*Article* **The Correlation of Arterial Stiffness Parameters with Aging and Comorbidity Burden**

**Francesco Fantin 1,\*, Anna Giani 1, Monica Trentin 1, Andrea P. Rossi 1, Elena Zoico 1, Gloria Mazzali 1, Rocco Micciolo <sup>2</sup> and Mauro Zamboni <sup>3</sup>**


**Abstract:** The aim of the study was to evaluate the relationships between carotid-femoral pulse wave velocity (PVW-cf), cardio-ankle vascular index (CAVI) and CAVI0 (which is a mathematical elaboration of CAVI, theoretically less dependent on blood pressure), age and comorbidity burden. Furthermore, 183 patients (119 female, mean age 67.5 ± 14.3 years) referred to the Geriatric Ward and Outpatient Clinic at Verona University Hospital were included; demographic, clinical and blood analysis data were collected. Charlson Comorbidity Index (CCI), PVW-cf, CAVI and CAVI 0 were obtained. Significant correlations were found between CAVI, CAVI0, PVW-cf and both age (r = 0.698, r = 0.717, r = 0.410, respectively *p* < 0.001 for all) and CCI, (r = 0.654; r = 0.658; r = 0.448 respectively and *p* < 0.001 for all), still significant after adjustment for several variables. In a stepwise multiple regression model, considering several variables, CCI was the only predictor of PWV-cf, whereas age and CCI were significant predictors of both CAVI and CAVI 0. In conclusion, all arterial stiffness indexes are associated with CCI and aging; the latter correlation is more evident for CAVI and CAVI 0 than for PVW-cf. Arterial stiffness parameters can complement the characterization of patients affected by a remarkable comorbidity burden across aging; arterial stiffening might mirror the complexity of these individuals.

**Keywords:** CAVI; CAVI0; PWV; comorbidity; aging

#### **1. Introduction**

Vascular aging is associated with arterial wall remodeling, with progressive stiffening and reduced compliance; arterial stiffness is an independent predictor of cardiovascular morbidity and mortality [1]. It is therefore of remarkable importance to evaluate arterial stiffness in older individuals and in those adult patients who, owing to the presence of vascular and metabolic comorbidities, display high cardiovascular risk. Carotid-femoral pulse wave velocity (PVW-cf) and cardio-ankle vascular index (CAVI) are two common and feasible techniques aimed at detecting signs of vascular stiffening. As compared to PWV-cf, CAVI can evaluate arterial stiffness from a larger proportion of the arterial tree and is considered less dependent on blood pressure at the time of measurement [2,3]. Thus, although pulse wave analysis is considered the gold standard technique to evaluate vascular stiffness [4,5], CAVI can provide a more comprehensive assessment of arterial stiffness [3]. Furthermore, in order to further relieve the dependence of CAVI by blood pressure, in 2016 the mathematical expression of CAVI was elaborated and CAVI 0 was then suggested [6–8], and the association between CAVI and CAVI 0 has been widely demonstrated [9,10].

A massive number of pathological conditions have been shown to be related to increased arterial stiffening. For instance, PVW-cf is known to be associated to aging [11],

**Citation:** Fantin, F.; Giani, A.; Trentin, M.; Rossi, A.P.; Zoico, E.; Mazzali, G.; Micciolo, R.; Zamboni, M. The Correlation of Arterial Stiffness Parameters with Aging and Comorbidity Burden. *J. Clin. Med.* **2022**, *11*, 5761. https://doi.org/ 10.3390/jcm11195761

Academic Editors: Ignatios Ikonomidis, Andrea Grillo and Paolo Salvi

Received: 22 July 2022 Accepted: 20 September 2022 Published: 28 September 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

cardiovascular risk factors [1] and metabolic syndrome [12]. On the other hand, increased CAVI is described in older subjects [13,14], in hypertensive patients [15,16], in the presence of vascular calcification and inflammation [17], in diabetic individuals and with concomitant metabolic diseases, [18,19], and in the presence of dyslipidemia [20,21]. Furthermore, weight loss is associated with CAVI reduction [22], and a positive association is described between CAVI and the presence of epicardial and visceral adipose tissue [23]. Increased CAVI is a predictor of cardiovascular events [24] and it is also associated to coronary artery disease [13], cerebral ischemia [25] and carotid arteries plaques [26].

Interestingly, although several comorbidities, as considered per se, are shown to be related to increased stiffness, less is known about the possible effect of the comprehensive comorbidity burden. The role of arterial stiffening in the characterization of complex patients with relevant comorbidity burden, considered across aging, is yet to be deeply explored; however, it may shed light on riveting pathophysiological issues. Furthermore, particular attention should be paid to cardiovascular comorbidities and risk factors, given their direct involvement in arterial structures. The aim of the study was to examine the correlation between arterial stiffness indexes, comorbidities and cardiovascular risk factors in a group of adults and older adults.

#### **2. Materials and Methods**

The study population included 183 subjects, 119 females and 64 males, hospitalized at Geriatric Clinic of Verona University Hospital or referred to Outpatient Clinic (medical nutrition or arterial hypertension, of any age). Exclusion criteria were: (I) limb amputation or history of surgical treatment to aorta, carotids, or femoral arteries; (II) severe peripheral arterial disease or proximal arterial stenosis; (III) atrial fibrillation or other major arrhythmias. A detailed clinical history, with particular mention to cardiovascular diseases and risk factors, and physical examination were recorded for each patient. To evaluate the comorbidity burden, Charlson comorbidity Index (CCI) was calculated for each patient, using anamnestic patient-reported data.

