Cardiometabolic Health and Carotid-Femoral Pulse Wave Velocity in Children: A Systematic Review and Meta-Regression

Abstract

OBJECTIVE:

To identify, in children the normal rate of carotid-femoral pulse wave velocity (cfPWV) progression, and whether presence of cardiometabolic risk factors is associated with cfPWV.

STUDY DESIGN:

Electronic databases (PubMed, Google Scholar) were searched from inception to May 2018, for all studies which reported cfPWV in children (<19 y). Random effects meta-regression quantified the association between time (y) and cfPWV, and a systematic review was performed to determine whether cardiometabolic risk factors are associated with cfPWV.

RESULTS:

Data from 28 articles were eligible for inclusion, including 9 reference value (n=13,100), 5 cardiovascular risk (n=5,257), 10 metabolic risk (n=2,999), and 8 obesity-focused (n=8,760) studies. Meta-regression findings (9 studies): the increase in cfPWV per year (age) was 0.12 (95%CI: 0.07, 0.16) m/s per year, and when stratified by sex the confidence intervals overlapped. Systematic review findings: Cardiometabolic risk factors were positively associated with cfPWV, including positive associations with blood pressure, impaired glucose metabolism, and metabolic syndrome. However, obesity was not consistently associated with cfPWV.

CONCLUSION:

Arterial stiffness in children progresses with age and is associated with cardiometabolic risk factors. Although further longitudinal studies are warranted, the presented reference data will be valuable to epidemiologists tracking children, and to scientists and clinicians prescribing therapies to mitigate risk in a population which is increasingly more vulnerable to cardiovascular disease.

Compared with chronological age, the concept of biological age is a proposed measure that would more accurately reflect structural and functional changes taking place in the body as it ages and predict health outcomes.^1,2 The biological age of the vasculature, or vascular aging, is most commonly estimated by measuring arterial stiffness². Arterial stiffening occurs along the length of the arterial tree, but is most evident in the aorto-illiac pathway^3,4. The elastic aorta is directly proximal to the heart and is responsible for dampening the speed and amplitude of retrograde pressure waves that increase the heart’s workload during systole⁵. Arterial stiffness can be estimated by measuring the transit time of the forward-traveling pulse wave between two arterial sites, or pulse wave velocity (PWV)⁶. Using the PWV approach, central aortic arterial stiffness can be estimated through measurement of blood pressure waveforms at the carotid and femoral (cf) arteries.

We conducted a systematic review and meta-regression to identify normal vascular aging, classified as cfPWV progression. Owing to heterogeneity in the literature, only a systematic review was performed to determine whether cardiometabolic risk factors are associated with cfPWV.

METHODS

This systematic review and meta-analysis were carried out in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines⁷.

Two authors searched electronic databases (PubMed and Google Scholar) using the following Boolean operators: (children[Title/Abstract] OR child[Title/Abstract]OR youth[Title/Abstract] OR adolescent[Title/Abstract] OR adolescence[Title/Abstract] or youth[Title/Abstract]) AND (pulse wave velocity[Title/Abstract] or PWV[Title/Abstract]). The reference lists of all identified articles and relevant reviews were also examined. The search was limited to articles published in English from database inception to May 05, 2018.

We screened article titles and abstracts for relevance. The full-texts of potentially eligible articles were obtained to review for eligibility based on the following inclusion criteria) the outcome was cfPWV, the subjects were aged ≤18 y, (at least 100 subjects were included within each experimental or observational group, and) articles were published in English. Eligibility was not limited by study design or comparison group. Two authors completed the study selection, with consultation from a third author in the case of discrepancies.

Three authors extracted the data from each eligible study. Extracted data included bibliographic information (author, publication year), participant characteristics, details of the study design, and study outcomes. Subsequently, the studies were organized by cardiometabolic factors that may influence vascular aging, ie, those focusing on CVD or those focusing on diabetes. Grouping occurred when at least three studies were available for a given factor.

Studies were organized by reference value studies, and three cardiometabolic factors cardiovascular, () diabetes and metabolic syndrome, and () obesity. One primary author was responsible for the synthesis of data for the reference value studies, and one author for each category of cardiometabolic factor (cardiovascular, diabetes and metabolic syndrome, and obesity). For each study, means and associated standard deviations were entered into a database. When means or standard deviations values were not published, the manuscript author(s) were first contacted, then if no response was provided, the values were estimated based on methods from the Cochrane Handbook for Systematic Reviews of Interventions⁸. For articles reporting multiple time points for a given population, only the final time point was used. Wherever possible data from individual studies are reported as the estimate and associated 95% confidence interval (95% CI). Where the 95% CI has not been provided the significance (P value) is reported. Beta values (β) were the preferred estimate of the association between cardiometabolic risk factors and cfPWV to enable direct comparison with the meta-regression outcomes.

