Carotid artery intima-media thickness (CIMT), a marker of atherosclerosis, is increased in youth at risk for future cardiovascular disease. Some pediatric studies have used CIMT as a primary outcome in clinical trials, yet data are limited on the standardization of methodology in children. The goal of this study was to evaluate reproducibility of CIMT measurements using two different measurement techniques.
Carotid artery ultrasound studies of children and adolescents obtained as a component of a research study in Kawasaki syndrome were retrospectively analyzed. The CIMTs of both common carotid arteries (CCAs) were measured by one of two sonographers at the time in the cardiac cycle when resolution subjectively was determined to be optimal (Opt-CIMT). These sonographers blindly remeasured a random sample of studies of their own and each other’s, using the same method. Another observer made CIMT measurements using exclusively frames on the R wave (R-CIMT). A fourth observer independently measured a random sample of studies twice with the R-CIMT method.
Carotid artery images from 184 subjects (mean age, 14.7 ± 2.2 years) were analyzed. The intraclass correlation coefficient for interobserver variability was 0.86 (95% confidence interval [CI], 0.69–0.94) compared with 0.85 (95% CI, 0.65–0.93) for the right and 0.86 (95% CI, 0.67–0.94) versus 0.95 (95% CI, 0.87–0.98) for the left CCA for Opt-CIMT and R-CIMT, respectively. R-CIMT was significantly thicker than Opt-CIMT (right CCA, 0.439 ± 0.030 vs 0.428 ± 0.024 mm, P < .001; left CCA, 0.446 ± 0.030 vs 0.434 ± 0.025 mm, P < .001).
Pediatric CIMT measurements have excellent reproducibility when the same methodology is applied but vary significantly throughout the cardiac cycle. This report highlights the need to standardize CIMT measurements in the youth and supports the use of electrocardiographic timing, as recommended in adults, in pediatric longitudinal studies.
Carotid artery intima-media thickness (CIMT) has been used as a noninvasive modality to evaluate the presence of atherosclerosis since the early 1990s. In adults, increased CIMT is associated with coronary artery disease and is predictive of future cardiovascular events, including stroke and myocardial infarction. CIMT is robust and reproducible in the evaluation of changes over time to serve as an end point in clinical trials assessing the impact of antihypertensive and lipid-lowering medications on cardiovascular risk in adults. To evaluate early, subclinical disease, assessment of CIMT also has been used in children and young adults with known risk factors for cardiovascular disease, including hypercholesterolemia, hypertension, obesity, and type 1 diabetes mellitus. Some pediatric clinical trials have also used CIMT as a primary outcome measure. Despite the clear value of this tool in the assessment of cardiovascular risk in high-risk children and adolescents, its application has been restricted by a number of factors, including variable protocols for data acquisition and analysis and limited data on reproducibility. Different methods used for research purposes limit the ability to make comparisons and generalizations of reported findings. Furthermore, the magnitude of the difference in CIMT during the cardiac cycle can affect risk stratification in asymptomatic individuals because of a mismatch between acquisition protocol and CIMT normative data, as shown in a recent adult study. Another important aspect in the pediatric population is image resolution, because the CIMT values are lower overall. More recent guidelines recommend measurement of CIMT on the R wave for standardization purposes.
The aim of this study was to evaluate the reproducibility and the impact of the timing of measurement within the cardiac cycle on CIMT values in the pediatric population. Precise and reliable noninvasive testing for atherosclerosis in youth will improve the ability to examine cardiovascular risk.
The present study was a retrospective review of data gathered from a subgroup of children and young subjects (ages 11–29 years) who were enrolled in a prospective Kawasaki disease research study conducted between 2007 and 2011 at Boston Children’s Hospital. Subjects either had history of Kawasaki disease or were healthy volunteers. All subjects underwent carotid artery ultrasound performed by one of the two sonographers trained for the study as part of the study protocol designed in 2006 and 2007. Subjects were included in the present retrospective study if they were in the pediatric group (11–19 years of age) at the time of carotid artery ultrasound testing.
