Functional versus morphological assessment of vascular age in patients with coronary heart disease

Communicating cardiovascular risk based on individual vascular age (VA) is a well acknowledged concept in patient education and disease prevention. VA may be derived functionally, e.g. by measurement of pulse wave velocity (PWV), or morphologically, e.g. by assessment of carotid intima-media thickness (cIMT). The purpose of this study was to investigate whether both approaches produce similar results. Within the context of the German subset of the EUROASPIRE IV survey, 501 patients with coronary heart disease underwent (a) oscillometric PWV measurement at the aortic, carotid-femoral and brachial-ankle site (PWVao, PWVcf, PWVba) and derivation of the aortic augmentation index (AIao); (b) bilateral cIMT assessment by high-resolution ultrasound at three sites (common, bulb, internal). Respective VA was calculated using published equations. According to VA derived from PWV, most patients exhibited values below chronological age indicating a counterintuitive healthier-than-anticipated vascular status: for VAPWVao in 68% of patients; for VAAIao in 52% of patients. By contrast, VA derived from cIMT delivered opposite results: e.g. according to VAtotal-cIMT accelerated vascular aging in 75% of patients. To strengthen the concept of VA, further efforts are needed to better standardise the current approaches to estimate VA and, thereby, to improve comparability and clinical utility.

Communicating cardiovascular risk to the patients remains challenging for doctors. Successful uptake of advice and sustained adherence to recommendations, including lifestyle and medication, may heavily depend on appropriate illustration and communication of such risk to the individual patient 1 . Cardiovascular risk prediction models (e.g., Framingham or SCORE) allow an estimation of absolute risk. However, these models have limitations: the widely used Framingham-based risk scores seem to underestimate coronary heart disease risk in older adults, while SCORE algorithms exclusively predict fatal cardiovascular events 2 . Nevertheless, current guidelines recommend their use for total cardiovascular risk estimation for adults > 40 years of age unless they are categorised as being at high-risk or very-high risk 3 .
The concept of vascular age is considered a helpful alternative to communicate risk of individual patients 4 . It relates the chronological age to the vascular status of an individual and is easy to understand. Its use has been suggested as a powerful tool aiding primary and secondary prevention strategies 5 . There are various approaches to calculate vascular age, relying on functional or morphological surrogates of atherosclerosis. From a conceptual point of view, vascular age is the result of the exposure over time to unfavorable levels of cardiovascular risk factors leading to accelerated development of atherosclerosis. Hence, structural measurements may be expected PWV measurement and PWV-derived vascular age. Measurements of PWV (PWVba, PWVcf, PWVao) and the corresponding AIao were performed in 501 patients by trained personnel following predefined standard operating procedures using the Vascular Explorer (Enverdis, Jena, Germany). Patients were told to refrain from alcohol and hot drinks for 12 h prior to the examination. Patients rested for 5 min supine in a temperature-controlled room. During the examination, patients were asked not to move, speak, or cough. For the measurement of the PWVao it was additionally required to approximate the length of the aorta by measuring the distance (in cm) between pubic bone and suprasternal notch. In some subjects, unrealistically small values for the distance between notch and pubic bone (i.e., < 40 cm) were documented without a definite cause; these subjects were excluded from subsequent analyses (n = 29).
Prior to the determination of the stiffness parameters, systolic and diastolic blood pressure was measured at the left and right upper arm and lower leg. The PWV measurement was done with blood pressure cuffs inflated at the right upper arm (upper arm cuff, UAC) and the right lower leg (lower leg cuff, LLC) to oscillometrically detect pulsatile pressure changes in the brachial and anterior tibial artery. The examination was divided into two parts. First, PWVba was determined by inflating both cuffs at the predetermined diastolic pressure level for 15 s. Afterwards, PWVao and AIao were obtained by applying a cuff pressure of 35 mmHg above the systolic blood pressure at the upper arm. Then, the Vascular Explorer software analysed all recorded pressure waves of each cycle and calculated the mean measured values. The pressure wave who´s measured values came closest to the mean values was chosen for quantitative analysis. In case of motion artefacts or inappropriate signal quality, the examiner was requested to repeat the measurement. If automated detection and selection algorithms of the Vascular Explorer software failed to yield optimal results, the selection could be manually overridden by the examiner.