The study was approved by the Ethical Committee of the University of Verona. All participants gave informed consent to be involved in the research study.

#### *2.1. Anthropometric Variables*

With the subject barefoot and wearing light indoor clothing, body weight was measured to the nearest 0.1 kg (Salus scale, Milan, Italy), and height to the nearest 0.5 cm using a stadiometer (Salus stadiometer, Milan, Italy); whenever patients could not assume the erect position, the last anamnestic height was recoded. BMI was calculated as body weight adjusted by stature (kg/m2).

#### *2.2. Blood Pressure and Arterial Stiffness Measurements*

CAVI, blood pressure and heart rate were measured and mean arterial pressure (MAP) and pulse pressure (PP) were calculated using VaSera-1000 (Fukuda-Denshi Company, LTD, Tokyo, Japan), as per the manufacturer's recommendations. BP cuffs were placed simultaneously on the four limbs and inflated two by two (right and left side) to increase the accuracy of measurements. ECG was obtained by two electrodes placed on both arms; to obtain phonocardiography, a microphone was placed on the sternum (second rib space). This device calculates CAVI, on the basis of the Bramwell–Hill Formula [27,28], measuring heart-ankle PWV by the following equation:

$$\text{CAVI} = a \ast \left( \ln \frac{P\_{\text{s}}}{P\_{\text{d}}} \ast \frac{\text{PWV}^2 \ast 2\rho}{P\_{\text{s}} - P\_{\text{d}}} \right) + b \ast$$

where *a* and *b* are constants, *ρ* is considered the blood density, *Ps* stands for systolic blood pressure (SBP), and *Pd* stands for diastolic blood pressure (DBP). By means of this device, heart-ankle PWV (haPWV) was calculated as the ratio between aortic valve to ankle length

(automatically derived by software) and the time T taken by pulse wave to run this distance (T = tb + tba, tb = time from the second heart sound to the dicrotic notch at the brachial pulse wave form, tba = time from brachial to ankle pulse waves) [29]. CAVI 0 was derived by proper electronic calculator [30] following the formula:

$$\text{CAVI}\,0^\circ = \frac{\text{CAVI} - b}{a} \ast \frac{\frac{P\_b}{P\_d} - 1}{\ln\left(\frac{P\_s}{P\_d}\right)} - \ln\left(\frac{P\_s}{P\_{ref}}\right)^2$$

and considering Pref as a standard pressure of 100 mmHg.

#### *2.3. Pulse Wave Velocity*

The pulse wave analysis was performed noninvasively using a portable device called PulsePen (Diatecne, Milan, Italy) [31], and its software to obtain central aortic pressure values, an assessment of arterial pulse wave contours, an estimation of reflection waves and measurements of PWV. We previously provided a detailed description of PWA calculation, in previous studies [12,32]; we obtained carotid-femoral PWV (PWV-cf), which is considered representative of elastic arteries [33]. As recommended by consensus documents [34], the carotid-femoral distance was multiplied by a correction factor of 0.8.

#### Biochemical Analysis

Venous blood samples were obtained after the subjects fasted overnight. Plasma glucose was measured with a glucose analyzer (Beckman Instruments Inc, Palo Alto, CA, USA). Cholesterol and triacylglycerol concentrations were determined with an automated enzymatic method (Autoanalyzer; Technicon, Tarrytown, NY, USA). High-density-lipoprotein (HDL) cholesterol was measured by using the method of Warnick and Albers. LDL cholesterol was calculated using the Friedwald formula [35]. Creatinine was measured by a modular analyzer (Roche Cobas 8000; Monza, Italy); eGFR was calculated by Cockroft– Gault formula.

#### *2.4. Statistical Analyses*

Results are shown as mean value ± standard deviation (SD). Variables not normally distributed were log-transformed before analysis. Pearson correlation coefficient was used to estimate correlations between variables. Independent samples t-tests were used to compare baseline characteristics of female and male patients. Analysis of variance (ANOVA) was performed when comparing continuous data, after stratifying the population upon age classes and comorbidities and to evaluate the effect of independent variables included in regression models.

A significance threshold level of 0.05 was used throughout the study. All statistical analyses were performed using SPSS 23.0 version for Windows (IBM, Armonk, NY, USA).

#### **3. Results**

The study population included 183 individuals, mean age 67.5 ± 14.3 years, 65% (*n* = 119) female. The main characteristics of the study population are listed in Table 1.


**Table 1.** Main characteristics of the study population.

**Table 1.** *Cont.*


BMI: body mass index, HDL: high density lipoprotein; LDL: low density lipoprotein; GFR: glomerular filtration rate; SBP: systolic blood pressure; DBP: diastolic blood pressure; PP: pulse pressure; MAP: mean arterial pressure; CAVI: cardio-ankle vascular index; PWV-cf: pulse wave velocity carotid-femoral; CCI: Charlson Comorbidity Index.

#### *3.1. Univariate Analysis*

As shown by univariate analysis (Table 2), all arterial stiffness indexes display a positive relationship with age (CAVI r = 0.698, CAVI0r= 0.717, PVW-cf r 0.410, *p* < 0.001 for all of them). Furthermore, CAVI, CAVI 0 and PVW-cf resulted correlated to higher comorbidities, as measured by CCI (CAVI r = 0.654, *p* < 0.001; CAVI 0 r = 0.658, *p* < 0.001; PWV r = 0.448 and *p* < 0.001). Both CAVI and CAVI 0 showed a significant inverse relation with DBP (r = −0.296 and r = −0.389, respectively, *p* < 0.001 for both), MAP (r = −0.209, *p* = 0.005 and r = −0.274, *p* < 0.001, respectively) and a positive relation with PP (r = 0.165, *p* = 0.025 and r = 0.219, *p* = 0.003, respectively).

