This study was performed as part of a larger project of ageing effects on cardiovascular function in normal subjects without a cardiovascular diagnosis, hypertension diabetes, or renal disease requiring dialysis (Cardiac Energetics and Function in Normal Human Ageing, ClinicalTrials.Gov identifier NCT01504828). Exclusion criteria were as above and also known claustrophobia. To get a consistent spread of ages we recruited subjects stratified by age ranges from 20 to 79 years. These and other general data are shown in Table 1. Informed consent was obtained for all patients, and this study was approved by a UK National Health Service Research Ethics Committee (NRES Committee North East - Newcastle & North Tyneside 1, reference number 12/NE/0057). All subjects had measurements of pulse wave velocity by the Vicorder device and Phase Contrast MRI on the same day within 2 h.

The Vicorder measurements were performed by trained research nurses. Patients (non fasting) were laid on an a couch in a quiet clinic room, with the head raised to approximately 30⁰, so that the skin and muscles over the carotid were relaxed though not too tense. Pulse wave velocity was measured by a cuff placed over the right carotid and the right thigh. The length between the carotid and femoral arteries was done by measuring the length between the suprasternal notch and the mid-point of the thigh cuff (Table 2). Measurements were taken until pressure waveforms over the carotid and thigh area were clear and reproducible. At a later date the lead investigator reviewed all tracings and if necessary selected only those data with clear pressure waveform upstrokes. Additional measurements with the Vicorder device were performed with a cuff placed on the right upper arm. These included oscillatory blood pressure measurement and using a global transfer function, central aortic pressures [13] and augmentation and augmentation index. Inter-observer measurements of pulse wave velocity with these methods (N = 32) were 7.8 ± 1.6 and 7.6 ± 1.6 m/s with mean difference of 0.2 ± 0.8 m/s (linear regression r = 0.89), and intra-observer measurements (N = 3) of 6.1 ± 1.0 and 6.1 ± 0.9 m/s with mean difference of 0.03 ± 0.25 m/s.

Phase contrast MRI acquisitions were specifically designed to acquire time-velocity curves at two slices locations (aortic arch and descending aorta approximately 10 cm apart) to estimate transit time (ΔT) as shown in Fig. 1. Subjects were scanned using a Philips Achieva 3 Tesla scanner and a six channel cardiac coil was employed for signal reception. Phase contrast MRI data was acquired by experienced radiographers using a high temporal resolution sequence (repetition time / echo time / flip angle / number of excitations / slice thickness = 5 ms/2.9 ms/10⁰/1/8 mm, SENSE factor 2, field of view 300 mm × 225 mm, reconstructed voxel size 1.17 mm², velocity encoding = 150 m/s, 44 phases, breath hold duration ~19 s) at both slice locations which were the aortic arch (AAo) and the descending (DAo) aorta (Fig. 1). Additional scout images were acquired to facilitate positioning of the phase contrast MR acquisitions and to ensure that the slices were positioned perpendicular to aorta at both locations. Q-flow analysis package (Philips, Viewform) was employed to extract time-velocity curves as shown in Fig. 1 and to estimate precise distance (ΔX) between the two slice locations. The time-velocity curves were then used to determine transit time (ΔT) and compute MR pulse wave velocity = ΔX/ΔT using an in house Matlab based program. A reproducibility analysis of this technique was performed in seven normal volunteers, repeating scans at two sessions, and within each session performing the scanning twice. For session one the values were 4.9 ± 1.6 and 4.5 ± 1.6 m/s (mean difference 0.4 ± 0.9) and session two 5.5 ± 1.6 and 5.2 ± 1.2 m/s (mean difference 0.3 ± 0.7).

Both techniques were significantly and similarly related to age (Table 2 and Fig. 2, Vicorder R = 0.594, and MRI R = 0.572), with R² values in both accounting for between 35 and 33 % respectively of the variances in the model (both P < 0.0001). Vicorder and MRI were also significantly correlated (Fig. 3a, R = 0.271, 7 % of variance, P < 0.05). Bland Altman plot (Fig. 3b) showed that in general the mean difference between the 2 measures was 1.6 m/s greater for Vicorder. Also, particularly at higher measurements of pulse wave velocity, there were four individual cases that were greater than two standard deviations greater or less than the mean difference.

Hemodynamic data are presented by age group in Table 3. Vicorder pulse wave velocity was significantly related to all these hemodynamic variables. In general, the extent of significant correlations with these measurements were less with the MRI (Table 4), though there were highly significant correlations with aortic pulse pressure, augmentation and augmentation index. Multivariate analysis was performed separately for both techniques to compare how they related to other physiological variables (Tables 5 and 6). For Vicorder, age alone was significantly related to pulse wave velocity. For MRI age and heart rate were significantly related. In both cases the slope (Beta) against age was similar (Vicorder 0.603 and MRI 0.632).