Study quality was assessed using the Effective Public Health Practice Project Quality Assessment Tool (EPHPP). EPHPP lists several 6 criteria and is suitable for systematic reviews combining original research with different study designs⁹. Using the quality criteria, an article is assigned a global rating: 1 (strong), 2 (moderate) or 3 (weak).

Meta-regression was conducted to determine the rate of change in cfPWV with age. For each study, the means and standard deviations for each age group were entered into software designed specifically for meta-analyses (Open Meta-Analyst, version 12.11.14, http://www.cebm.brown.edu/openmeta/). To account for both within- and between-study variability, univariate meta-regression analyses were run as random effects models using the DerSimonian–Laird method¹⁰. Stratified analyses were performed to calculate sex-specific effects. The intercepts (i.e., cfPWV at age 0) and slopes (i.e., change in cfPWV per year) are reported in absolute units with 95% CI. Two authors (LS, MM) conducted the data analysis.

RESULTS

Figure 1 depicts the article selection process along with reasoning for exclusions. A total of 378 potentially eligible articles were identified through electronic databases, and no additional articles were identified through manual searches of reference lists. Of the 378 total articles, 323 were excluded after title and abstract screening. Subsequently, 55 full-text articles were acquired and assessed for eligibility, from which 28 met the inclusion criteria. Meta-regression identified the normal rate of vascular aging, and systematic review determined whether cardiometabolic risk factors are associated with cfPWV.

FIGURE 1.

Study Selection Process

Nine studies were identified, including 3 longitudinal^11–13 and 6 cross-sectional studies^14–19 (Table 1I; available at www.jpeds.com). The quality of the studies ranged from 1 (strong) to 3 (week), with a median score of 1. The baseline age for the longitudinal studies ranged from 14.5 – 15.1 y, with a 2–5 y follow-up. The age range across the cross-sectional studies was 6–18 y. Across the longitudinal studies, cfPWV increased in children at the rate of 0.15–0.25 m/s per year, which is in-line with the 0.08–0.18 m/s per year increase observed in the 6 cross-sectional studies.

TABLE 1.

Summary of childhood aortic pulse wave velocity studies reference studies (meters per second, m/s)

First Author	Yr	Ref	Quality	Country	Study Design	Population	Device	Δ PWV/y
Bjornstad	2015	11	3	USA	Longitudinal, 2 y	n=222 (47% F), 15.1–17.1 y, T1D	SphygmoCor, tonometry	All: 0.25
								F: NR
								M: NR
Chen	2012	40	1	Sweden	Longitudinal, 3 y	n=162 (58% F), 14.5–17.5 y	SphygmoCor, tonometry	All: 0.16
								F: 0.09
								M: 0.27
Dabelea	2013	12	1	USA	Longitudinal, 5 y	n=298 (46% F), 14.5–19.2	SphygmoCor, tonometry	All: 0.15
								F: NR
								M: R
Diaz	2018	14	1	South America	Cross-Sectional	n=586 (77% F), 13.8–17.5 y	Arteriometer, tonometry	All: 0.13
								F: NR
								M: NR
Fischer	2012	15	1	Germany	Cross-Sectional	314 (50% F), 8–17 y	Vicorder, oscillometry	All: 0.08
								F: 0.12
								M: 0.07
Mora-Urda	2017	16	1	Spain	Cross-Sectional	n=350 (n=48% F),8–11 y	SphygmoCor, tonometry	All: 0.18
								F: 0.19
								M: 0.16
Reusz	2010	17	1	Hungary	Cross-Sectional	n=l,008 (51%), 7–19 y	Pulsepen, Tonometry	All: 0.10
								F: 0.10
								M: 0.11
Silva	2016	18	1	Angola	Cross-Sectional	n=157 (61% F), 8–10.5 y	Compilor, Tonometry	All: 0.08
								F: NR
								M: NR
Thurn	2015	19	1	Germany/Turkey	Cross-Sectional	n=l,003 (54% F), 6–18 y	Vicorder, oscillometry	All: 0.11
								F: 0.10
								M: 0.12

Quality: scores were derived from the Effective Public Health Practice Project Quality Assessment Tool (EPHPP), where 1 = strong, 2 = moderate, and 3 = weak.