Height and weight; systolic, diastolic, and mean blood pressure; and fasting lipid profile, including total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and triglycerides, were also obtained as part of the study protocol. Body mass index and body surface area were calculated. Blood pressure Z scores (Boston Children’s Hospital normative database) and body mass index Z scores (Centers for Disease Control and Prevention) were calculated. This study was approved by the institutional review board of Boston Children’s Hospital.
Carotid Artery Ultrasound and Measurements
A linear array probe (L11-3 MHz, iE33 Ultrasound System; Philips Medical Systems, Andover, MA) was used to obtain three 10-sec loops taken at the anterior, posterior, and lateral angles on the right and left common carotid arteries (CCAs). We did not use a Meijer’s arc, which is a tool used to standardize angles of insonation for carotid ultrasound studies, and therefore referred to these angles as anterior, posterior, and lateral angulations. However, these angles would corroborate to approximately 60°, 105°, and 150° for the right and 300°, 255°, and 210° for the left CCA on the arc. The images were stored as Digital Imaging and Communications in Medicine loops. The resolution was 1,024 × 768 pixels/cm. In general, gain and compress adjustments were made on default vascular settings (P mode, no harmonics) to optimize the images. The frame rate was kept at >50 frames/sec and persistence at low. The CIMT of the far wall of the CCA was measured online by one of the two sonographers (S.T. or S.O.) using a frame from each of the loops with anterior, posterior, and lateral angulations over the length of a 1 cm segment located at the distal aspect of the right and left CCA just proximal to the bifurcation into the internal and external carotid arteries, using commercially available, semiautomated edge detection software (QLAB; Philips Medical Systems). The sonographer made the CIMT measurements on the frame in which he or she subjectively noted the resolution to be optimal, regardless of the timing within the cardiac cycle ( Figure 1 A). The measurements were accepted only if the edge detection software was able to automatically measure 95% of the 1-cm segment. As per original prospective study protocol, after 3 to 9 months of study acquisition and online measurements, the sonographers were blinded to their original CIMT measurements and frame choice for the measurements and asked to remeasure CIMT in a random subset of their original studies and the other sonographer’s studies, offline, using the same “optimal resolution” method they had previously used (QLAB) for intra- and interobserver variability ( Table 1 ).
|Original studies||Remeasurement for intraobserver variability||Remeasurement for interobserver variability|
|“Optimal resolution” method|
The purpose of the present retrospective study was to evaluate the impact of standardizing the timing of measurement of CIMT within the cardiac cycle, and therefore a separate observer (D.G.) made offline CIMT measurements on the studies of the included subjects using the same software and images but exclusively using the frames coincident with the R wave on the electrocardiogram ( Figure 1 B.) This observer was blinded to the measurements originally taken by the sonographers. A fourth observer (E.S.S.T.) independently measured a random subset of the studies measured by D.G. offline, using the same method D.G. used, to determine interobserver variability. Six months later, the same observer (E.S.S.T.) remeasured a subset of the studies measured by E.S.S.T. offline, using the same method, for intraobserver variability ( Table 1 ). The original sonographers were not asked to do offline R-wave measurements, so that no measurement bias was introduced in the ongoing prospective study.
For both the right and left carotid arteries, up to three measurements of CIMT were averaged for the analysis. The intraclass correlation coefficient (ICC) was used as a measure of reproducibility, and ICCs are reported with 95% confidence intervals. Agreement was also estimated by percentage error measurement, defined as 100 times the absolute value of the difference in the two measurements divided by the mean of the two measurements. Bland-Altman plots were used for comparison and reproducibility of the “optimal resolution” and R-wave measurements; statistical significance was assessed with a paired t test. Pearson or Spearman correlation coefficients were used to explore the relationships of CIMT values, age, biometrics, blood pressure, and lipid profile with the CIMT value and percentage error, and an unpaired t test was used for the binary variables gender and case or control status. Differences in intraobserver variability among the observers were assessed by analysis of variance using the Welch modification for unequal variances, and the Games-Howell post-hoc test was used to evaluate the significance of differences between groups. A two-sided P value < .05 was used to indicate statistical significance. Analyses were performed with SPSS version 20 (SPSS, Inc, Chicago, IL).