For the calculation of PWVba, the Vascular Explorer software a) imputes further distances (i.e., distance suprasternal notch to centre of the UAC or LLC, respectively) derived from patient´s height, and b) utilises the pulse transit time (i.e., the derived travel time of the pulse wave between two selected reference points, Eq. (1); from: manual of the vascular explorer). The approach for the measurement of PWVao has been validated against invasive measurements of the PWVao, and a correlation coefficient of 0.91 was observed 9 .
Pressure waves originating at the aortic outflow tract do not only travel into the peripheral arterial tree, but are also reflected at branching points or in areas with changes in impedance or diameter, e.g. the aortic bifurcation (projected at the pubic bone) 20 . According to the model applied by the Vascular Explorer for PWV measurement the second pressure wave in the brachial artery pressure contour represents the wave reflection at the aortic bifurcation. The additional distance travelled by this reflected wave compared to the initial pressure wave in the brachial artery thus represents an approximation of two times the length of the aorta. By measuring the time difference between the arrival of both pressure waves in the brachial artery (return time, RT), PWVao can be determined non-invasively (Eq. (2)). Jug = Jugulum (= suprasternal notch), Sym = Symphysis (= pubic bone), RT = return time. The PWVcf is derived from the PWVao and PWVba by applying an unpublished formula (company secret). Transfer functions that are also spent from the Vascular Explorer software then provide central pressure waves allowing the calculation of AIao via peripheral pressure waves 11 .
Finally, the PWVao, the PWVba and the AIao can be used to calculate a vascular age (PWV-VA) depending on respective PWV and AI measurements: VA PWVao , VA PWVba , and VA AIao . The vascular age was based on nomograms provided by McEniery et al. 13 . Each measured value was assigned to an age, at which it corresponds to the 50 th percentile in the named cohort (see Supplementary Figure S1). Vascular age could reach a minimum of 20 and a maximum of 100 years.
Vascular ultrasound and cIMT-derived vascular age. Measurement of cIMT was performed in 501 subjects using high-resolution ultrasound on a Vivid Q system equipped with a 10 MHz linear transducer (GE, Fairfield, USA). Examiners were certified according to a protocol provided by an external supervisor (Meijer Medical Ultrasound), and iteratively completed quality checks every two months. The cIMT assessment included bilateral measurements of the far walls of the common carotid artery (CCA), the carotid bulb, and the internal carotid artery (ICA), thus resulting in 6 values per patient. Ultrasound images were frozen at the end-diastolic phase (R wave peak of concomitantly run electrocardiogram) and had to display at least 10 mm of vessel wall length. If image quality was insufficient (right bulb: n = 35; left bulb: n = 47; right ICA: n = 53; left ICA: n = 61), the CCA of that particular side was examined more extensively, i.e. from 3 different angles. All images were stored in DICOM format and imported into a dedicated workstation for cIMT quantification (Syngo US Workplace, Siemens, Munich, Germany).
Images were semi-automatically analysed using the Syngo Arterial Health Package 21 . If needed, the reader could edit the intima-media layer delineations suggested by the software. Subsequently, the crude measurement values per site were stored digitally. Besides crude values also the average of all 6 means is reported (total-cIMT).
Calculation of cIMT-derived vascular age was based on measurements of the right and left CCA (VA rCCA and VA lCCA ) and the total-cIMT (VA total-cIMT ) and calculated using nomograms provided by the ARIC investigators 17 considering age, sex, and ethnicity, respectively. Each measured value was assigned to an age at which it corresponds to the 50 th percentile in the ARIC cohort (see Supplementary Figure S2). According to the recommendations of the ARIC investigators, vascular age could range between 20 and 79 years; values higher than 79 years were beyond the nomogram-verified range and therefore labelled " > 80 years" by the Syngo software. Moreover, the software allowed to calculate vascular age only for individuals aged between 40 and 70 years for whom 6 measured values were available. As 229 subjects did not meet these criteria (163 men and 42 women outside the age range, 19 men and 5 women with incomplete cIMT data), morphological vascular age could be calculated in 272 patients.