Abbreviations: F, female; M, male; NR, not reported

The meta-regression results, reported in Figure 2, A, suggest that among children each additional year of age was associated with a 0.12 (95% CI: 0.07, 0.16) m/s increase in cfPWV. To explore linearity of the association, we included and age quadratic term and found it to be statistically non-significant (P = .943). This term was therefore omitted from the regression models. Of the 9 studies included in the meta-regression, 3 studies^11–13 were longitudinal (collective n = 682), and these studies contribute 6 data point points out of 55. If these 3 studies are excluded the increase per year drops from 0.12 (95%CI: 0.7, 0.16) to 0.08 (95%CI: 0.06, 0.11) m/s, indicating that the cross-sectional studies may under-estimate the change in cfPWV with age.

FIGURE 2.

Univariate random-effects meta-regression scatter plots: carotid-femoral pulse wave velocity (cfPWV) change in (a) both sexes (9 studies, n=4,757); females only (5 studies, n=1,567); and (c) males only (5 studies, n=1,367). Each circle represents an age category within an individual study, and the size of the circle is proportional to study weighting.

Abbreviations: CI, confidence interval, m/s, meters per second

Five studies reported independent cfPWV values for males and females^{13,15–17,19}. The data suggest that cfPWV increases 0.07–0.27 m/s per year in males and 0.09–0.19/y in females. The meta-regression outcomes from these 5 studies are reported in Figure 2, B and C. Across both sexes, cfPWV increased at 0.15 (95% CI: 0.11, 0.18) m/s per year (data not shown), for females cfPWV increased at the rate of 0.13 (95% CI: 0.08, 0.18) m/s per year (Figure 2, B), and for males cfPWV increased by 0.16 (95% CI: 0.12, 0.20) m/s per year (Figure 2, C).

We identified 5 CVD-related studies (Table 2; available at www.jpeds.com), all of which were cross-sectional and measured cfPWV with tonometry^20–24. The quality of the studies ranged from 1 (strong) to 2 (moderate), with a median score of 1. Four studies included blood pressure^20,22–24 and one study focused on retinal vessel calibers²¹, as described below.

TABLE 2.

Summary of studies examining the influence of cardiovascular disease risk factors on childhood aortic pulse wave velocity (meters per second, m/s)

First Author	Yr	Ref	Study	Quality	Country	Study Design	Population	Device	Outcome(s)
Batista	2015	20	n/a	2	Brazil	Cross-Sectional	n=231 (50.2% F),9–10 y	Compilor, tonometry	SBP & DBP ↑ associated with higher tertiles of cfPWV; DBP positively associated with cfPWV (β=0.157)
Gishti	2015	21	n/a	1	Netherlands	Cross-Sectional	n=4007, 5–8 y	Complior, tonometry	Retinal venular caliber positively associated with cfPWV (0.04 z-score per z-score in venual caliber. No association with arteriolar caliber
Lurbe	2016	22	n/a	1	Spain	Cross-Sectional	n=415 (46.4% F), 8–18 y	SphygmoCor, tonometry	Compared with normotensives, ↑cfPWV in those with isolated systolic hypertension & systolic-diastolic hypertension
McCloskey	2014	23	n/a	1	Australia	Cross-Sectional	n=289 (49.3% F),7–11 y	SphygmoCor, tonometry	cfPWV positively associated with 10 mmHg Δ in DBP (β=0.30), SBP (β =0.12) and MAP (β=0.26)
Peluso	2017	24	n/a	1	Uruguay	Cross-Sectional	n=315 (44.4% F), 8–18 y	SphygmoCor, tonometry	Children with ≥90th percentile of cSBP or cDBP had ↑ cfPWV compared to those <90th percentile (5.2 vs. 5.0 m/s)

Quality: scores were derived from the Effective Public Health Practice Project Quality Assessment Tool (EPHPP), where 1 = strong, 2 = moderate, and 3 = weak.

Abbreviations: DBP, diastolic blood pressure; cfPWV, carotid-femoral pulse wave velocity; cDBP, central diastolic blood pressure; cSBP, central systolic blood pressure; F, female; MAP, mean arterial blood pressure; SBP, systolic blood pressure; SD, standard deviation

