Cardio-ankle Vascular Index Is More Closely Associated Than Brachial-ankle Pulse Wave Velocity With Variables Predicting Future Cardiovascular Disease in Patients With Diabetes


 Background: This study aimed to compare the usefulness of arterial stiffness parameters, cardio-ankle vascular index (CAVI) and brachial-ankle pulse wave velocity (baPWV), for evaluating future cardiovascular disease (CVD) in subjects with diabetes.Methods: The study subjects were 277 patients with type 1 or type 2 diabetes. All subjects were evaluated for vascular stiffness using CAVI (n = 154) or baPWV (n = 123). Carotid intima-media thickness (IMT) and the Suita score were also measured because these are established risk factors for future CVD. Associations of both CAVI and baPWV with these established parameters were evaluated in all subjects, and then in 172 subjects with adjustment for covariates by using propensity score matching.Results: In all subjects, CAVI and baPWV correlated significantly with both IMT (r = 0.470, P < 0.001, and r = 0.256, P = 0.004, respectively) and the Suita score (r = 0.558, P < 0.001, and r = 0.360, P < 0.001, respectively). The correlation between CAVI and IMT was more significant than that between baPWV and IMT (Z = 2.03, P = 0.042).　Similarly, the correlation between CAVI and the Suita score was more significant than that between baPWV and the Suita score (Z = 2.07, P = 0.039). After adjustment by matching, significant correlations between CAVI and IMT (r = 0.459, P < 0.001) and between CAVI and the Suita score (r = 0.526, P < 0.001) were preserved, though only the association between baPWV and the Suita score was significant (r = 0.270, P = 0.011) while that between baPWV and IMT showed no significance. Again, CAVI showed a significantly stronger association with the Suita score than baPWV (Z = 1.99, P = 0.046).Conclusions: CAVI is more closely associated than baPWV with variables predicting future CVD in patients with diabetes.


Background
Subjects with dyslipidemia, hypertension and diabetes mellitus, who smoke, have a high risk of developing cardiovascular disease (CVD) [1]. Furthermore, CVD risk factor clusters reportedly raised CVD risk in a general population cohort [2]. Therefore, evaluating the risk of developing CVD is important for improving CVD mortality in patients with diabetes who have CVD risk factors. Among many tools for CVD risk assessment, brachial-ankle pulse wave velocity (baPWV) [3] and cardio-ankle vascular index (CAVI) [4] are useful and noninvasive. Both baPWV and CAVI evaluate vascular stiffness, which is recognized as a surrogate marker predicting CVD risk [5][6][7][8]. The advantages of CAVI and baPWV for evaluating atherosclerosis and for CVD risk assessment remain controversial [9][10][11]. However, to our knowledge, no studies have compared usefulness for CVD risk assessment between baPWV and CAVI in patients with diabetes.
Thus, we aimed to compare usefulness for cardiovascular risk assessment between baPWV and CAVI in patients with diabetes. We selected carotid intima-media thickness (IMT) and the Suita score for evaluating the risk of CVD development. Carotid IMT is widely recognized as a surrogate marker for the risk of CVD in the future [12,13]. Additionally, the Suita score provides suitable risk factor categories for predicting the ten-year probability of coronary heart disease (CHD), and is more accurate for predicting CHD risk than the Framingham risk score in the Japanese population [14].

Study subjects
Two hundred seventy-seven patients with type 1 or type 2 diabetes (173 men and 104 women including 9 type 1 diabetes patients, average age 64.8 ± 11.5 years) were recruited as study subjects. All subjects were ambulatory and were followed at the Department of Neurology, Hematology, Metabolism, Endocrinology and Diabetology, Yamagata University Faculty of Medicine and Division of Diabetes and Metabolic Diseases, Department of Internal Medicine, Nihon University School of Medicine. Patients with atrial brillation, peripheral arterial disease, malignant diseases, collagen diseases, acute and chronic in ammatory diseases, and/or receiving steroid hormone therapy or other immunosuppressants, were excluded from this study.