**Table 2.** Univariate Correlations between CAVI, CAVI0, PWV-cf and the main clinical variables.


\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001. HDL: high density lipoprotein; LDL: low density lipoprotein; GFR: glomerular filtration rate; CCI: Charlson Comorbidity Index; SBP: systolic blood pressure; DBP: diastolic blood pressure; MAP: mean arterial pressure; PP: pulse pressure; CAVI: cardio-ankle vascular index; PWV-cf: pulse wave velocity carotid-femoral.

Moreover, CAVI 0 is directly correlated to CAVI (r = 0.955, *p* < 0.001) and both CAVI and CAVI 0 relate to PVW-cf (r = 0.430 and r = 0.438 respectively, *p* < 0.001 for both).

#### *3.2. Subgroup Analysis: Cardiovascular Comorbidities and Risk Factors*

As outlined by subgroup analyses, patients with hypertension diagnosis, as compared to patients without, had increased arterial stiffness indexes (mean PWV-cf 10.05 ± 4.67 vs. 8.63 ± 3.36, *p* = 0.017; mean CAVI 9.25 ± 2.13 vs. 8.19 ± 1.85, *p* = 0.001; mean CAVI 0 15.82 ± 6.57 vs. 12.95 ± 4.58, *p* = 0.003). Mean CAVI, CAVI 0 and PVW-cf were also increased in diabetic patients, when compared to normoglycemic subjects (mean PWV-cf 12.53 ± 5.42 vs. 9.013 ± 3.96, *p* < 0.001; mean CAVI 10.15 ± 2.50 vs. 8.64 ± 1.91, *p* < 0.001; mean CAVI 0 18.27 ± 7.31 vs. 14.11 ± 5.30, *p* < 0.001). Furthermore, the subgroup of patients with previous CV events, as compared to subjects without, had increased CAVI and CAVI 0, whilst PVW-cf was not significantly different between groups (mean CAVI 10.82 ± 2.46 vs. 8.65 ± 1.90, *p* < 0.001; mean CAVI 0 20.13 ± 7.52 vs. 14.18 ± 5.58, *p* < 0.001). When stratifying the study population upon CCI, as CCI increased, we outlined a progressive increase in CAVI (Figure 1A), CAVI 0 (Figure 1B), and PVW-cf (Figure 1C), which remained significant after adjustment for age, sex, MAP and GFR.

#### *3.3. Regression Analysis: Arterial Stiffness Predictors*

Stepwise multiple regression models were performed (Table 3) in order to evaluate the combined effect of independent variables on arterial stiffness parameters. In the first model PWV-cf was considered as a dependent variable; among several independent variables (age, GFR, MAP, CCI, LDL and triglycerides) only CCI resulted as significant predictor of PWV-cf (*p* < 0.001), accounting for 20.5% of its variance. Interestingly, when considering CAVI as dependent variable, and age, GFR, MAP, CCI, LDL and triglycerides as independent variables, both age and CCI resulted to be significant predictors (*p* < 0.001 and *p* = 0.012, respectively), explaining almost 53% of CAVI variance. Likewise, as shown in the third model which considered CAVI 0 as dependent variable and again age, GFR, MAP, CCI, LDL and triglycerides as independent variables, age and CCI (*p* < 0.001 and *p* = 0.010, respectively) could predict CAVI 0, accounting for 55.8% of its variance.

**Table 3.** Stepwise regression analysis, considering PWV-cf, CAVI, and CAVI 0 respectively as dependent variables, and age, glomerular filtration rate, mean arterial pressure, Charlson Comorbidity Index, LDL-Cholesterol and Triglycerides as independent variables.


CCI: Charlson Comorbidity Index; CAVI: cardio-ankle vascular index; PWV-cf: pulse wave velocity carotid-femoral.

#### **4. Discussion**

The present study shows significant positive correlations between all parameters of arterial stiffness and CV risk factors, comorbidities, and aging. The positive correlation with age is stronger for CAVI and CAVI 0 than PVW-cf. Moreover, our data confirm and complement previous knowledge showing that age and comorbidity can predict arterial stiffness parameters.

We could demonstrate a positive relationship between CAVI, CAVI 0, PVW-cf and the main CV risk factors, even after adjustment for age, sex, MAP and GFR. In line with previous evidence [36], in our population all the arterial stiffness indexes resulted increased in hypertensive patients, reflecting the vascular remodeling, characterized by wall stiffening typical of this condition. In particular, we outlined a significant increase in CAVI among hypertensive subjects, which is consistent with the results of Nagayama and colleagues [16], who demonstrated increased CAVI values in a cohort of 2300 individuals, describing a sharper increase after the SBP threshold of 140 mmHg.

As is predictable considering vascular involvement in the diabetes mellitus course [37], we described increased arterial stiffness indexes among diabetic patients and among subjects with impaired fasting glucose levels, as compared to normoglycemic individuals. These results are in line with previous finding regarding both PVW-cf [37] and CAVI [18,38]; the latter was found increased in diabetic patients, showing however a progressive decrease after 8 weeks of glucose lowering therapy, consistent with HbA1c reduction [18]. Moreover, we found a significant correlation between all the arterial stiffness indexes and metabolic syndrome components, confirming previous evidence [12,39,40], and corroborating the hypothesis of increased arterial stiffening as a crucial change in the presence of metabolic disorders or metabolic syndrome.

