Arteriosclerosis is the process of arterial stiffening that has significant pathophysiological implications and is strongly associated with cardiovascular events [1]. It is predominantly a product of age and pulse pressure, reflecting the effects of repetitive strain on the elastic fibers of the arterial wall [2]. Pulse wave velocity is a well established marker of arterial stiffening with important prognostic implications, and as a result of such has now been incorporated into management guidelines for the treatment of hypertension [3]. Carotid femoral pulse wave velocity (cf-PWV) is the most commonly used technique for the measurement of PWV, however it relies on the use of surface measures of distance to represent the convoluted path of the underlying arterial tree, introducing potential errors in PWV calculation.

Use of Magnetic Resonance Imaging (MRI) for the measurement of PWV (MRI-PWV) has become increasingly prominent due to its ability to measure directly the central aortic PWV coupled to recent advances in the technique allowing high temporal resolution imaging and reduced image acquisition time [4]. The two techniques correlate well with each other, but with a significant and consistent bias between the two techniques across studies with cf.-PWV consistently being approximately 1.6–1.7 ms⁻¹ higher than MRI-PWV [5, 6]. This is consistent with observations of 1.9 ms⁻¹ difference between the aortic pulse wave velocity measured invasively and the cf-PWV, however this previous study also used a direct carotid-femoral measurement which is known to overestimate distance travelled and thus PWV by approximately 25% [7]. Previous work has demonstrated both a clear discrepancy in the distances measured using a tape measure and the actual arterial pathlength travelled by the pulse wave due to the effects of age and obesity confounding both arterial length and body contour measurements [8]. However the muscular and elastic properties of the aorta and its branch vessels is also known to change throughout their length which in turn affects their stiffness and by extension PWV [6, 9].

It is thus not currently clear whether the differences in PWV between the two techniques are due to inaccuracies in distance measurement in cf-PWV, the lower temporal resolution of MRI-PWV or the changing elastic properties throughout the arterial tree. The effects of this bias are also not clear. The aim of the current study was therefore to evaluate whether the differences between cf.-PWV and MRI-PWV in a mixed cohort of patients with and without diabetes mellitus and symptomatic cardiovascular disease can be accounted for by inaccuracies of currently used distance measurements and whether recalculation of cf.-PWV using accurate distance measurements results in improved detection of cardiovascular disease.

One hundred fourteen study participants completed the study protocol undergoing cf.-PWV, whole body MRA and cardiac MRI. Of these 90 also underwent MRI PWV measurement using phase contrast MRI of whom 88 had analysable data, and thus used in the final analysis. Demographics and CMR measurements of the 4 groups are described in Table 1.

PWV distance measurements were significantly lower using MRA arterial centrelines than on external tape measure (tape measure distance = 537 ± 48 mm vs. MRA distance = 452 ± 33 mm, mean diff = 85.4 ± 54.0 mm, p < 0.001). This is predominantly due to an underestimation of the carotid arteries to the arch distance measurement, and an over-estimation of the arch to femoral artery measurement (See Table 2). As a result, cf.-PWV_EXT was significantly higher than the recalculated cf.-PWV_MRA (cf-PWV_EXT = 10.98 ± 2.63 ms⁻¹ vs. cf.-PWV_MRA = 9.2 ± 2.04 ms⁻¹, p < 0.001). Despite this the carotid femoral PWV calculated using the tape measure and using the MRA centerline showed a high degree of correlation (rho = 0.90, p < 0.001). A backward stepwise linear regression model using the difference between external tape measure differences and arterial centerline distances as the dependent variable and age, sex, height, weight, waist circumference, and hip circumference as independent variables was performed to better understand the source of the differences between the two measurement techniques. Age (B = −0.75 ± 0.45, p = 0.1), height (B = −187 ± 50.3, p < 0.001), waist (B = 1.29 ± 0.5, p = 0.011), and hip circumference (B = 2.17 ± 0.62, p = 0.001) were all significantly related to distance measurement discrepancy and combined accounted for 42.9% of the difference between the techniques.