The 4 studies on cfPWV and blood pressure in children consistently show that higher systolic blood pressure (SBP) and diastolic blood pressure (DBP) are associated with higher cfPWV. In 9–10 y old school children, the mean SBP and DBP were higher with increasing tertiles of cfPWV and DBP was positively associated with cfPWV (β=0.16 m/s, 95%CI: 0.03, 0.29) after adjusting for SBP and sex²⁰. Similar results were seen in a twin pair study of children with a mean age of 9 (range 7–11 y)²³. PWV was higher for each 10 mmHg increase in DBP (β=0.30, 95% CI: 0.20, 0.41), SBP (β=0.12 m/s, 95% CI: 0.06, 0.19), and mean arterial pressure (MAP; β=0.26 m/s, 95% CI: 0.17, 0.36) adjusting for age and sex²³. In the within–pair analysis of monozygotic twins, blood pressure was not significantly associated with cfPWV. Conversely, in dizygotic twins, DBP and MAP were significantly associated with cfPWV, suggesting a genetic component to the association of blood pressure and arterial stiffness. In children aged 8–18 y, a study in Uruguay defined high central blood pressure as ≥90^th percentile of central SBP and/or central DBP²⁴. Those with high central blood pressure had a higher mean cfPWV compared with those <90^th percentile (5.22 [SD: 0.89] m/s versus 5.01 [SD: 0.68] m/s, P=0.047). Further, when children 8–18 y were defined as normotensive, isolated systolic hypertensive, and systolic-diastolic hypertensive, the mean cfPWV was higher in those with isolated systolic hypertension (5.70 m/s) and systolic-diastolic hypertension (6.44 m/s) compared with the mean cfPWV of 4.91 m/s in normotensives after adjustment for age and sex²².

We identified one study which evaluated in children the association of cfPWV with retinal vessel calibers, considered to be early markers of microvascular damage. In 4,007 children (median age 6 y), wider retinal venular caliber was associated with higher cfPWV (β=0.04 cfPWV z-score, 95% CI: 0.01, 0.07 per one unit z-score increase in venular caliber), but was no association with arteriolar caliber was observed²¹.

We identified 10 publications (Table 3; available at www.jpeds.com) investigating metabolic abnormalities, with 7 examining the association of diabetes with cfPWV^{11,12,25–29}, 2 the association of indicators of glucose metabolism with cfPWV^23,30, and 1 the association between metabolic syndrome and cfPWV³¹. Only 1 study included follow-up of the study participants¹¹, with the remaining studies cross-sectional. Of the included 7 diabetes studies, with duration of diabetes ranging from 2 to 12.8 y, only 2 studies were not limited to Type 1 diabetes,^30,31 and 5^12,28–31 were conducted in the SEARCH CVD, an ancillary study to the SEARCH For Diabetes in Youth Study. The studies^26,30,31 examining glucose metabolism and the metabolic syndrome study were conducted in general pediatric populations. Quality of the studies ranged from 1 to 2 (mean 1.4).

TABLE 3.

Summary of studies examining the influence of metabolic risk factors on childhood aortic pulse wave velocity (meters per second, m/s)

First Author	Yr	Ref	Study	Quality	Country	Study Design	Follow-up (y)	Population	Diabetes (Type, y)	Device	Outcome(s)
Alman	2014	25	DMDA-T1D	1	USA	Cross-sectional	na	n=190 (43% F), 13–17 y	Type 1, 5 y	SphygmoCor, tonometry	↑ ICH metrics negadvely associated with cfPWV (β=−0.02)
Bjornstad	2015	11	AdDIT	2	USA	Cross-sectional & follow-up	2 y	n=297 (50% F), 12–18 y	Type 1, 8.7 y	SphygmoCor, tonometry	Achieving 4–6 vs 2–3 ISPAD/ADA goals associated with ↓ cfPWV at baseline (5.2 vs. 5.7 m/s) and follow-up (5.7 vs. 6.1 m/s)
	2016	26	Generation XXI	2	Canada	Cross-sectional	na	T1DM: n=199 (51% F): controls: n=178 (54% F), 10–16 y	Type 1, 6.2 y	SphygmoCor, tonometry	Compared to controls, T1D had ↑ SBP (113 ± 10 vs. 110 ± 9 mmHg) and DBP (62 vs. 58 mmHg), cfPWV & (5.3 vs. 5.1 m/s), & FMD (7.5 vs. 6.5 %)
Correia-Costa	2016	30	SEARCH CVD	2	Portugal	Cross-sectional	na	n=315 (47% F),8–9 y	N/	PulseTrace, Doppler	In univariate regression cfPWV was associated with fasting insulin (p=0.13) & HOMA-IR, but not FBG (p=0.03). In multivariate regression, cfPWV was associated with HOMA-IR (β=0.11) & the absence of dipping (β=0.16)
Dabelea	103	12	SEARCH Diabetes	1	USA	Cross-sectional & follow-up	5 y	n=298 (% 47F), 14,5 (2.8) y at baseline	Type 1, 4.8 y	SphygmoCor, tonometry	Baseline presence of MetS (P = 0.004) but not ↑ HBA1c (P=0.460) associated with ↑ cfPWV over time. ↑ HbA1c over time associated with ↑ cfPWV over time (P=0.04)
Lamichhane	2014	27	n/a	1	USA	cross-sectional	na	n=237, >10 y	Type 1, >3 months	SphygmoCor, tonometry	High intake of sugar sweetened beverages, diet soda, eggs, potatoes and high-fat meats, sweets/deserts, and highfat dairy was positively associated with cfPWV (β=0.01 per unit of a dietary score)
McCloskey	2014	23	n/a	1	Tasmania	Cross-Sectional	na	n=289 (49.3% F), 7–11 y	N/A	SphygmoCor, tonometry	cfPWV positively associated with FBG (β=0.25), insulin (β=0.02), insulin/glucose ratio (β=0.12); HOMA-IR (β=0.10), C-peptide (β=0.53)
Pandit	2011	31	SEARCH Diabetes	2	India	Cross-sectional	na	n=236 (51% F), 6–17 y	NA	Prosound alpha-10, Ultrasound	↑ cfPWV in those with Mets >−0.8 vs. <0.8 cutpoint of continuous MetS score (4.3 vs. 3.7 m/s)
Shah	2012	29	SEARCH Diabetes	1	USA	Cross-sectional	na	n=225 (47% F), mean age 14.7 y	Type 1, 4.8 y	SphygmoCor, Tonometry	No correlation between adiponectin & cfPWV and (r = 0.08)
Urbina	2010	28	SEARCH CVD	1	USA	Cross-sectional	na	T1DM: n=535 (47% F) mean age 14.6 y; controls: n=241 (58% F), mean age 17.8 y	Type 1, 5.7 y	SphygmoCor, tonometry	Age-adjusted cfPWV ↑ in T1DM vs. controls (5.4 vs 5.1 m/s, P=<0.001).

