Multiple non-invasive peripheral vascular function parameters with obesity and cardiometabolic risk indicators in school-aged children

Our subjects were recruited between September 2014 to May 2015 from the Minhang province of Shanghai, which were nested in the 2013–2015 China Child and Adolescent Cardiovascular Health (CCACH) Study [15]. The CCACH study was a nationwide cross-sectional study covering Chinese children and adolescents. The full criteria of participants were described elsewhere [15]. Students aged 7–17 years with complete demographic, anthropometric, blood pressure, biochemical measurement, and free of congenital heart diseases, diabetes mellitus, peripheral vascular diseases or hypertension were invited into this study. Among respondents whose parents agreed to participate, a cluster sampling was conducted by enrolling at least 40 students per age group with balanced sex.

The enrollment was on-going until the target sample size of 540 in total was reached, which ensured power > 0.8 for correlations with r > = 0.12 and alpha of 5%.

Data of demographic information (gender, age), anthropometric measurements and cardiometabolic indicators were directly extracted from CCACH database [15]. All measurement were performed at school levels. Height(cm), waist circumference(cm) (WC) and weight (kg) were measured to the nearest 0.5 cm or 0.5 kg with bare foot and light clothing, using standard stadiometers and weight scales in adherence to protocols of the Physical Fitness and Health Surveillance of Chinese School Students [15]. BMI was calculated as body weight (kg) divided by squared height (m²), and BMI SD score was further computed [16]. Students with overweight and obesity were identified according to national classification criteria [Table 10 in [17]]. Blood pressure were assessed on the right arm with an appropriately sized cuff using an oscillometric device (HEM-7012, Omron Healthcare, Kyoto, Japan) after at least 5 min’ rest. The measurements were repeated three times in at least 30-s intervals and the mean values were used. Body fat mass and serum insulin levels were available in 182 sub-samples aged 15–17 years. Body fat mass (%) was measured by a trained and qualified nurse using dual energy X-ray absorptiometry (DEXA, Hologic, Inc., Bedford, MA). Serum insulin levels were measured by chemiluminescence methods using Human Insulin kits purchased from Beckman Coulter (California, USA). Fasting glucose and lipid profiles (TC, Total Cholesterol; TG, Triglycerides; HDLC, high density lipoprotein cholesterol; LDLC, low-density lipoprotein cholesterol) were available in 512 subjects. All blood biochemistries were measured on the morning following an overnight fast. Homeostasis model of assessment for insulin resistance index (HOMA-IR) [fasting insulin (μIU/ml) × fasting glucose (mmol/L)/22.5] [18] was also calculated.

Vascular function was assessed at school levels within 2 weeks after subjects’ enrollment. Endo-PAT™ 2000 device (Itamar Medical Limited, Israel) [19] was applied in adherence to official device operation manual. The detailed description of Endo-PAT application can also be found in a recent methodological review [13]. Briefly, Endo-PAT™ is a computer-based system based on the use of PAT technology, which measures post-ischemic vascular responsiveness following upper arm blood flow occlusion. The device comprises two pneumatic probes for fingertip plethysmograph on both hands that register arterial pulse wave amplitudes (PWAs) at 5-min baseline assessment. In case of concern on bad fit of probe in young children, a quality control took place during the assessment to make sure all subjects’ fingertip could be inserted all the way to the end of the probe. Then the brachial artery of the non-dominant arm is occluded by inflating a sphygmomanometer to supra-systolic pressure for 5 min until cuff release, and another 5-min post-deflation hyperemia is recorded. RHI is automatically calculated as the ratio of average PWAs in 90–150 s post-deflation duration to the baseline average PWAs in the occluded arm compared to the control arm, which reflects pressure changes by the reactive hyperemia. The Framingham reactive hyperemia index (F-RHI) is another automatically outputted PAT index derived from Framingham Heart Studies [13]. It is the natural log transform of the same ratio as RHI, while without a correction of occluded baseline amplitude [0.2276 ∗ ln(mean occluded baseline amplitude) − 0.2], and is using shorter post occlusion times (90–120 s). In addition, we manually calculated the peak response that used the maximum post-deflation PWAs among averages of each 30 s amplitude intervals instead of a fixed post-occlusion duration. Peak response can better reflect the “true” peak hyperemic response of children and adolescents as time-to-peak is delayed in this population compared to that in adults [13]. Higher PAT ratio reflects better endothelial function.

Endo-PAT™ not only measures endothelial function but also assesses arterial stiffness by measuring the peripheral augmentation index (AIx) from the radial pulse wave analysis. AIx indicates the proportion of difference between backward reflected peak (P2) and systolic peak (P1) by P1. Lower AIx reflects better vascular compliance. Because AIx is heart rate related, the result is corrected to a standard of AIx at heart rate of 75BPM (AIx75).