Characteristics of study subjects
We determined clinical characteristics including sex, age, body mass index (BMI), smoking habit, systolic and diastolic blood pressures, anti-hypertensive drug use, statin use, and biochemical variables in all subjects. Biochemical variables, including lipid metabolic parameters, uric acid and HbA1c were measured after an overnight fast. Low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, uric acid, creatinine and HbA1c were measured using an automatic analyzer. The estimated glomerular ltration rate (eGFR) served as an indicator of renal function. eGFR was estimated by the following formula: eGFR (mL/min/1.73 m 2 ) = 194 × Serum creatinine −1.094 × Age −0.287 × 0.739 (if female) [15]. Blood pressure was measured with the patient in a sitting position at the hospital in the morning.

Examination of atherosclerosis and cardiovascular risk
Carotid IMT, CAVI and baPWV were measured as variables associated with atherosclerosis. Carotid IMT was established as a suitable surrogate marker for the risk of future CVD development [12,13]. A total of six segments of the near and far walls in the common carotid artery, at the bifurcation, and in the internal carotid artery on the right and left were measured with B-mode imaging of ultrasonography, as described in a previous report [16]. The maximum IMT, including bilateral plaque, was de ned as the IMT in all study subjects [16]. Carotid ultrasonographic measurements were performed by experienced clinician.
The IMT measurements showed a variability of 8.0%, as previously reported [16]. CAVI 4 and baPWV [3] are indicators of arterial stiffness. CAVI is an index of arterial stiffness based on the stiffness parameter β [4], while baPWV re ects the stiffness from the aorta to the lower limb arteries [3].
CAVI [5,6] and baPWV [7,8] have been recommended to surrogate markers for CVD. CAVI was measured using a Vasera VS-1000 vascular screening system (Fukuda Denshi, Tokyo, Japan). The maximum CAVI on both sides was recorded in each of the study subjects (n = 154) enrolled at Yamagata University Hospital and the maximum baPWV measurement was performed using a form PWV/ABI (Omron Healthcare Co., Ltd. Kyoto, Japan) in subjects (n = 123) enrolled at Nihon University Itabashi Hospital. CAVI and baPWV in this study were de ned as the largest CAVI and baPWV between those of both sides.
For predicting the risk of CVD development, we used the Suita score. The Suita score is an established cardiovascular risk score based on risk factor categories for predicting CHD in the Japanese population [14]. The Suita score consists of the sum of each of these four risk categories and indicates the ten-year probability of CHD [14].

Statistical analysis
All subjects were divided into the CAVI (n = 154) and baPWV groups (n = 123). Furthermore, to adjust for clinical characteristics, we performed one-to-one propensity score matching [17]. Each patient in the CAVI group was paired with a patient in the baPWV group based on the propensity score in this one-to-one matching. Propensity scores were calculated using logistic regression analysis with the covariates of sex, age, BMI, smoking, systolic and diastolic blood pressure, receiving statins and antihypertensive drugs, HDL cholesterol, LDL cholesterol, triglyceride, uric acid, eGFR and HbA1c. Patients with the nearest propensity score within the caliper were paired. A caliper size in the range of 0.20 to 0.25 standard deviation (SD) is recommended [18,19]. This study de ned a caliper as 0.20 SD. As a result, 86 patients were in both the CAVI and the baPWV group after propensity score matching. The Mann-Whitney U test and the chi-square test were performed to compare clinical characteristics between the CAVI and baPWV groups before or after adjustment by matching. Bonferroni's multiple comparison test was used to compare the mean values of CAVI and baPWV between age groups. Pearson's correlation coe cient and univariate linear regression analysis were used to identify whether CAVI and the Suita score were signi cantly associated with clinical characteristics, IMT and the Suita score in subjects before or after adjustment by matching. In this univariate linear regression analysis, we assumed CAVI and baPWV to be dependent variables, and sex (men), smoking habit (current), anti-hypertensive drug use or statin use to be independent variables. The observed Z test was used to compare and analyze statistical signi cance between correlation coe cients. The difference in variance between CAVI and baPWV was evaluated using the F-test of equality of variances. Data are presented as means ± SD, number (%), coe cient of covariation (r), Z values (z), F-value or β coe cients. A value of P < 0.05 was considered to indicate statistical signi cance. All analyses were performed with IBM SPSS Statistics for Windows Version 25 J (IBM Corp., Armonk, NY, USA).

