Participants
The study was a single centre observational sub-study of the multicentre SUMMIT (SUrrogate markers for Micro- and Macrovascular hard endpoints for Innovative diabetes Tools) study which was designed to analyse cardiovascular biomarkers in diabetes. Recruitment criteria, strategy and study protocol have been described in detail previously [10, 11], but in summary, subjects were recruited and categorised into 4 groups based on their history of type 2 diabetes and cardiovascular disease (CVD) as follows: Group 1: Type 2 diabetes mellitus (T2DM) with a prior clinical diagnosis of cardiovascular disease that included coronary artery disease (CAD), cerebrovascular disease and/or lower extremity arterial disease (LEAD); Group 2: Type 2 diabetes mellitus with no clinical evidence of cardiovascular disease; Group 3: Absence of diabetes mellitus with clinical evidence of CAD, cerebrovascular disease and/or LEAD; Group 4: Healthy controls, with no biochemical evidence of diabetes mellitus (see below) and no clinical evidence of cardiovascular disease. All participants had a detailed clinical history and examination performed, bloods taken for renal function, cholesterol and HbA1c, and underwent whole body magnetic resonance angiography (MRA), cardiac magnetic resonance imaging (CMR) and cf.-PWV.
Carotid-femoral pulse wave velocity
Carotid-femoral PWV was measured using a SphygmoCor device (Atcor Medical, West Ryde, Australia). A blood pressure (BP) cuff was attached to the left arm, and three electrocardiogram (ECG) electrodes (lead I) were attached. External distance measurements were performed using a tape measure, and a proximal (carotid to sternal notch) and distal (sternal notch to umbilicus and umbilicus to femoral) measure with final distance used for calculation being the proximal distance subtracted from the distal distance. After resting for 5 min, the BP was measured three times at one-minute intervals, and the mean value of the two final measurements was recorded. The carotid and femoral pulses were captured using the SphygmoCor device with cf.-PWV automatically calculated as the measured distance divided by the differences in time between the R wave of the ECG and the foot of the carotid and femoral pulse curves. For the remainder of the paper, cf.-PWV calculated using the external tape measure distance shall be referred to as cf.-PWVEXT.
Magnetic resonance imaging
Images were acquired on a 32 RF receiver channel, 3 Tesla MRI scanner (Magnetom Trio, Siemens, Erlangen, Germany). For whole body coverage, a combination of six RF coils including head matrix, neck matrix, spine matrix, two body matrix and peripheral angiography phased array RF surface coils were used. Subjects where placed supine, head first into the magnet bore. The imaging protocol was carried out in 3 phases: i) MRA of the thoracic and neck, and distal lower limbs, ii) CMR including late gadolinium enhancement (LGE) and iii) MRA of the abdomen, pelvis and proximal lower limb.
Whole body magnetic resonance angiography protocol
Whole body magnetic resonance angiography (WB-MRA) was performed using 4 overlapping data sets covering: head, neck and thorax (station 1), abdomen and pelvis (station 2), upper legs (stations 3) and lower legs (station 4). All stations were acquired using a coronal spoiled FLASH (Fast Low Angle Shot) sequence (repetition time TR = 2.6–3.47 ms; echo time TE = 0.96–1.21 ms; flip angle = 16–37°; pixel area = 1.1–1.5 mm2 and slice thickness = 1–1.4 mm, slight variation according to station and participant body habitus). [12].
First, station 1 and 4 pre-contrast images were acquired. Following this an injection of 10 ml of 0.5 M Gadoterate meglumine (Guerbet, Villipinte, France) followed by a 20 ml saline flush were administered at a rate of 1.5 ml/s. Acquisition of station 1 was triggered when the contrast reached the aortic arch (timed using a fluoro sequence through the aortic arch), following which three sequential acquisitions of station 4 were performed. Cardiac MRI was performed after this first injection (see next section for details) before stations 2 and 3 image acquisition began. After the cardiac acquisition was completed, pre contrast images were acquired of both stations following which post-contrast images were acquired after an injection of 15 ml gadoterate meglumine and 20 ml saline flush, both administered at a rate of 1.5 ml/s.