On the same day, BaPWV was measured as another index of arterial stiffness for comparison (automatic waveform analyzer BP-203RPE-I; Colin Medical Technology, Komaki, Japan). Before measurement, four cuffs on both brachia and ankles were fitted with oscillometric sensors [20] with at least 10 min’ rest at supine position in a dedicated, somber and quiet testing room (22 ± 2 °C). The time intervals (T) between the wave fronts of the brachia and those of the ankles were calculated. The distance (L) between the heart and sampling points was calculated automatically according to the subjects’ height. BaPWV is calculated using the following formula: BaPWV = (La-Lb) /T, where ‘La’ is the path length from the heart to the ankle and ‘Lb’ is the path length from the heart to the brachium.

Two trained and experienced operators completed the vascular assessment blinded to other clinical data and did not participant in the following analysis and manuscript preparation.

Characteristics of study population were summarized in all and by obesity status. Student’s t tests and chi-square tests were performed for group comparisons in continuous and categorical variables. We summarized age-dependent trends of Endo-PAT parameters (RHI, F-RHI, Peak response, AIx75) in all and by BMI groups, using two-way fractional polynomial prediction plots. Because of few students with obesity especially in girls, groups with overweight and obesity were combined in order to generate a smoother function of growth with age.

We performed two correlation analysis of Endo-PAT parameters with age, adiposity measures (weight, height, BMI, WC, and body fat mass) and cardiometabolic indicators (blood pressure, glycemic and lipid profiles). First was bivariate correlation by spearman method. Second was generated independent of baseline somatic growth, by using fractional polynomial regression analysis.

Fractional polynomials have been validated to predict non-linear growth trajectories [21]. Each Endo-PAT parameter was modelled by function of age, where height, gender and baseline PWAs of occluded arm were covariates. The model deviance of powers was chosen from the set {− 2, − 1, − 0.5, 0, 0.5, 1, 2, 3} (a power of zero is the log function) to identify the best-fitted fractional polynomial. Fractional polynomial was generated by “fp” function in STATA 15.1 SE (Stata, College Station, TX). By default, “fp” will fit degree-2 factional polynomial (FP2) model and compared it with degree-1(FP1) model by χ ₂ distribution with 2 degrees of freedom. If FP2 significantly improves the model and higher-degree models do not further alter the model, we accept FP2 model otherwise we instead it of FP1 model. Using RHI as an example, FP1 and FP2 model could be formulated as [21]:

p: power.

β: fixed coefficients describing the average shape of the trajectory.

μ: deviation of trajectory from average.

After fitting the power(s) for each model, individual-specific occasion level residuals, representing the deviation from fitted trajectory, were used for the second correlation analysis. The results were therefore modelled against baseline somatic growth, of which we decided were age, gender, height and baseline PWA. Of note is that, the abovementioned correlation analysis was also performed in BaPWV, in comparison with results of Endo-PAT parameters. As vascular response with growth and metabolic profiles might perform differently in children and adolescents with severely high BMI [10, 22], we repeated the above correlation analysis limited in students with obesity.

Characteristics of 545 subjects were presented in all and by groups of non-OB (non-overweight or -obese, n = 392), overweight (n = 80) and obesity (n = 73) in Table 1. Mean age was 12(±3) with a range of 7–17 years. 296(54.3%) were males and the proportion increased in group of overweight (58.8%) and obesity (78.1%). Thirty-seven of seventy-three subjects with obesity had BMI SD score ≥ 3. Students with overweight and obesity had higher WC, body fat mass, blood pressure levels, glycemic levels (fasting glucose, insulin levels, HOMA-IR), TC, TG, LDL-C and lower HDL-C than non-OB group. BaPWV(p = 0.001) and baseline PWA(p < 0.001) were significantly higher in students with overweight and obesity. Despite lack of significance, F-RHI, peak response and Aix75 were lower, and RHI was higher in individuals with overweight and obesity.

Figure 1 depicts the age-dependent trends of Endo-PAT parameters. Overall, the trends showed that RHI, F-RHI and peak response were higher and AIx75 were lower in elder students in both genders. The trends appeared to slow down during adolescence, especially in students over 15 years, compared to lower-grade children. When stratified by groups, students with overweight and obesity showed similar trends as non-OB students.