MRI-PWV showed moderate correlation with cf.-PWV_EXT (rho = 0.49, p < 0.001) but showed a consistent bias with a significantly lower MRI-PWV than cf.-PWV_EXT (MRI-PWV = 8.12 ± 2.9 ms⁻¹ vs. cf.-PWV = 10.92 ± 2.68 ms⁻¹, p < 0.001). When cf.-PWV was recalculated using the inter-arterial distance from whole body MRA, correlation between the two techniques improved (R = 0.55, p < 0.001), and a significant reduction in the difference between the two techniques (MRI-PWV = 8.12 ± 2.92 ms⁻¹ vs. cf.-PWV_MRA 9.08 ± 2.06 ms⁻¹. Despite this improved correlation, and a significant reduction in the bias and a reduction in the inter-technique variability, as demonstrated in the Bland Altman plots in Fig. 2, the differences between the techniques remained significantly different (mean difference = 0.96 (95%CI 0.36–1.55 ms⁻¹). Recalculation of the PWV using the MRA measurements improved the correlation between PWV and age (cf-PWV_EXT rho = 0.5, cf.-PWV_MRA rho = 0.65, p < 0.001 for both) and pulse pressure (cf-PWV_EXT rho = 0.5, cf.-PWV_MRA rho = 0.52, p < 0.001 for both), see Table 3 for a full list of correlates.

Neither cf.-PWV_EXT nor cf.-PWV_MRA differentiated between the four groups although the MRI PWV showed a trend towards differentiation between the groups (see Table 4). When the groups were combined into having CVD or no-CVD, a continued lack of difference between these groups was observed for both cf.-PWV_EXT (p = 0.68) and cf.-PWV_MRA (p = 0.39), however a significant difference was seen in MRI-PWV (CVD + ve MRI-PWV = 9.22 ± 3.37 vs CVD-‘ve MRI-PWV 7.22 ± 2.12 ms⁻¹, p = 0.01).

In the current study we have shown a significant difference between the external distance measurement used for the calculation of PWV and the actual arterial pathlength. When this distance discrepancy is corrected for using intra-arterial distances measured on MRI there is a significant improvement in the agreement between PWV measured using tonometry or MRI.

The continued difference of approximately 1 ms⁻¹ between the two techniques after accounting for this difference in distance measurements suggests part of the differences in the PWV is due to differences in the underlying arterial stiffness between the underlying vessels. Previous work has shown the PWV to increase throughout the aorta, although even in the distal aorta where the PWV was the highest, this was still 1 ms⁻¹ lower than cfPWV [6]. Thus differences in the two measurements are almost certainly secondary to differences in carotid and iliac stiffness. Of these the iliac stiffness is the most likely to contribute to this difference as the carotid artery demonstrates a similar distensibility to the abdominal aorta, while the femoral arteries are significantly stiffer than both [14, 15]. The distorting effect of the iliac stiffness on this measurement has clinically significant implications, as even the corrected cfPWV did not differentiate between those with and without cardiovascular disease, whereas the aortic stiffness as measured on MRI was able to detect a significant difference in the PWV between these.