Quality: scores were derived from the Effective Public Health Practice Project Quality Assessment Tool (EPHPP), where 1 = strong, 2 = moderate, and 3 = weak.

Abbreviations: AdDIT, adolescent type 1 diabetes cardio-renal intervention trial; cfPWV, carotid-femoral pulse wave velocity; DBP, diastolic blood pressure; DMDA-T1D; determinants of macrovascular disease in adolescents with type 1 diabetes; F, female; FBG, fasting blood glucose; HbA1C, Glycated hemoglobin; HOMA-IR, Homeostatic model assessment of insulin resistance; ICH, ideal cardiovascular health; ISPAD, International Society for Pediatric and Adolescent Diabetes; METs, metabolic syndrome; SBP, systolic blood pressure; T1DM,Type 1 diabetes mellitus

Pulse wave velocity was found to be greater among children and adolescents with diabetes as compared with that observed among normoglycemic controls. However, in cross-sectional analyses, the difference in cfPWV between children with Type 1 diabetes and those without was not statistically significant (5.25 [SD: 0.75] vs. 5.10 [SD: 0.87] m/s)²⁶. Increased peripheral, rather than central arterial stiffness in children with diabetes as compared with normoglycemic controls was observed in the SEARCH CVD Study²⁸.

Good glycemic control, assessed as adherence to 4–6 (as compared with 2–3) International Society for Pediatric and Adolescent Diabetes/American Diabetes Association goals was associated with lower cfPWV¹¹. Among children with Type 1 diabetes, boys were observed to have greater cfPWV as compared with girls²⁸. In the SEARCH CVD study of children with Type 1 diabetes, male sex together with age, obesity, and blood pressure, was found to be a consistent predictor of increases in arterial stiffness^28,29.

The 2 studies^23,30 conducted in the general pediatric population suggest that abnormalities of glucose metabolism show a positive association with cfPWV, even in the absence of clinically diagnosed diabetes. Correia-Costa et al observed a modest correlation (r=0.12) of HOMA-IR (Homeostatic Model Assessment of Insulin Resistance) with cfPWV among children aged 8–9 y without clinically diagnosed diabetes³⁰. In 7–11 year-old children, McClosky et al observed that several markers of insulin resistance were associated with cfPWV, including: glucose (β=0.25 m/s, 95% CI: 0.00,0.50); insulin (β=0.02 m/s, 95% CI: 0.01, 0.04); insulin/glucose ratio (β=0.12 m/s, 95% CI: 0.04, −0.21); HOMA-IR (β=0.10 m/s, 95% CI: 0.03, 0.17); C-peptide (β=0.53 m/s, 95% CI: 0.17, 0.89)²³.

In an analysis based on a continuous metabolic syndrome score among children aged 11.4 (SD 2.8) y, Pandit et al observed a higher average cfPWV (4.3 [SD: 0.6] m/s) among those with a high metabolic score as compared with respective values (3.7 [SD: 0.5] m/s) observed in children with a low metabolic score³¹. In this study, cfPWV increased with the increase in the number of metabolic syndrome components. A similar observation of greater PWV among children with metabolic syndrome as compared with those without was reported by Dabelea et al¹².