All parameters were significantly correlated with age, weight, height, WC and BMI in all (all p < 0.001). The correlations were also shown as significant when limited in subjects with obesity, except for height with BaPWV (Table 2). RHI, F-RHI, peak response and BaPWV were positively correlated, while AIx75 showed negative correlation with the abovementioned indicators with moderate magnitude. Stronger correlations were observed in students with obesity. RHI, F-RHI and peak response were not correlated, while AIx75 (r = 0.154 and 0.573 in all and obese only) and BaPWV (r = 0.109 and 0.408 in all and obese only) showed significant correlation with body fat mass (p < 0.05), with stronger correlations in students with obesity.

The correlations with blood pressure, glycemic and lipid profiles were relatively weak and mixed in all parameters (Table 2). RHI, F-RHI and peak response were not correlated with SBP, but showed inverse correlation with DBP (p < 0.05). In contrast, AIx75 and BaPWV were significantly correlated with SBP (p < 0.05), and BaPWV was also correlated with DBP (r = 0.316 p < 0.001 in all and r = 0.198 in obese only). Notably, all parameters except AIx75 were significantly positively correlated with TG (RHI: r = 0.150; F-RHI: r = 0.105; peak response: r = 0.089; BaPWV: r = 0.168), and the directions of correlations in RHI, F-RHI, peak response were reversed in students with obesity (RHI: r = − 0.225; F-RHI: r = − 0.193; peak response: r = − 0.155).

As shown in Table 3, the best-fitted curved powers for RHI, F-RHI, peak response, AIx75 and BaPWV was − 0.5, − 0.5, − 0.5, − 2 and {0,0.5} correspondingly. An overall 14, 13 and 19% of variance in RHI, F-RHI and BaPWV could be explained by age, height, gender and baseline PWA. Twenty-four and thirty-nine percent of variance was contributed by the same covariates for peak response and F-RHI respectively. Age was the strongest determinant of RHI (p = 0.001), F-RHI (p < 0.001), peak response (p = 0.010) and BaPWV (p < 0.001). Gender (p < 0.001) and baseline PWA (p = 0.008) were also significant predictors for BaPWV. AIx75 was primarily explained by height and gender (p < 0.001).

After accounting for age, height, gender and baseline PWA, the correlations with adiposity measures dropped greatly in all parameters (Table 4). RHI, F-RHI and peak response were only significantly correlated with DBP (RHI: r = − 0.132, p = 0.002; F-RHI: r = − 0.204, p < 0.001, peak response: r = − 0.176 p < 0.001). AIx75 was not correlated with any indicators in all, but with body fat mass in students with obesity (r = 0.523, p = 0.007). None of Endo-PAT parameters was correlated with glycemic and lipid profiles, except an inverse correlation observed between RHI and TG in students with obesity (r = − 0.322, p = 0.008). Conversely, although the magnitude decreased, BaPWV was significantly correlated with all adiposity measures and most cardiometabolic indicators (insulin, HOMA-IR, TG, HDL-C, LDL-C), especially for body fat mass, blood pressure and insulin levels. The correlations with cardiometabolic indicators became stronger after adjusting for baseline somatic growth, and consistent in students with obesity.

To our knowledge, we are the first to build parallel comparisons of alternate PAT algorithms, as well as characteristics of endothelial function and arterial stiffness by Endo-PAT technique in the same young population. This has not been well explored before, and we controlled for somatic growth using effective modelling. Fractional polynomials help detect non-linear trajectories in growth, while the coefficients are not as interpretable as linear models [21]. This study helps narrow the research gap by demonstrating a bigger picture of Endo-PAT value for identifying cardiometabolic risk factors in very early of life. However, we acknowledged that some limitations existed that might affect the results. First was insufficient size by age and gender (Additional file 1: Table S2). Although we reached overall power and pre-specified size for each age, the response rate in certain age was extremely low due to academic pressure during graduation. Besides, few biochemical data were not available in all subjects. These might contribute to the inaccuracy of findings due to limited power calculation in some analysis. However, we believe this might not easily undermine our main conclusion since the whole picture showed overall consistent tendency across different parameters. Secondly, as sex steroids made different contribution on vascular reactivity and endothelial function during puberty [11, 23], puberty stage assessment is preferred to increase validity which was not covered in this study. Our study was cross-sectionally sampled in a certain area. Thus, we highlighted the value of longitudinal study in future investigation, as it exactly deciphers the timing of maturation and pathophysiological characteristics within an individual. Finally, it is noteworthy that there is no established norm of Endo-PAT use in paediatric studies so far [13, 14, 25, 26]. RHI of < 1.67 or 1.35 for adults might mistakenly classify a large proportion of children as dysfunction [25, 43] and analysis in relation to thresholds of Endo-PAT parameters was not considered in this study. Rather, our results raised the possibility that the Endo-PAT system itself was still beyond to be widely applied in general young population, and this could not be easily resolved by alternate PAT algorithms or data normalization.