The 3D WB-MRA datasets were viewed offline (Carestream PACS Client Suite Version 10.1 sp1, Rochester, NY, USA). An arterial centreline was drawn between the bifurcation of the right common carotid artery and the right common femoral artery. From this a curved multiplanar reformat of the vessel was generated. From this the distance from the carotid bifurcation to the centre of the aortic arch was measured (proximal distance) (See Fig. 1). The distance from the common carotid to the common femoral artery was measured, from which the aortic arch to the common femoral artery bifurcation distance (distal distance) was calculated as the carotid to arch distance subtracted from the total distance, with these representing the measures taken using the external tape measure technique. The distances were measured twice and an average of the two used for subsequent analysis. This distance was then used to recalculate the cf.-PWV using MRA measured distance (distal – proximal distance) divided by the carotid-femoral time interval produced by the SphygmoCor device. For the remainder of the article, the cf.-PWV calculated using the MRA derived arterial distance shall be referred to as cf.-PWVMRA.
Cardiac magnetic resonance (CMR) protocols
Cardiac magnetic resonance (CMR) imaging utilised a spine matrix and six-element body array matrix RF coils. TurboFLASH two-chamber, four-chamber and short axis localiser, and two-chamber and four-chamber cine images were acquired. Left ventricular assessment involved the acquisition of a horizontal long axis, vertical long axis and stacked short axis cine images using a retrospectively gated TrueFISP sequence (repetition time TR = 3.4 ms; echo time TE = 1.48 ms; flip angle = 50–60°; pixel area = 1.4 mm by 1.9 mm; slice thickness = 6 mm; inter-slice gap = 4 mm). Left ventricular mass (LVM), end diastolic volume (EDV), end systolic volume (ESV), stroke volume (SV) and ejection fraction (EF) were calculated from the short axis stack.
MRI pulse wave velocity measurement
A retrospective ECG-gated gradient-echo pulse sequence with velocity encoding was applied to measure the through plane flow at two predefined locations in the ascending and abdominal aorta. The first slice was positioned axial through the aortic arch at the level of the pulmonary bifurcation, and the second slice was placed axial through the descending aorta immediately proximal to the renal arteries.
Imaging parameters included the following: echo time of 4.83 ms, repetition time of 14 ms, flip angle 20°, slice thickness of 8 mm, field of view at 350 mm, matrix size 256 × 256. The temporal resolution was optimised to ensure that 128 phases per cardiac cycle were obtained and a VENC of 150 cm/s. To determine the distance between the two aortic slices, a 2D gradient echo FLASH (fast low angle shot) was acquired of the aorta in a ‘candy stick’ double-oblique orientation. TR/TE 40/1.2 ms; flip angle 15°, slice thickness of 8 mm, 23 cardiac phases, 1 averages, a pixel size of 1.5 × 1.5 mm2, bandwidth of 475 Hz/ pix and breath hold scan time of average 9 s. Aortic PWV was calculated from the MRI images using the transit time method [4]. Image analysis was performed by a single observer using Segment version 1.9 R4339 (http://segment.heiberg.se) [13], blinded to the clinical status of the subjects. The up-slope of the arriving pulse wave at each location was calculated from the flow curves. The distance was measured along the aorta between the two analyses planes using candy stick FLASH and the time delay calculated as the time delay between the arrival of the foot of the pulse wave at the ascending aorta and abdominal aorta.
Statistical methods
Descriptive statistics were used for the analysis of the demographic and clinical features of the cohorts with data expressed as mean ± standard deviation (SD) for continuously distributed variables, and n(%) for nominal data.. Normality of distribution was tested using Shapiro Wilk test; if the test failed, where possible standard transformations such as square root, reciprocal or logarithmic transforms were used to generate a Gaussian distribution. To test the null hypothesis that samples originate from the same population, a paired sample t-test was used for comparison of distance measurements and between technique PWV measurements. Spearman correlation coefficient and Bland-Altman plots were performed to further examine the differences between carotid-femoral and MRI PWV. One-way analysis of variance (ANOVA) and Kruskal–Wallis ANOVA by ranks was used for the non-parametric data to compare demographic data and PWV measurements between the four groups. Backward stepwise linear regression was used to determine the contribution of body metrics and contours to discordances between tape measure and MRI measured arterial pathlengths. All data were analysed using SPSS statistical package (version 21.0, SPSS Inc. Chicago, Illinois). Significance was assumed when p < 0.05.