Associations between CAVI or baPWV and clinical characteristics
Mean values of CAVI and baPWV in all study subjects were 8.69 and 17.74 m/sec., respectively. Age showed signi cant associations between CAVI and baPWV both before (r = 0.654 and 0.494, respectively, P < 0.001) and after (r = 0.632 and 0.475, respectively, P < 0.001) adjustment by matching (Table 1). CAVI and baPWV rose with age ( Fig. 1) in all subjects. Additionally, variations in baPWV in each age group or all subjects were signi cantly greater those in CAVI (Supplementary Table S1). Adjustment by matching did not change the in uence of age on CAVI and baPWV (data not shown).
Univariate linear regression analysis indicated signi cant associations between baPWV and antihypertensive drug use (β = 1.547, P = 0.029) before adjustment by matching. After adjustment by matching, however, current smoking was found to show a signi cant correlation (β = -1.195, P = 0.026), while the signi cant association with antihypertensive drug use was unchanged (β = 1.796, P = 0.031).

Discussion
Our results indicate both CAVI and baPWV to be signi cantly associated with carotid IMT and the Suita score, widely used surrogate variables for assessing atherosclerosis and the risk of developing CVD in the future. Notably, CAVI was found to be more closely associated than baPWV with variables predicting CVD development in patients with diabetes. To our knowledge, this is the rst study to demonstrate a difference in clinical signi cance between CAVI and baPWV in patients with diabetes.
We can speculate as to why CAVI more closely re ects atherosclerosis and CVD development in the future than baPWV in patients with diabetes. One possibility is that peripheral arterial stiffness in uences CAVI and baPWV measurements. CAVI re ects central arterial stiffness because the basic formula for calculating CAVI is based on heart-ankle PWV. 4 CAVI is more in uenced by central arterial stiffness than baPWV because the route of the pulse wave in baPWV re ects the status of both central and peripheral arteries, including those from the aortic annulus to the brachium [20]. Atherosclerosis, renal and cardiac function are reportedly more closely associated with central arterial stiffness (carotid-femoral PWV) than baPWV [21]. Our results also revealed a signi cant inverse association between CAVI and eGFR before and after adjustment by matching, whereas the signi cant association between baPWV and eGFR was seen only before adjustment. Furthermore, a recent study of patients with chest pain syndrome showed that carotid IMT had a somewhat more signi cant association with CAVI than with baPWV [9].
Another factor is that CAVI shows higher reproducibility than baPWV. We found variations in baPWV in each age group as well as in all subjects to be signi cantly larger than those in CAVI. We thus suggest that the blood pressure at the time of measurement may in uence the difference in variation between CAVI and baPWV. baPWV is decreased in association with reductions in systolic and diastolic blood pressures when antihypertensive agents are administered, but CAVI does not change [22]. Furthermore, baPWV showed a signi cant increase when systolic blood pressure elevated more than 10 mmHg, while CAVI showed no signi cant change [23]. These studies suggested that baPWV is more susceptible than CAVI to changes in blood pressure. Therefore, we speculate that baPWV shows less reproducibility than CAVI due to blood pressure variability at the time of measurement. The blood pressure variability in patients with diabetes is high compared to that in subjects with normal glucose tolerance [24]. In addition, this variability increases with age [25].
The accuracy of the path length formula also in uences the variation in both CAVI and baPWV measurements. Magnetic resonance imaging results established that the estimated path length from the heart to the ankle in CAVI well re ects the true path length [26]. Conversely, the arterial path length formula in the baPWV measurement is calculated based on the path length from the aortic annulus to the ankle subtracted from the path length from the aortic annulus to the brachium, and this arterial path length is overestimated because the estimated path length from the aortic annulus to the brachium in the baPWV measurement is shorter than the actual path length [27].