Definitions.
Variables used for the statistical analysis were defined according to current guidelines. Hypertension was defined as systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg 22 . Patients with a fasting glucose ≥ 7.0 mmol/l or 2-h plasma glucose ≥ 11.1 mmol/l were defined as diabetics. Individuals exhibiting a fasting glucose < 7.0 mmol/l and 2-h plasma glucose < 11.1 mmol/l but ≥ 7.8 mmol/l were labelled as "impaired glucose tolerance" (IGT) 23 . Impaired fasting glucose (IFG) was defined as fasting glucose ≥ 6.1 mmol/l but < 7.0 mmol/l with 2-h plasma glucose < 7.8 mmol/l 23 . Dyslipidaemia was defined as LDL cholesterol > 1.8 mmol/l 24 and nutritional status was classified according to WHO 25 . Central obesity was defined as waist circumference ≥ 88 cm for women and ≥ 102 cm for men 26 . Data analysis. Patient characteristics are presented using absolute and relative frequencies, mean (SD), and median (quartiles), as appropriate and were compared across categories using χ2 test. Due to sex differences in stiffness parameters, results of PWA and cIMT were compared between male and female patients. We used linear regression for PWVao, PWVcf, AIbr, AIao and AIao75 and censored regression analysis for VA rCCA , VA lCCA and VA total-cIMT , including age adjustment. In cases, where parameters (PWVba, VA PWVao , VA PWVba and VA AIao , rCCA-cIMT, lCCA-cIMT, total-cIMT and mean of the maximum cIMT) did not fulfil requirements for linear

Results
The characteristics of the study population (n = 536) stratified by sex are summarised in Table 1. The majority of both men (75.7%) and women (71.6%) were between 60-79 years of age. Sex differences were scarce with men showing a higher percentage of current and former smokers, while central obesity was more frequent in women.
In total, 501 subjects of the study population underwent the vascular explorer examination and the cIMT measurement. Reasons for dropouts for both crude values and derived vascular ages are explained in Methods (cf. 2.2, 0). www.nature.com/scientificreports/ Median blood pressure measurements performed with the Vascular Explorer showed no significant differences between men (n = 416) and women (n = 85): right arm MAP in men was 104 mmHg (quartiles 94-114 mmHg) vs 104 mmHg (94-115 mmHg) in women (age-adjusted P = 0.459); respective measurements on left arm were 105 mmHg (101-115 mmHg) in men vs 113 mmHg (101-122 mmHg) in women (age-adjusted P = 0.169). No sex differences were found for PWVao, PWVba, PWVcf, mean of the maximum cIMT, PWA-VA and cIMT-VA using regression analyses including age as covariate.
Linear regression analysis for PWVcf revealed considerable associations for mean arterial pressure, age, smoking and duration of coronary heart disease (see Table 3). For total-cIMT, linear regression analysis revealed considerable associations only for age and female sex (see Table 3).
Correlation between vascular age derived from PWA vs cIMT was poor. Spearman coefficients ranged from rho = 0.06 for VA PWVba with VA rCCA to rho = 0.16 for VA AIao and VA lCCA (see Supplementary Table S1). Further analyses revealed only moderate correlation between the vascular age within type of assessment (i.e., within PWV-VA and within cIMT-VA). The highest correlation was found between VA PWVao and VA AIao (rho = 0.57; see Supplementary Table S2 and Supplementary Table S3).
