured amplitudes of their spectroscopically distinct lipid-methyl group.17
The NMR variables examined in this manuscript were total LDL particle concentration (LDLp) expressed in nmol/L units and total HDL particle concentration (HDLp) expressed in μmol/L units.
The methods for acquisition and interpretation of the MRI images collected in the Carotid MRI substudy have been described previously.10, 14
Briefly, a contrast-enhanced MRI exam of the thickest 1.6-cm segment of the thicker carotid artery was performed according to a standard protocol on a 1.5T whole-body scanner as follows: A 3-dimensional time-of-flight magnetic resonance angiogram (MRA) was acquired through both carotid bifurcations. Detailed black blood MRI (BBMRI) images were then acquired through the extracranial carotid bifurcation, known to have a thicker maximum wall by the most recent ultrasound study, unless the contralateral carotid bifurcation wall appeared to the technologist to be thicker on the MRA. These BBMRI images consisted of 16 axial T1-weighted, fat-suppressed slices (thickness=2 mm; acquired in-plane resolution=0.51 × 0.58 mm2
; total longitudinal coverage=3.2 cm) oriented perpendicular to the vessel and centered at the thickest part of the internal or common carotid artery wall. These 16 slices were acquired 5 minutes after the intravenous injection of gadodiamide (Omniscan, GE Amersham), 0.1 mmol/kg body weight, with a power injector.
Seven readers were trained to interpret the MRI images and contour the wall components using specialized software (VesselMass, Leiden University Medical Center). Readers were blinded to the characteristics of the study population. Each reader drew contours to delineate the lumen, outer wall, lipid core, and calcification. Eight slices centered at the slice with the thickest wall were analyzed. All exams were assigned quality scores by the reader based on image quality and protocol adherence; exams that failed were not analyzed.
Using semi-automated software, vessel walls were divided into 12 radial segments, and mean thickness values were generated for each segment (). Standard deviation of the wall thickness measurement was also computed and provides a measure of the plaque eccentricity. Area measurements were calculated for the lipid core and calcification contours. Volumetric data were computed by integrating area measurements over all 8 slices examined.
Figure 1 Black blood MRI (BBMRI) slices through the carotid bifurcation and plaque. A long axis BBMRI image adjacent to the slice shown in Figure 1a was used to orient 8 precontrast (yellow lines) and 16 postcontrast (yellow and blue lines) slices through the (more ...)
Maximum segmental wall thickness (in mm) was defined as the maximum wall thickness of 12 segments at the slice with the largest lipid contour area. Vessel wall area and lumen area (in cm2) were computed at the slice with the largest mean segmental wall thickness. Volume measurements for total wall volume (in mm3) was computed by integrating area measurements over 8 slices. Normalized wall index (NWI) was then calculated. NWI was derived as carotid wall area divided by the total vessel area. The total vessel area in these measurements includes the lumen area. Lipid-rich core was represented as a binary variable (present or absent).
All analyses were weighted and appropriately accounted for the stratified random sampling design of the ARIC Carotid MRI substudy. Analyses were conducted using SAS version 9.1 for descriptive statistics or SUDAAN for domain analysis. Wall thickness and wall volume were analyzed in the full data set. Due to the resolution constraints of the MRI scan, we restricted consideration of lipid core to those 1,131 participants whose maximum wall thickness was ≥1.5 mm. Only 4 lipid cores were excluded using this cut point. Among the 1,131 participants who were included in the analyses, a total of 542 participants had an identifiable lipid-rich core.
For continuous MRI variables, standardized regression coefficients (β) are presented for linear regression models, standardizing by one standard deviation (SD) of exposure and outcome with adjustment for covariates. These β coefficients can be interpreted as number of SD difference in the dependent variable (e.g., total wall volume) associated with a single SD difference in plasma cholesterol or lipoprotein level or their ratio. For the dichotomous lipid-rich core variable (presence or absence), standardized odds ratio (OR) are presented for logistic regression models adjusting for covariates. These ORs can be interpreted as the odds for the presence of lipid-rich core associated with a single SD difference in plasma cholesterol or lipoprotein level or their ratio. Covariates included for adjustment were age, race, gender, smoking status, body mass index, blood glucose, hypertensive medication use, cholesterol-lowering medication use, aspirin use, diabetic medication use, hs-CRP, and plasma triglyceride concentration. Similarly, the coefficients of determination (R2) were described for each model including the above-mentioned covariates as well as lipid or lipoprotein particle measure of interest. R2 can be interpreted as proportion of the variance in the dependent variable (e.g., total wall volume) as explained by the set of independent variables in the model including the cholesterol or lipoprotein particle concentration parameter of interest.
Because the presence of lipid-rich core on carotid MRI has been shown to highly correlate with carotid wall thickness,10
we conducted additional analyses to examine the presence of lipid-rich core, while controlling for carotid wall thickness in addition to the covariates mentioned above.