We further outlined increased CAVI and CAVI 0 in patients with previous CV events, still in line with several studies that described increased CAVI in subjects with known coronary artery disease and cerebral ischemia [25,41–43]. Altogether, our and other results suggest that different vascular diseases, affecting different segments of the arterial tree, share the common finding of increased arterial stiffness. Our findings actually complement previous observations because they show that the heterogeneity of the vessels involved may be more accurate by testing CAVI, instead of PVW-cf, since the first is more representative of a large proportion of the arterial tree [3].

Although several conditions are known to be associated to worse values of CAVI and PVW-cf, the possible association with the comprehensive comorbidities burden is not entirely explored. In this regard, although less is known about increased comorbidity index, CAVI has already been depicted as increased among frail individuals; thus, our results confirm and complement previous evidence by Xue and colleagues who described higher CAVI values in elderly frail patients (relaying on Fried's frailty definition) [44]. Noteworthily, moving one step further, we observed a positive relationship between all the arterial stiffness indexes and both comorbidities number and CCI, still significant after adjustment for age, sex, MAP and GFR. The pathophysiological background of this finding might rely on the vascular remodeling, which occurs during healthy aging [45], and pathological conditions [1]. Certainly, arterial stiffening is the common denominator of several diseases included in the CCI calculation, and therefore a double-sided relation might be inferred: first, arterial wall stiffening in otherwise healthy aging subjects might increase the risk of developing a huge number of vascular-associated conditions. On the other hand, presenting relevant comorbidities (primarily involving or not the vascular system) might promote a complex network of tissue remodeling processes, leading to arterial wall stiffening. Thus, more than a single disease, the comprehensive burden of multiple co-existing conditions might contribute to widespread and increased vascular stiffening. According to the latter interpretation, we could demonstrate that considering arterial stiffness parameters as dependent variables, the comorbidity burden described by CCI is a strong predictor of all the arterial stiffness indexes, and along with age it can explain a consistent percentage of CAVI and CAVI 0 variance. Arterial stiffness might thus be

considered as part of the expression of a multidimensional decline; in particular we found that CAVI, as compared to PVW-cf, was more strongly related to the comorbidity burden, and, once more, a possible explanation lies in the wider proportion of arterial segments that is simultaneously investigated by CAVI, therefore including a broad spectrum of pathological conditions.

According to consolidated knowledge [45], our study confirms a significant association between aging and arterial stiffness: each arterial stiffness index displays significant relation with age; the strength of the association is higher for CAVI and CAVI 0, as compared to PVW-cf. Furthermore, on top of several variables, age is shown to be a significant predictor of CAVI and CAVI 0, yet not of PWV-cf. The remarkable relationship between age and CAVI was previously demonstrated by Shirai et al. [46], who described increased CAVI in elderly subjects as a possible expression of age-related arterial wall sclerosis. Although PVW-cf is still considered as the primary arterial stiffness evaluation technique for outcome prediction [47], CAVI could be also considered as a more reliable index in the elderly, due to its lower dependence from blood pressure [48].

A few limitations of the study should be recognized: this is a cross-sectional study, and therefore we were not allowed to test the power of arterial stiffness indexes in predicating long term cardiovascular risk. Our study was predominantly performed in female patients and given the increased prevalence of cardiovascular diseases in the male population, we need to test our hypothesis in a more represented male population. Further, information regarding medical therapy was not available, but we acknowledge that the possible effect of different medications on arterial stiffness may be of interest. As concerns arterial stiffness parameters, we acknowledge that the augmentation index, which is deemed to be an important parameter, was not significant in our findings, and therefore excluded from the results.

In conclusion, our study, conducted on a relatively wide and heterogeneous cohort of patients, demonstrated that PVW-cf, CAVI and CAVI 0 are associated to CV risk factors and higher comorbidity burden, even after adjustment for several variables. Furthermore, our data outline a strong correlation between arterial stiffness indexes and age. Our findings might complement the pathophysiological understanding of the cardiovascular impairment in subjects with older age and remarkable comorbidity burden. Therefore, in the clinical setting, arterial stiffness evaluation, which is a feasible and easily available technique, may complement the characterization of complex patients.

**Author Contributions:** Conceptualization, F.F. and M.Z.; methodology, F.F., A.G., M.T., A.P.R. and E.Z.; formal analysis, R.M. and F.F.; investigation, F.F., A.G. and G.M.; data curation, F.F., A.G. and R.M.; writing—original draft preparation, F.F., A.G. and E.Z.; writing—review and editing, F.F. and M.Z.; supervision, M.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Approval Code: CE 191CESC University Hospital, Verona, approval Date: 11 February 2015.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Detectable Bias between Vascular Ultrasound Echo-Tracking Systems: Relevance Depends on Application**

**Afrah E. F. Malik 1, Alessandro Giudici 1,2, Koen W. F. van der Laan 1, Jos Op 't Roodt 3, Werner H. Mess 4, Tammo Delhaas 1, Bart Spronck 1,5 and Koen D. Reesink 1,\***