To visualise accelerated vascular aging (i.e., vascular age exceeding chronological age), we created bar charts showing the distribution of the difference of vascular and chronological age (see Fig. 1). Since analyses were performed with "81 years" being the maximum VA of the cIMT examination, only patients aged < 61 years could possibly enter the group of individuals exhibiting pre-aging of > 20 years. In general, vascular age derived from PWA was similar or even smaller than patients´ age, the latter being the case in 67%, 60% and 48% of men, and 78%, 66% and 50% of women for VA PWVao , VA PWVba , and VA AIao , respectively. By contrast, regarding VA rCCA , VA lCCA and VA total-cIMT , 28%, 33% and 34% of men and 24%, 33% and 38% of women belonged to the group of maximum VA, with 65%, 63% and 75% of men and 67%, 69% and 79% of women exhibiting a pre-aged vascular status. Table 2. Results of the PWA and the cIMT measurement. PWA pulse wave analysis, cIMT carotid intimamedia thickness, PWVao aortic pulse wave velocity, PWVba brachial-ankle pulse wave velocity, PWVcf carotidfemoral pulse wave velocity, AIbr brachial augmentation index, AIao aortic augmentation index, AIao75 aortic augmentation index standardised to a heart rate of 75 bpm, VA Vascular age, rCCA right common carotid artery, lCCA left common carotid artery. 1 linear regression for PWVao, PWVcf, AIbr, AIao and AIao75, censored regression for VA rCCA , VA lCCA and VA total-cIMT and quantile regression for PWVba, VA PWVao , VA PWVba , VA AIao , rCCA-cIMT, lCCA-cIMT, total-cIMT and mean of the maximum cIMT; age-adjusted. 2 maximum value over a length of 10 mm. 3 mean value over a length of 10 mm.   Figure 1. Percentage distribution of VA rCCA minus age (a)), VA lCCA minus age (b)), VA total-cIMT minus age (c)), VA PWVao minus age (d)), VA PWVba minus age (e)) and VA AIXao minus age (f)). Visualisation of accelerated vascular aging (i.e., vascular age exceeding chronological age) by the distribution of the difference of vascular and chronological age. A negative value represents vascular pre-aging. Since analyses were performed with "81 years" being the maximum VA of the cIMT examination, only patients aged < 61 years could possibly enter the group of individuals exhibiting pre-aging of > 20 years. In general, vascular age derived from PWA (first row) was similar or even smaller than patients´ age while VA derived from cIMT (second row) delivered opposite results. VA vascular age, PWVao aortic pulse wave velocity, PWVba branchial-ankle pulse wave velocity, AIao aortic augmentation index, rCCA right common carotid artery, lCCA left common carotid artery, total-cIMT total carotid intima-media thickness.

Discussion
Despite their meaning in today's cardiovascular risk prediction, cIMT measurement and the PWA showed largely different results in the determination of the individual vascular age as a correlate of vascular stiffening. On the one hand, while being easier to perform and more economic, PWA conducted with the Vascular Explorer seems to underestimate the cardiovascular risk systematically. On the other hand, cIMT is more reliable in detecting patients with high risk for cardiovascular disease but only half of our study population fulfilled the requirements for the vascular age determination via normal values from the ARIC cohort. We investigated Spearman´s rank correlations between the vascular ages derived from the cIMT measurement and the PWA. In general, correlations were poor between cIMT and PWA, which is in line with studies investigating crude values of both examinations 28 . Amongst measurements obtained with one method, we found moderate correlations (except for VA PWVba that showed poor correlation with both VA PWVao and VA AIao ). As underlying problem for these inconsistencies, we assume a methodological issue. The statistically determined distance travelled by the pulse wave and used for the calculation of the PWVba seems to overestimate the real distance. By contrast, the distance used for the calculation of the PWVao was measured by the examiner. Poor correlation may also be attributed to the fact that, compared to PWVba, measurement of PWVao and PWVcf with the Vascular Explorer is based on a model, in which the second systolic pressure wave detected at the brachial artery is solely the result of pulse wave reflection at the aortic bifurcation. While there is evidence of good correlation of this approach with invasive PWV measurement 9 , data from Trachet et al. suggest that the second wave is rather a result of brachial artery stiffness than aortic stiffness 29 .