Eight studies were identified (Table 4; available at www.jpeds.com)^30,32–38 that included obese children, all of which were cross-sectional. The quality of the studies ranged from 1 (strong) to 2 (moderate), with a median score of 1. All studies measured cfPWV with tonometry, 7 studies measured body composition using anthropological methods^{30,32–34,36,38,39}, 4 using dual-energy X-ray absorptiometry^32,33,37,38, and 2 using bio-impedance analysis^30,35.

TABLE 4.

Summary of studies examining the influence of obesity on childhood aortic pulse wave velocity (meters per second, m/s)

First Author	Ref	Yr	Study	Quality	Country	Study Design	Population	Device	Body Comp.	Outcome(s)
Arnberg	32	2012	n/a	1	Denmark	Cross-Sectional	n=182 (62% F), 12–15 y, all overweight	SphygmoCor, tonometry,	Anthro., DXA	Mean cfPWV was 4.8 (0.72) m/s. cfPWV positively associated with BMI (β=0.05), WC (β=0.02), AF:GF (β=1.5) & AF:BF (β=10.3), but not BF%.
Charakida	33	2012	ALSPAC	1	UK	Cross-Sectional	n=6,576 (51% F), 10–11 y, normal-weight & obese	SphygmoCor, tonometry,	Anthro., DXA	Mean cfPWV was 7.7 (1.2) m/s for normal-weight and 7.0 (1.0) m/s for obese (P=<0.001). cfPWV negatively associated with BMI (β=−0.077, WHtR (β=−3.9), BF% (β=−0.02) & trunk fat% (β=−0.03).
Correia-Costa	30	2016	Generation XXI	1	Portugal	Cross-Sectional	n=315 (%47 F), 8–9 y, normal-weight & obese	PulseTrace, Doppler	Anthro., BIA	Median cfPWV in was 5.0 (P25-P75: 4.6–5.2) m/s for normal-weight and 5.1 (4.8–5.5) m/s for obese (P=<0.001). cfPWV positively associated WhtR (ρ=0.13) and BF% (ρ=0.21), but not zBMI (ρ=0.10, P=0.070)
Hanvey	34	2017	PEAS KGS	1	AU	Cross-sectional	n=187 (55% F),14 y, normal-weight & overweight	SphygmoCor, tonometry,	Anthro.	Mean cfPWV was 4.6 (0.7) m/s in normal-weight and 4.7 (0.7) m/s in overweight (P=0.97). BMI growth trajectory (birth-adolescence) was not significantly associated with cfPWV.
	35	2017	HELP	1	UK	Cross-Sectional	n=174 (63% F), 12–19 y, all obese	SphygmoCor, tonometry,	Anthro., BIA	Mean cfPWV was 7.1 (1.2) m/s. cfPWV positively associated with zBMI (β=0.49), FMI (BIA, β=0.05), zWC (β=0.26), & SAD (β=0.05).
Lurbe	36	2012	n/a	2	Spain	Cross-Sectional	n=501(47% F), 8–18 y	SphygmoCor, tonometry,	Anthro.	cfPWV was 5.7 (1.6) m/s in normal-weight and 4.9 (1.0) in obese (P=<0.001). cfPWV negatively associated with zBMI (β=−0.193)
Ryder	37	2016	n/a	2	USA	Cross-Sectional	n=252 (52% F), 8–20 y, all overweight	SphygmoCor, tonometry,	DXA	Mean cfPWV was 7.0 (1.2) m/s. cfPWV was not significant associated with BF%
Sakuragi	38	2009	looks	1	AU	Cross-Sectional	n=573 (49% F), 9–10 y, low BF% & high BF%	SphygmoCor, tonometry,	Anthro., DXA	cfPWV was 4.4 (0.5) m/s in low BF% and 4.6 (0.5) m/s in high BF%. cfPWV positively associated with BF% (β=0.022), BMI (β=0.054), & WC (β=0.023)

Quality: scores were derived from the Effective Public Health Practice Project Quality Assessment Tool (EPHPP), where 1 = strong, 2 = moderate, and 3 = weak.