This study demonstrated signi cant associations of systolic blood pressure with both CAVI and baPWV.
The latter is closely associated with blood pressure because baPWV depends on blood pressure at the time of measurement [22]. Furthermore, subjects with hypertension had a higher average baPWV than subjects without hypertension [28]. Indeed, we found taking antihypertensive medication to be signi cantly positively associated with baPWV. The reasons for systolic blood pressure being signi cantly correlated with CAVI are, however, uncertain. CAVI is independent of blood pressure at the time of measurement because the determination of CAVI is based on measurement of the stiffness parameter β that is not in uenced by blood pressure [4]. However, CAVI was reported to be signi cantly, but more weakly than baPWV, correlated with blood pressure in subjects receiving routine health checkups [23]. Moreover, a cross-sectional study in elderly individuals indicated that systolic blood pressure in high-CAVI group (CAVI ≥ 9) patients correlated signi cantly with CAVI, whereas blood pressure in the low-CAVI (CAVI < 9) group patients showed no signi cant associations with CAVI [29]. The results of these and our studies allow us to conclude that the CAVI, as well as baPWV, might be in uenced by blood pressure.
The elevations of CAVI and baPWV with age have previously been documented [6, 28], and our study also showed age-related increases in CAVI and baPWV. Increasing aortic stiffness with age contributes to the observed increases in CAVI [30] and baPWV [31]. Our results also demonstrated that CAVI and baPWV well re ect increased aortic stiffness with age in patients with diabetes.
We demonstrated CAVI to be inversely associated with BMI before and after adjustment by matching, while baPWV showed an association only before adjustment. Previous cross-sectional studies indicated CAVI and baPWV to correlate negatively with BMI [32,33]. The reasons for these inverse associations have yet to be clari ed. Body length affects the measurements of both CAVI and baPWV because the formulas for calculating these indices include pulse wave velocity from the heart to the ankle [4] and the brachium to the ankle [3], respectively. BMI is calculated as body weight/height squared, and BMI would thus be expected to have an inverse association with CAVI and baPWV. Additionally, in our study, baPWV showed a negative association with triglyceride. BMI was positively associated with triglyceride (data not shown). It is reasonable to speculate that the positive association between BMI and triglyceride in our study explains the signi cant inverse association between baPWV and triglyceride before and after adjustment by one-to-one case matching.
The invert association between smoking and baPWV was indicated after adjustment by matching in our study, although acute or chronic smoking causes a signi cant increase in arterial stiffness [34]. The reasons why smoking was negatively associated with smoking remain unclear. Subjects with smoking habit were younger, and indicated higher prevalence of the use of antihypertensive medication than subjects without (data not shown). We concluded that receiving antihypertensive medication and the age in subjects with smoking habit may cause the invert association between baPWV and smoking habit.

Conclusion
Our study demonstrated that CAVI is more closely associated than baPWV with the variables predicting the risk for developing CVD in the future in patients with diabetes. Thus, CAVI might be a more suitable tool for cardiovascular risk assessment than baPWV. CAVI appears to be particularly useful for assessing patients with diabetes who show a high risk for developing CVD. Abbreviations CAVI; cardio-ankle vascular index, baPWV; brachial-ankle pulse wave velocity, CVD; cardiovascular disease, IMT; intima-media thickness, CHD; coronary heart disease, BMI; body mass index, LDL; lowdensity lipoprotein, HDL; high-density lipoprotein, eGFR; estimated glomerular ltration rate, SD; standard deviation.