To compare the vascular status according to the cIMT and the PWA, we determined the difference between vascular age and chronological age. Although the study participants by definition belonged to a group considered at high risk for cardiovascular diseases, PWA found accelerated vascular aging in less than 50% of our patients. These unexpected findings may be attributed to the reference values used by the Vascular Explorer software for the VA determination. In the reference study an alternative device, the SphygmoCor (AtCor Medical, Sydney, Australia), had been used to measure the AI in 4 001 healthy individuals aged 18 to 90 years, and the carotidfemoral (but not aortic) PWV in a subset of 998 subjects 13 . In contrast to the Vascular Explorer, this device analyses carotid and femoral artery wave forms which are sequentially recorded by gated electrocardiogram 30 . So far, the association between PWV/AI measured with the Vascular Explorer and the SphygmoCor device was determined in a small study cohort (n = 44) 31 . A strong correlation for AI (r = 0.89, P < 0.0001) and a moderate correlation for PWV (r = 0.57, P = 0.0002) between the measured values of both devices was reported. Furthermore, regarding the PWV, Bland-Altman-Analysis showed a systemic bias by the Vascular Explorer with overestimating low PWV values and underestimating high PWV values. With most of our patients being situated in the latter PWV range, the reference values obtained with a different device may be an explanation for the unexpected results of the VA determination. The need for device-specific normal values is emphasised by moderate or poor correlation between PWV measured with the Vascular Explorer and the Arteriograph (TensioMed, Budapest, Hungary) or the VICORDER (Skidmore Medical, UK) 31,32 . In a comparable study cohort of medicated patients with coronary heart disease (n = 160, mean age = 61.5 years; 71.7% men) examined with the Arteriograph, mean PWVao was 12.24 m/s compared to 8.16 m/s in our study 33 . Differences in that order of magnitude raise doubts regarding the reliability of these PWV measurements and hence their utility for the diagnosis of an end organ damage of the arteries 34 .
For the determination of our patients´ individual VA via cIMT, both the mean of the left and right CCA and the mean of all means were used. There is evidence that one single site of measurement may be sufficient to assess individual cardiovascular risk 35 . Best reproducibility was shown for the mean cIMT compared to the maximum and minimum value 21 . In contrast to PWA, when comparing the patients´ vascular age to the chronological age, we found accelerated vascular aging in most of our patients -as we had expected. However, determination of vascular age via cIMT was limited to patients aged 40-70 years (according to specifications of pre-sets of the software package) with 6 measured values. As a result, cIMT-VA was available for about half of our study sample. Furthermore, reference values for determination if cIMT-VA have been derived in the early 1990ies in the US by the ARIC investigators. Hence, the comparison of the technical equipment as the characteristics of the population under study is compromised. Alternatively, in the community-based Gutenberg-Heart Study performed in Germany between 2007 and 2008, normal CCA-cIMT values were established in a subgroup of 1,025 subjects without classic cardiovascular risk factors or previous cardiovascular disease 36 . These normal values allow the VA CCA determination in a slightly larger age range (35-74 years of age). The insecurity about normal values for the cIMT is illustrated by the fact that the former threshold of 0.9 mm for the diagnosis of a target organ damage is being questioned in the latest guidelines of the ESC/ESH 22 .
We acknowledge that our study has certain limitations. Our analyses involved data based on a German study population (98.5% Caucasian) and might not be transferable to other ethnicities. We examined medicated CHD patients. Consequently, the results of the vascular examination may have been influenced by drug therapy, but we did not detect major effects of medication on actual measurements. The length of the aorta for PWA was measured along the body surface, resulting in falsely higher values in case of abdominal obesity. Using a measuring loop and having the examiner palpate the pubic bone may have resulted in higher precision in the length determination. In our analyses, however, we did not find confirmation of a systematic deviation caused by obesity. Finally, cold exposure may enhance the augmentation of wave reflection thus overestimating AI 37 . We strived to maintain a stable peripheral temperature thus avoiding cooling effects when measuring stiffness parameters.
In conclusion, vascular age if measured by cIMT but not if measured by PWA showed the accelerated aging expected in the investigated high-risk sample. The findings of the PWA performed with the Vascular Explorer may be explained, amongst other things, by the usage of inappropriately derived normal values. Device-specific normal values are needed to strengthen the reliability of the vascular age determination via PWA, which is

Data availability
The datasets generated and analysed during the current study are not publicly available due to data protection but are available from the corresponding author on reasonable request. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.