Abbreviations: AF:BF, android to body fat ratio; AF:GF, android to gynoid fat ratio; ALSPAC, Avon Longitudinal Study of Parents and Children; BIA, bio-impedance analysis; BF%, body fat percentage; BMI, body mass index; cfPWV, carotid-femoral pulse wave velocity (cfPWV); F, female; FFM, fat free mass; FMI, Fat Mass Index; HELP; Healthy Eating and Lifestyle Trial (HELP); LoOKS, Lifestyle of Our Kids Study; PEAS KGS; Parent Education and Support (PEAS) Program Kids Growth Study; SAD, sagittal abdominal dimension; USA, United States of America; WC, waist circumference; WHtR, waist: height

Three studies reported a positive association between cfPWV and body mass index (BMI) (β range=0.05 to 0.44 m/s)^32,35,38, 2 a negative association with BMI (β range = −0.08 to 0.19)^33,36, and 2 no association with BMI^30,34. Three studies did report a positive association between cfPWV and waist circumference (β range = 0.02 to 0.27)^32,33,38, and 2 a positive association between cfPWV and waist: height (β = 3.9, P=<0.001 and P=0.13, P=0.019)^33,38. Two studies reported a positive association with body fat percentage (β=0.02, 95% CI: 0.02, 0.03³⁸, and p=0.21, P=<0.001³⁰), 1 reported a negative association (β=−0.02, P=<0.001)³³, and 2 reported no association^32,37. One study reported a negative association with trunk fat percentage (β=−0.03, P=<0.001)³³, and 1 study reported a positive association between cfPWV and fat mass index (β=0.05, 95% CI: 0.01, 0.10)³⁵, sagittal abdominal dimension (β=0.05, 95% CI: 0.01, 0.10)³⁵, android to body fat ratio (β=10.3, 95%CI: 0.10, 20.5)³², and android to gynoid fat ratio (β=1.49, 95% CI: 0.24, 2.75)³².

DISCUSSION

Meta-regression analysis of 9 trials suggests that central arterial stiffness, approximated through measurements of cfPWV, increases in children by 0.12 m/s per year. The strength of these findings is supported by the high quality of the included trials. Additionally, seven of the trials measured cfPWV using the “gold-standard” tonometry^{11,14–19,40}, 3 of the trials were longitudinal^11,12,40, and 8 of the trials examined healthy children^{11,14–19,40}. It should be noted that 1 of the longitudinal studies examined children with Type 1 diabetes¹²; however, the cfPWV progression estimate (0.15 m/s per year) for this study was in agreement with the other 2 longitudinal trials (0.15–0.25 m/s per year) and the cross-sectional studies (0.08–0.18 m/s per year). Collectively, these findings are consistent with a recent review of longitudinal adult-based studies², which suggests a 0.2–0.7 m/s increase in cfPWV per 5 y in adults.

Subgroup analysis was conducted on the 5 studies reporting cfPWV values for both boys and girls. The findings suggest that cfPWV increases 0.13 (95% CI: 0.08, 0.18) m/s per year in girls and 0.16 (95% CI: 0.12, 0.20) m/s per year in boys. However, considering the wide overlap in confidence intervals, we cannot conclude that the progression of cfPWV is differential by sex. A recent review did report that cfPWV appears to increases at a greater rate in adult men versus women, but the differences are more pronounced at ages greater than 60 y of age². It should be noted that the recent review of studies in adults incorporated only longitudinal studies², whereas 4 of the 5 studies included in the current sub-group analysis were cross-sectional. A systematic review of cross-sectional studies in adults reported that only 15 of 54 showed higher PWV in men⁴¹. The one longitudinal study included in the current analysis did report a greater rate of cfPWV progression for boys (0.27 m/s per year) versus girls (0.16 m/s per year), but this finding is based on two assessments conducted over a 3 y follow-up period. Longer-duration longitudinal studies are required to confirm the findings of the current study, including whether sex differences do exist.

In this review, studies of CVD-related factors consistently show an association between blood pressure and cfPWV. The 4 studies of blood pressure in children showed an association of higher SBP and DBP with higher cfPWV^20,22–24. In adults, reports of the temporal association of blood pressure and central PWV are conflicting. Blood pressure has been associated with central PWV in cross-sectional^41,42 and longitudinal studies^43–47. However, central PWV has also been reported to be a determinant of increases in blood pressure^48–51. Thus, the association between elevated blood pressure and central arterial stiffening may be bidirectional^52,53. All of the studies in this review were cross sectional, thus longitudinal studies are needed to determine the temporal associations of blood pressure and central arterial stiffness in children and their effects on long-term health.

Findings from the current review suggest a positive association of abnormalities in glucose metabolism with cfPWV. These findings are consistent with observations from studies of adults, which have shown that cardiometabolic factors, including impaired glucose homeostasis and the metabolic syndrome are associated with PWV progression². Of interest, one study reported that, when compared with normoglycemic controls, increased peripheral PWV (leg or arm) was more common than increased cfPWV in subjects with Type 1 diabetes²⁸. However, findings from a study by Bjornstad et al and from the SEARCH CVD study²⁵ do suggest that lowering of blood glucose levels through adherence to diet, exercise, and therapeutic regimen among children with diabetes is associated with a reduction in cfPWV.¹¹ These findings underscore the critical importance of good glycemic control to cardiovascular health.