The Vicorder and MRI are fundamentally two different techniques. The Vicorder measurements rely on pressure waveforms whereas MRI is based on flow. The other major difference is the region from which the pulse wave velocity is measured. The MRI is measured in a relatively small segment (10 cm) in the descending aorta, whereas the Vicorder measurements go from the carotid to femoral artery. This can explain why the Vicorder measurements are on average 1.6 m/s greater than the MRI as the peripheral arteries are stiffer than the aorta [14, 15]. The measurement of length with MRI is accurate, though with the Vicorder an anatomical external surface measure along the length of the major blood vessels is used, which is a potential source of error. Despite these differences it is reassuring that both are closely and similarly related to age.

Vicorder pulse wave velocity has been assessed comprehensively in healthy subjects by Müller and colleagues [16]. In comparison with the present study, they obtained mean values of pulse wave velocity for 20–29 year olds of 5.8 ± 0.7 m/s and for 40–49 years of 7.1 ± 0.9 m/s which is slightly lower than our data particularly at the younger age (6.7 ± 0.9 and 7.5 ± 1.2 m/s respectively, Table 2). A likely explanation for this difference is in the measurement of the carotid-femoral length. Müller et al. have used the length from the suprasternal notch to the top of the femoral cuff, whereas we have taken the measurement to the middle of the cuff. Thus, our longer length will record higher pulse wave velocities. We have done this as the centre of the cuff has been shown to be where the pressure measurement is actually recorded [10]. Rogers et al [7] have looked at phase contrast MRI to measure pulse wave velocity in different regions of the aorta. Their values obtained from the proximal descending aorta which is a similar area as used in our study are comparable to our study, though values particularly in the older age groups are lower. However, this study was in a smaller cohort of patients and significant variability is seen in the older subjects. For instance, in the 20–29 year old Rogers et al. obtained a pulse wave velocity of approximately 5 m/s (compared to 4.5 ± 1.5 m/s in the present study), and for 60–69 years old approximately 9 m/s compared to 6.8 ± 2.1 m/s in the present study. In a larger study of 162 normal subjects Hickson and colleagues [17] have compared the SphygmaCor and Vicorder devices and phase contrast MRI. Their values of phase contrast MRI pulse wave velocity in the descending thoracic aorta are very similar to the current study. They did however show a closer relationship of Vicorder pulse wave velocity measurements to phase contrast MRI measurements than in the current study (r = 0.64 vs r = 0.27 in current study). This higher level of correlation may be due to the larger number of subjects studied and the measurement of pulse wave velocity along the whole length of the descending aorta by phase contrast MRI. Nevertheless Bland Altman plots show a similar pattern to the current study with higher overall mean values for Vicorder, and individual points outside two standard deviations particularly at higher levels of pulse wave velocity. Thus, the values of both Vicorder and MRI pulse wave velocity obtained in this study match very closely to published data.

There are two issues of particular clinical significance with these data. Firstly, measurement of pulse wave velocity is recommended in the 2013 European Society of Cardiology/European Society of Hypertension guidelines for assessment of cardiovascular risk [18], though there are continuing issues regarding clinical application that have been recently highlighted [19]. The issue that is most relevant to this study is that there are a myriad of devices that can measure pulse wave velocity and that these are not necessarily interchangeable. In that regard, our data is very important comparing these two techniques. In particular, as MR imaging becomes more widely adapted in clinical cardiovascular diagnostics, how this performs compared to other techniques is very important. The second issue of clinical significance is that with both techniques there were approximately 10 % of cases in which data was not obtainable or usable. This was more prevalent in older subjects. One of the advantages with the Vicorder system is that it is easy to use, and so in this study all examinations were performed by research nursing staff. However, despite this there are cases where the carotid waveform is simply not clear enough to identify the upstroke and therefore not usable, and this appears higher than previously reported [11]. For the MRI scanning there were similarly issues with both claustrophobia and tolerating the scan, and also poor data quality which is predominantly in the older subjects. The scans were performed by experienced radiographers, though there were issues with obtaining scans perpendicular to the descending aorta in older subjects as a result of increased aortic curvature. As these techniques move from a purely research domain to clinical tools, these data are important to understand how these techniques perform. Our data suggests that assessing pulse wave velocity in older subjects requires careful attention to these technical issues to avoid loss of potentially important data, and more information is required to determine how other techniques perform in these older subjects.

These data are in normal subjects, so in subjects with cardiovascular disease with increased vascular stiffness the relationships described may be altered, and also there may be additional issues with acquiring high quality data. Gender differences have been described with measures of vascular stiffness [20], but this study was not statistically powered to detect gender differences. There are other methods to measure pulse wave velocity with MRI. For instance, other areas of the aorta can be assessed using phase contrast MRI. Hickson and colleagues [17] have shown that the abdominal aorta has slightly higher rates of age-related increases in pulse wave velocity relative to the descending thoracic aorta. Other phase contrast techniques include determination of flow-area [21], and cross-correlation methods [22] (where flow is determined at several points along the descending aorta). Ibrahim et al. [23] have shown that the transit time (as used in this study) and cross-correlation methods resulted in more reproducible measurements compared to the flow-area method. Our MRI reproducibility studies have more variation than the data published by Ibrahim et al. explained by their much larger study number. A dependency of heart rate on augmentation index has been reported for the SphygmoCor device [24]. As there is no published data on this issue with the Vicorder device, we have not normalised data to heart rate.