**Abstract:** The Esaote MyLab70 ultrasound system has been extensively used to evaluate arterial properties. Since it is reaching end-of-service-life, ongoing studies are forced to seek an alternative, with some opting for the Esaote MyLabOne. Biases might exist between the two systems, which, if uncorrected, could potentially lead to the misinterpretation of results. This study aims to evaluate a potential bias between the two devices. Moreover, by comparing two identical MyLabOne systems, this study also aims to investigate whether biases estimated between the MyLabOne and MyLab70 employed in this study could be generalized to any other pair of similar scanners. Using a phantom set-up, we performed *n* = 60 measurements to compare MyLab70 to MyLabOne and *n* = 40 measurements to compare the two MyLabOne systems. Comparisons were performed to measure diameter, wall thickness, and distension. Both comparisons led to significant biases for the diameter (relative bias: −0.27% and −0.30% for the inter- and intra-scanner model, respectively, *p* < 0.05) and wall thickness (relative bias: 0.38% and −1.23% for inter- and intra-scanner model, respectively *p* < 0.05), but not for distension (relative bias: 0.48% and −0.12% for inter- and intrascanner model, respectively, *p* > 0.05). The biases estimated here cannot be generalized to any other pair of similar scanners. Therefore, longitudinal studies with large sample sizes switching between scanners should perform a preliminary comparison to evaluate potential biases between their devices. Furthermore, caution is warranted when using biases reported in similar comparative studies. Further work should evaluate the presence and relevance of similar biases in human data.

**Keywords:** echo-tracking; vascular ultrasound; arterial properties; arterial stiffness

#### **1. Introduction**

Arterial properties, such as diameter, wall thickness, and distension, are extensively investigated in the literature [1–6], considering the valuable information they provide about cardiovascular health. Moreover, they are used to quantify arterial stiffness: an independent predictor of cardiovascular diseases [7]. Local arterial stiffness can be characterized by the distensibility coefficient and Young's elastic modulus, among other indices. These indices require the assessment of the instantaneous diameter change and wall thickness by means of ultrasound echo-tracking [8–10].

Efforts made by prof. Hoeks and his group [2,11,12] represent seminal endeavors for the utilisation of ultrasound echo-tracking in the field of large artery (patho-)physiology.

**Citation:** Malik, A.E.F.; Giudici, A.; van der Laan, K.W.F.; Op 't Roodt, J.; Mess, W.H.; Delhaas, T.; Spronck, B.; Reesink, K.D. Detectable Bias between Vascular Ultrasound Echo-Tracking Systems: Relevance Depends on Application. *J. Clin. Med.* **2023**, *12*, 69. https://doi.org/ 10.3390/jcm12010069

Academic Editors: George N. Kouvelos, Andrea Grillo and Paolo Salvi

Received: 30 September 2022 Revised: 15 November 2022 Accepted: 17 December 2022 Published: 21 December 2022

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Their efforts mainly focused on analysing radiofrequency (RF) signals to estimate arterial distensibility. Hoeks' group developed the necessary software and hardware (ART.LAB) to be integrated ultimately with the MyLab70 (Esaote Europe B.V. Maastricht, the Netherlands) and to present a top-class research-oriented echo-tracking system. In the past two decades, the MyLab70/ART.LAB combo has been used extensively in expert centers to quantify arterial elastic and geometrical properties, predominantly in a research context [13–15]. At the same time, ultrasound manufacturers continued to incorporate such technology into commercial devices with the objective of bringing the technology to clinical practice. Today, the MyLab70/ART.LAB has reached its end-of-service-life, forcing ongoing longitudinal epidemiological and interventional studies to switch to another scanner, which, because of technical differences, may not necessarily provide identical results. For instance, the Maastricht Study opted for the MyLabOne with RFQAS and RFIMT functionalities, plus a radiofrequency output license (Esaote) [14], with technology based on the original radiofrequency tracking [2]. The newer system represents a portable, integrated, and more affordable substitute to MyLab70/ART.LAB. The two systems, however, have different technical characteristics, and hence, their use at two-time points of a longitudinal study might result in a bias in the follow-up measurement. Not considering or correcting for such an ultrasound system-related bias may lead to misinterpretation of results and, thereby, erroneous conclusions.

The primary aim of the present study was to compare MyLabOne- and MyLab70 based echo-tracking systems to assess potential bias in quantifying diameter, wall thickness, and distension. In addition, we explored if such biases might also arise when comparing two MyLabOne systems with identical specifications. We conducted the comparative measurements in two steps, using a phantom set-up and three ultrasound systems: MyLabOne I, MyLab70, and MyLabOne II. First, we compared MyLab70 and MyLabOne I. We refer to this comparison as inter-scanner model comparison. Next, we estimated the biases between MyLabOne I and MyLabOne II. We refer to this comparison as intra-scanner model comparison. To confirm that the inter- and intra-system model biases estimated in this study are not spurious but that they originate from real intra- and inter-system model differences, we also performed an intra-device comparison. To this end, we assessed the differences between two measurement sets performed with MyLabOne I.

#### **2. Materials and Methods**

#### *2.1. Ultrasound Scanners*

Measurements in this study were performed using three different ultrasound systems: MyLabOne I, MyLabOne II, and MyLab70. A summary of the specifications of these systems is shown in Table 1. MyLab70 was equipped with a linear array transducer operating at 7.5 MHz and had a practical axial resolution of 0.125 mm. MyLabOne systems were equipped with linear array transducers operating at 10 MHz. The systems had a practical axial resolution of approximately 0.120 mm. All three systems were operating in fast B-mode, with high frame rates of 498 and 524 for MyLab70 and the MyLabOne systems, respectively. These high frame rates are achieved by generating multiple M-lines separated by 0.98 mm in the longitudinal direction of the probe. The number of M-lines is, however, different between the two scanners: *n* = 19 for MyLab70 and *n* = 14 for MyLabOne. In addition, all three scanners enabled the recording of raw radiofrequency signals sampled at 50 MHz for MyLab70 and 33 MHz for MyLabOne.