Studies by Pandit et al and Dabelea et al which examined the association of metabolic syndrome with cfPWV, provided a broader perspective, suggesting an increase in cfPWV with an increase in the number of metabolic abnormalities in children and adolescents, thus further emphasizing the need for control of conditions such as hyperlipidemia and insulin resistance at an early age.^{12, 31}

Our findings suggest a mixed association between body composition and cfPWV in obese children. Although 4 studies reported a positive association between indices of body composition and cfPWV^30,32,35,38, 2 reported a negative association^33,36 and 4 reported no association^30,32,34,37. It should be recognized than these studies incorporated a range of body composition; however, limiting the findings to studies which used DXA to measure body fat percentage, 1 reported a positive association^32,38, 1 a negative association³³, and 2 no association³⁷. The one study reporting a positive association found that each 1% increase in body fat is associated with 0.022 m/s (95% CI: 0.016, 0.029) increase in cfPWV; according to our meta-regression finding, a ~5.5% increase in body fat would equate to a cfPWV being accelerated by 1 y. However, one study reported a similar (β = −0.02 m/s, P=<0.001), albeit negative association.

Our findings are partially contrary to a previous meta-analysis of 14 studies, which reported that obese children had a 0.45 m/s greater PWV than non-obese children³⁹. However, the studies incorporated in the previous meta-analysis utilized a range of PWV techniques, including carotid-radial, carotid-femoral, and carotid-only PWV. Further, the differences between groups were non-significant when only studies with low to medium risk of bias were included, where bias was determined using the Cochrane risk of bias tool.

Limitations and strengths of this study should be addressed to contextualize the findings, and to indicate areas of need to evolve the field. Most importantly, all but 3 studies included in this review were cross-sectional, limiting the capacity to infer causality. In our estimation of change in cfPWV with age, 6 cross-sectional reference studies included children in the age range 6 – 18 y and the 3 longitudinal studies included children with baseline age of 14.5 – 15.1 y, with a 2–5 y follow-up. Although our analysis of available cross-sectional and longitudinal data assumes linearity in the association of age with cfPWV, we are unable to confirm whether this is true across all ages. Additionally, although the majority of studies used the tonometric-based AtCor SphygmoCor device to measure cfPWV, several studies did use alternative devices, which may limit the capacity to directly compare studies. Lastly, the number of studies per cardiometabolic risk factor was small, particular for cardiovascular risk factors; therefore, the topic of investigation was heterogeneous.

Several strengths of this study should be mentioned. To ensure robust estimates, the inclusion criteria limited studies to >100 children per group. The quality of the included studies was generally high. Although it is acknowledged that longitudinal trials of sufficient duration are required to support the findings of this study (Table 5), there are sufficient data to support the use of cfPWV in tracking the progression of CVD risk in children and targeting cardiometabolic risk factors.

TABLE 5.

Summary of findings from the systematic review and meta-regression

What did we know prior to this study?

cfPWV is widely used to reflect aortic arterial stiffness, and can be used to track ‘vascular aging’ In adults, cfPWV naturally increases 0.2 to 0.7 m/s per 5 years In adults, cfPWV increases at a greater rate in those with cardiometabolic diseases

What didn’t we know prior to this study?

In children, the rate at which cfPWV increases has not been summarized In children, the effects of cardiometabolic risk on cfPWV have not been summarized

What does this study add?

In children, cfPWV increases 0.12 (95% CI: 0.07, 0.20) m/s per year The relationship between cfPWV and body composition/obesity is mixed High blood pressure, impaired glucose metabolism, and the metabolic syndrome are positively associated with cfPWV

How do we use this new information?

Findings from this study can serve as a motivation for further research into the association of subclinical disease in association with cardiometabolic risk in children Clinicians and public health practitioners can use these findings to promote healthy lifestyle in children and prevention of clinically overt cardiometabolic diseases

What needs to happen next to move the field forward?

Conduct longitudinal assessment of change in cfPWV and its risk factors in children and adolescents

Quality: scores were derived from the Effective Public Health Practice Project Quality Assessment Tool (EPHPP), where 1 = strong, 2 = moderate, and 3 = weak.

Abbreviations: cfPWV, carotid-femoral pulse wave velocity (cfPWV)

ABBREVIATIONS

BMI

body mass index

CVD

cardiovascular disease

DBP

diastolic blood pressure

PWV

pulse wave velocity

EPHPP

Effective Public Health Practice Project Quality Assessment Tool

HOMA-IR

Homeostatic Model Assessment of Insulin Resistance

MAP

mean arterial pressure m/s per year: meters per second per year

PWV

pulse wave velocity

SBP

systolic blood pressure