Our difference between cf.-PWV and MRI is consistent with prior studies, although is even more pronounced at 2.8 ms⁻¹ than previous reports of 1.8 ms⁻¹ [5, 6]. The most likely explanatory factor for this is the much higher BMI of the recruited participants in the current study compared with these two previous studies. Our findings of an 85 mm discordance between the external measurement technique and the actual arterial pathlength taken by the pulse wave is in agreement with prior work by Huybrechts et al. who found a difference in distance measurement of 81.8 mm between an arterial centreline and external tape measure distance [8]. This previous study found a significant confounding effect from age on pathlength measurement with only 4% of the discrepancy being accounted for by BMI. This is in contradistinction to our study where weight and hip circumference accounted for 42% of the measurement differences with only 1% of differences explained by age. These differences in findings are likely due to the much narrower age range combined with the more obese nature of our study participants compared to the wide age range and relatively normal weight in the previous study, as obesity is known to result in increased tape measure distances [16, 17]. However as those who are most likely to be assessed by PWV for risk assessment are likely to be older and overweight, rather than young and slim, our findings are likely to be more clinically relevant, and add weight to the proposal that direct caliper measured distance between the carotid artery and femoral artery should be the measurement technique of choice as these are least affected by body contours [7]. Creating a consistent technique for the accurate calculation of travelled pulse wave distance is crucial to the interpretation of these studies, and also to allow comparison between different techniques. We have shown that differences in PWV between MRI and tonometric measures can be largely accounted for by inaccuracies in these distance measurement techniques. Thus creation of suitable techniques may allow not just accurate PWV measurements but improved cross centre reproducibility and the ability to combine data from multiple techniques in systematic reviews and meta-analysis adding greater power to these. It may be possible to generate these measurements automatically from simple patient allometric measures such as sex, age, and height, and indeed several groups have tried exactly this [18, 19]. However one of these used the tape measure as the gold standard while the other used the aortic root to aortic bifurcation measurement, neither of which reflect the true path of the pulse wave as measured by carotid-femoral PWV. One option may be to use large whole body imaging studies such as the TASCFORCE and SHIP studies that would allow for the generation of an accurate aortic pathlength model based on thousands of individuals [20‐22]. While this would require substantial effort and external validation, the benefits may be justified in terms of improved inter-centre and inter-individual reproducibility due to the removal of an extra step in the process of PWV and its incumbent potential for human error.

In our study we observed no significant difference between the groups using either cf.-PWV or MRI PWV. This is likely due to the relatively small sample size of our study, as using the same carotid-femoral technique in a larger population, of which the current group is a sub-study of, significant differences were observed between all groups [11]. Our lack of correlation between PWV and left ventricular mass is interesting as it has previously been posited that the importance of arterial stiffening is due to its effects on the afterload of the ventricle resulting in an increased work load with resultant adverse remodelling [23]. Previous studies demonstrating a correlation between the two have been performed in healthy volunteers or those with cardiovascular risk factors but no overt clinical cardiovascular disease [24, 25]. Thus PWV may have a greater impact on left ventricular mass earlier in the arteriosclerotic disease process with later left ventricular remodelling being affected by other processes such as coronary atherosclerosis, however longitudinal studies will be required to further elucidate this relationship.

There are several limitations within our current study. We used a three point technique for measuring the PWV distance externally, while current ESC guidelines would advocate the use of a direct carotid-femoral distance measurement with a correction factor of 0.8 to account for the shorter internal pathlength of the pulse wave and the simplicity of this technique [3]. However this is largely based on a single MRI study of 98 individuals, and other studies have suggested the currently used technique provides superior correlation with invasive aortic PWV over a direct distance measurement technique [18]. Our cf.-PWV and MRI scans were performed on different days, which is known to affect PWV measurement due to differing haemodynamic states [26], and may account for the lower correlation between the two techniques than has been seen in previous studies, but this does not change the fact that recalculation of PWV using an accurate distance measurement improves agreement between techniques. Finally, the MRI PWV and cf.-PWV use different techniques for determining wave arrival time which is known to affect the measurement and reproducibility, [27] however each technique was optimised to maximise its own intrinsic accuracy and reproducibility.

In conclusion, differences in PWV measurement between carotid-femoral PWV and MRI measured central aortic PWV can be predominantly explained by inaccuracies introduced by the use of simple surface measurements to represent the convoluted arterial path between the carotid and femoral arteries. Correction for this may in future allow more direct comparison between the techniques thus strengthening systematic reviews and meta-analyses.