**Table 1.** Specifications of echo-tracking systems used in the study.

RF—radiofrequency; \* Practical here refers to the resolution estimated and based on the actual bandwidth measured from received RF signals in contrast to theoretical axial resolution.

#### *2.2. Phantom Configuration*

To perform the inter- and intra-ultrasound system model comparisons to estimate scanner biases for assessing diameter, wall thickness, and distension, a two-part phantom set-up was used (Figure 1). The first part consisted of a static silicone tube with an outer diameter of 12.4 mm and a wall thickness of 1 mm. This part was used to estimate biases between the ultrasound system pairs for diameter and wall thickness. The second part consisted of an eccentric wheel connected to a motor via a rod, which was inserted in the wheel 300 μm off centre. Thus, measuring the instantaneous location of the top surface of the wheel results in a sinusoidal distension waveform with a peak-to-peak amplitude of 600 μm. A silicone slab was mounted above the wheel to mimic the artery near wall. This near wall was held in a fixed position; hence, it did not contribute to the simulated vessel distension. The phantom set-up and the transducer lens were immersed in tap water at room temperature to enable ultrasound propagation.

**Figure 1.** Study set-up for ultrasound system bias estimation (**A**) and how it appeared in the ultrasound measurements (**B**). Two-part phantom set-up consisting of a silicone tube (1), used to assess inter- and intra-scanner biases in diameter and wall thickness, and an eccentric wheel with a silicone slab mounted on top of it (2), used to assess the bias in distension. (B.1): Example of a B-mode image for the silicone tube. The white dot markers indicate the outer tube–water echo interfaces (appearing as horizontal white lines) of the near (top) and far (bottom) walls of the silicone tube used to measure the diameter. The green line indicates the inner tube–water echo interface of the far wall, which, together with the far wall outer echo interface, was used to measure wall thickness. (B.2): Example of an M-mode acquisition of the wheel 'distension'. The white sinusoidal line reflects the motion of the wheel surface, while the two less echogenic parallel reflections above the undulating line indicate the silicone slab.

#### *2.3. Data Acquisition*

To compare MyLabOne I and MyLab70, we obtained *n* = 60 repeated RF acquisitions for both scanners. Post hoc analysis using data from the previously mentioned comparison revealed that after about 40 repeated measurements, the bias and confidence intervals (CI) remained relatively constant for all examined variables indicating that *n* = 40 repeated measurements were sufficient to provide a reliable estimate of the bias. Therefore, for the intra-device model comparison, we performed *n* = 40 repeated measurements using MyLabOne I and MyLabOne II for diameter and wall thickness. Moreover, we inferred from the inter-device comparison that part of the distension recordings might be lost potentially due to uncontrolled saturation; hence, *n* = 60 repeated distension measurements were performed for the intra-device model case. The MyLabOne I measurement set available for the inter-scanner model comparison was also used to perform an intra-device comparison. For all the studied variables, two groups were created by dividing this measurement set into two groups based on the order of performing the measurement (even and odd). Individual acquisitions were performed by repositioning the ultrasound probe at random distances (ranging between 1 and 3 cm) from the tube and the wheel. This was conducted to simulate in vivo situations due to the fact that the depth from the skin to the carotid artery varies between different individuals.

#### *2.4. Data Processing*

The diameter was defined as the distance between the near and far wall outer siliconewater reflections and is indicated by white dotted lines in Figure 1B.1. Wall thickness was defined as the distance in the far wall between the inner and outer silicone-water reflections. Both diameter and wall thickness were estimated based on longitudinal acquisitions covering 19 and 14 equidistant M-lines for Mylab70 and MyLabOne, respectively. On the contrary, cross-sectional acquisitions of the wheel motion were performed to estimate distension. Since the wheel thickness was 4 mm and the distance between the ultrasound M-lines was approximately 1 mm, only a few M-lines covered the wheel. Therefore, we estimated the distension based on a single M-line with the most wheel coverage [9]. This was deduced based on the brightness of the corresponding B-mode of the line, indicating a strong wheel reflection. RF signals were processed in MATLAB (MATLAB R2020b; The MathWorks, Natick, MA, USA) using proprietary wall-tracking software that was previously described in [2,12,16].

#### *2.5. Statistical Analyses*

RF recordings were acquired for at least five seconds for all scanners. Biases were quantified as means ± 95% CI and are reported in absolute and relative terms. For all three considered variables, the absolute bias was calculated as the difference between the average of estimates obtained with MyLabOne I and MyLab70 (i.e., MyLabOne I- minus MyLab70-derived values) and between those obtained with MyLabOne I and MyLabOne II (i.e., MyLabOne I- minus MyLabOne II-derived values) for the inter-device and intra-device comparisons, respectively, and tested with an independent sample Student's *t*-test. The relative bias was defined as the absolute bias normalized with respect to the mean value of both systems. Precision was assessed by the estimates' standard deviation (SD) and compared with F-test. Statistical analyses were performed using SPSS version 27 (SPSS, Chicago, IL, USA). A two-sided *p*-value <0.05 was considered statistically significant.

#### **3. Results**

#### *3.1. MyLabOne I vs. MyLab70*

Post hoc, 21 MyLabOne I distension recordings were excluded due to the uncontrolled saturation in the corresponding RF complex [17].

The diameter obtained with MyLabOne I was significantly lower than that obtained with MyLab70 (12.3830 vs. 12.4170 mm, *p* = 0.001), corresponding to a relative bias of −0.27%. However, the precision of the diameter measurements defined as the SD was similar for the two scanners (0.0533 vs. 0.0527 mm, *p* = 0.542). Compared to MyLab70, MyLabOne I resulted in a significantly higher wall thickness (1.0019 vs. 0.9981 mm, *p* = 0.004) which translated into a relative bias of 0.38%. Further, MyLabOne I yielded a

significantly higher standard deviation for wall thickness (0.0079 vs. 0.0062 mm, *p* < 0.001). The SD obtained with MyLabOne I for distension was significantly higher than that achieved with MyLab70 (17.8 vs. 12.1 μm, *p* = 0.047). However, we found no significant difference between the two scanners for distension (617.0 vs. 614.1 μm, *p* = 0.333) (Table 2 and Figure 2).

**Figure 2.** Overview of absolute values of all repeated measurements performed for the inter- and intra-scanner model comparisons. Measurements were performed with MyLab70, MyLabOne I, and MyLabOne II for diameter (**A**), wall thickness (**B**), and distension (**C**). Solid lines indicate the medians and dashed lines indicate the 25th and 75th percentiles.


**Table 2.** Diameter, wall thickness, and distension as determined by MyLabOne I, MyLab70, and MyLabOne II for the inter- and intra-device model and the intra-device comparisons.

SD—standard deviation; CI—confidence intervals.

#### *3.2. MyLabOne I vs. MyLabOne II*

Of the distension measurements performed for the intra-scanner model comparison, *n* = 23 were excluded for MyLabOne I and *n* = 15 were excluded for MyLabOne II due to uncontrolled saturation in the corresponding RF complex.

The comparison of the two MyLabOne systems yielded significantly different diameter estimates (12.3569 vs. 12.3945 mm, *p* < 0.001), corresponding to a relative bias of −0.30%. However, the two systems yielded similar SDs (0.0222 vs. 0.0267 mm, *p* = 0.343) for diameter. The two systems resulted in significantly different wall thickness measurements (0.9855 vs. 0.9976 mm, *p* < 0.001), translating into a relative bias of −1.23%. Further, the two scanners yielded significantly different SDs of the wall thickness (0.0110 vs. 0.0048 mm, *p* = 0.013). We found no significant difference between the distension estimates obtained with the two scanners (609.5 vs. 610.2 μm, *p* = 0.892). Similarly, there was no significant difference between the SDs of the distension estimates obtained with the two systems (25.5 vs. 21.4 μm, *p* = 0.591) (Table 2 and Figure 2).

#### *3.3. MyLabOne I vs. MyLabOne I*

The results of this comparison are shown in Table 2. The intra-device comparison yielded statistically non-significant differences for diameter (difference −0.0012 mm, *p* = 0.929), wall thickness (difference = −0.0026 mm, *p* = 0.198), and distension (difference −3.6 μm, *p* = 0.529). Similarly, the two measurement sets of MyLabOne I resulted in statistically non-significant SDs for all the examined variables (*p* > 0.05). Because of the considerable intercurrent time between the two available MyLabOne I measurement sets (i.e., one set for the inter- and one for the intra- system model comparisons) as well as the lack of consistency in measurement conditions/set-up status, we refrained from intra-device comparison based on the two available MyLabOne I measurement sets.

#### **4. Discussion**

Using a phantom set-up, this study assessed the inter- and intra-scanner biases between MyLabOne I- and MyLab70-based echo-tracking systems for measuring arterial diameter, wall thickness, and distension. Our results show detectable biases for diameter and wall thickness but not for distension. This held true for the comparison between the MyLab70 and a MyLabOne I system, as well as for the comparison between two MyLabOne systems. Biases were in the same order of magnitude in both comparisons. All biases were very small with respect to the values reported in the literature for studies comparing two echo-tracking systems (Table 3) [1,3,4]. Based on our results, research studies should adhere to one device unless switching is necessary. Whenever replacement is unavoidable, a comparison between the two systems should be performed to establish the amount of bias, even if the devices have the same vendor and model.

To the best of our knowledge, this is the first study to compare MyLabOne and MyLab70, as well as two MyLabOne systems with identical specifications. MyLab70 and MyLabOne share several common features. Indeed, they are both RF-based echo-tracking systems designed and manufactured by the same manufacturer. In addition, the two systems employ conceptually similar RF tracking approaches [2]. As shown in Table 3, the biases between MyLabOne and MyLab70 for all the examined variables were lower than the values reported by similar studies comparing two different devices/models. This indicates that MyLabOne is a good substitute for MyLab70.

To ensure that the non-significant bias found in the case of the distension was not due to insufficient statistical power, we performed a post hoc power analysis. This analysis was performed using G\*Power version 3.1.9.4: an open-source statistical power analysis tool available at https://www.gpower.hhu.de (accessed on 23 June 2021) [18]. The power analysis showed that a sample size of 39 would enable us to detect an effect size greater than 0.64 (power 80% and a two-sided α = 0.05). Note that based on our study design, we would, thus, be able to detect a bias greater than 64% of the device's precision.


**Table 3.** Studies found in the literature that compare two different ultrasound devices for measuring arterial diameter, wall thickness, and distension.

Bold texts highlight the current study.

Biases between MyLabOne I and MyLab70 systems for all examined variables did not exceed 0.5%, which appears clinically irrelevant for personalized risk stratification and diagnosis (e.g., in the context of cardiovascular risk assessment). However, the findings presented in this study may have direct implications for research studies, particularly follow-up designs. By alleviating the effect of device-related biases on the outcomes of these studies, the findings presented here have an indirect clinical relevance. For such studies, an appraisal of the relevance of the bias depends on multiple factors, with the sample size/statistical power being the most important. For instance, for the same value of the bias, a low population variability would lead to significant results with a small sample size, while larger variability requires a larger sample size for the results to be relevant.

Let one consider the lowest sample size (*n*min) beyond which the estimated biases would be considered relevant. In other words, studies with a sample size exceeding *n*min should consider the effect of the inter-/intra-scanner bias in their analysis and interpretation. Figure 3 shows the results of our calculations of *n*min for a range of population variabilities. Based on the results presented in Figure 3, research studies are recommended to consider their population variability/statistical power when evaluating the relevance of the biases detectable between MyLabOne and MyLab70 systems.

**Figure 3.** Lowest sample size (*n*min) for estimated biases to be considered relevant. *n*min is estimated using several values of standard deviation for (**a**) diameter, (**b**) wall thickness, and (**c**) distension. We assumed that variability within research studies conducted in humans could be expected to be larger than that observed here using a phantom and, hence, considered a range of variabilities (defined with SD) in quantifying *n*min. The black line represents *n*min as a function of SD, while the red area represents sample sizes for which estimated biases are considered significant. Black dots represent *n*min for a significant bias calculated using the SDs observed in this study, and they correspond to 21, 28, and 193 samples for diameter (**a**), wall thickness (**b**), and distension (**c**), respectively.

This study found differences between two identical-on-paper systems (MyLabOne I and MyLabOne II) for diameter, and wall thickness, indicating that the results found here for the inter-scanner comparison could not be generalized to any other pair of similar scanners. These findings also imply that caution is warranted when using systematic biases reported in similar comparative studies. A potential explanation for the intra-scanner model differences relates to the different operational periods between our two systems and the supposed 'wear' effect on data quality. Another possible explanation relates to the uncertainty in the manufacturing process, which is determined by the adopted tolerance, and the admissible variation in the end product.

The intra-device comparison was performed to check if the differences found in the cases of the inter- and intra- system model comparisons were spurious, originating from factors such as study set-up, environmental conditions, and wear effect or if they were real, originating from inter- and intra- system model differences. Intra-device differences for all the studied variables were not statistically significant (Table 2), confirming that the significant differences found in the cases of inter- and intra-scanner model comparisons originated from real device/device model differences. Compared to inter- and intra- system model differences, intra-device differences were smaller for diameter and wall thickness and larger for distension. We believe that the difference found in the case of distension was caused by the relative uncertainty of the distension estimate.

For studies switching between devices, a similar phantom set-up and approach could be used to calibrate the new system against the old one to avoid any effect that a systematic bias between the two systems could have on the study outcomes. Phantom set-ups are controllable and provide repeatable estimates; hence, they are superior to human data for calibration purposes. By using a phantom, one mitigates additional uncertainty in bias estimates originating from human data variability.

This study has several possible limitations: (1) Some distension recordings were excluded due to saturation in the RF complex. This problem was experienced with the MyLabOne systems but not with MyLab70. The eccentric wheel used in the phantom set-up was made of a strong reflector; hence, with certain gain settings, the peaks of the incoming RF signal may exceed the dynamic range (16 bit or 96 dB) of the scanner. While adjusting the gain setting was possible with MyLab70 during the RF acquisitions, this option was not available for MyLabOne, explaining the occurrence of saturation issues. (2) The set-up used here consisted of homogeneous materials, and diameter and wall thickness measurements were performed under static conditions. Tissue inhomogeneity

and vessel wall motion might alter bias estimates under in vivo situations. (3) In vivo arterial diameter and wall thickness (defined as intima-media thickness) are typically smaller than those of the phantom tube. Hence, the effect of the ultrasound scanners' limited axial resolution may be expected to be more pronounced during in vivo settings.

#### **5. Conclusions**

The present study found detectable inter- and intra-scanner model biases for diameter and wall thickness measurements but not for distension measurements. The existence of a detectable bias between two identical systems/models indicates that the biases estimated in the present study cannot be generalized to any other pair of scanners. Therefore, studies with large sample sizes and particularly those with longitudinal designs, in which a change in or an exchange of scanners is necessary, should check for the presence of biases between devices following our approach. Further work should evaluate the presence and relevance of biases in (existing) human studies.

**Author Contributions:** A.E.F.M. and K.D.R.: conceptualization. A.E.F.M., K.D.R., W.H.M., A.G., B.S. and T.D.: methodology. A.E.F.M., K.D.R., A.G. and B.S.: data analysis. A.E.F.M.: data processing. A.E.F.M. and K.D.R.: writing—original draft preparation. A.E.F.M., K.D.R., W.H.M., A.G., B.S., K.W.F.v.d.L., J.O.'t.R. and T.D.: writing—review and editing. T.D., W.H.M. and K.D.R.: supervision. All authors have read and agreed to the published version of the manuscript.

**Funding:** A.E.F.M. was supported by the European Union-funded Horizon 2020 project InSiDe (no. 871547). B.S. was supported by the European Union's Horizon 2020 research and innovation program (no. 793805).

**Data Availability Statement:** Data will be made available upon request to the corresponding author.

**Acknowledgments:** The authors would like to thank Arnold P. G. Hoeks for developing the Distension software code that was used to analyse the RF recordings in this study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

MDPI St. Alban-Anlage 66 4052 Basel Switzerland www.mdpi.com

*Journal of Clinical Medicine* Editorial Office E-mail: jcm@mdpi.com www.mdpi.com/journal/jcm

Disclaimer/Publisher's Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Academic Open Access Publishing

mdpi.com ISBN 978-3-0365-9331-9