Carotid artery ultrasound images of 184 children and adolescents (149 cases, 35 controls; mean age, 14.7 ± 2.2 years [range, 11.3–19.0 years]; 116 male [63%]) were included in the study ( Tables 2 and 3 ). In 97.8% ( n = 180) and in 96.2% ( n = 177) for the right and left sides, respectively, the sonographers could make the online “optimal resolution” CIMT measurements at all three angles. In 11 of these 184 studies, the sonographers could not make the CIMT measurements at all three angles; in one study on the right and two studies on the left side, CIMT measurement could be made at one angle, while in three studies on the right and five studies on the left side, CIMT measurements could be made at two angles only. In 6.7% of all online measurements, the “optimal resolution” CIMT measurements were made on the R wave. Because some studies were not available for offline analysis and some did not have electrocardiographic tracings on display to use for “R-wave” CIMT measurements (the electrocardiographic tracing had not been activated in the vascular setting on the ultrasound machine), this retrospective analysis could not be performed in all cases. In the subset of studies ( n = 126 and n = 136 studies on the right and left sides, respectively) in which offline measurements could be successfully performed, the third observer (D.G.) was able to make at least one offline measurement on the R-wave on the right or left CCA. In 95.2% ( n = 120) and in 97.8% ( n = 133) for the right and left sides, respectively, D.G. could make the measurements coincident with the R wave at all three angles. In eight studies, D.G. could not make the R-wave CIMT measurements at all three angles; in one study on each side, a CIMT measurement could be made at one angle, while in five studies on the right and two studies on the left side, CIMT measurements could be made at two angles only.
|Age (y)||14.7 ± 2.2|
|Weight (kg)||59.3 ± 15.7|
|Height (cm)||164.3 ± 11.3|
|BSA (m 2 )||1.64 ± 0.26|
|BMI (kg/m 2 )||21.7 ± 4.4|
|BMI ( Z score)||0.54 ± 2.09|
|Gender (male)||116 (63%)|
|Native Hawaiian/Pacific Islander||0|
|Total cholesterol (mg/dL)||153.1 ± 29.6|
|High-density lipoprotein cholesterol (mg/dL)||54.4 ± 11.7|
|Triglycerides (mg/dL)||77.6 ± 38.7|
|Low-density lipoprotein cholesterol (mg/dL)||84.3 ± 24.2|
|Systolic blood pressure (mm Hg)||116.2 ± 11.8|
|Systolic blood pressure ( Z score)||0.62 ± 0.96|
|Diastolic blood pressure (mm Hg)||70.2 ± 8.43|
|Diastolic blood pressure ( Z score)||1.30 ± 0.89|
|Mean blood pressure (mm Hg)||82.1 ± 8.4|
|Mean blood pressure ( Z score)||0.56 ± 0.83|
CIMT measurements coincident with the R wave strongly correlated with the original online measurements made using the “optimal resolution” method ( r = 0.726, P < .001, and r = 0.791, P < .001, for right and left CIMT, respectively). However, they were significantly thicker than the “optimal resolution” CIMT (right CCA, 0.439 ± 0.030 vs 0.428 ± 0.024 mm, P < .001; left CCA, 0.446 ± 0.030 vs 0.434 ± 0.025 mm, P < .001; Table 4 , Figures 2 A and 2B).
|Method||CIMT of the right CCA (mm)||CIMT of the left CCA (mm)|
|“Optimal resolution”||0.428 ± 0.024 ( n = 184)||0.434 ± 0.025 ( n = 184)|
|“Optimal resolution” ∗||0.429 ± 0.026 ( n = 126)||0.435 ± 0.026 ( n = 136)|
|R wave||0.439 ± 0.030 ( n = 126)||0.446 ± 0.030 ( n = 136)|
Inter- and Intraobserver Variability
Table 1 details the number of studies measured and remeasured by all of the observers.
CIMT Measurements with the “Optimal Resolution” Method
The ICC for interobserver variability between the two sonographers was 0.86 for both right and left CCA, as summarized in Table 5 and shown in Figures 3 A and 3B. Intraobserver variability was similar for each of these sonographers, ranging from 0.78 to 0.94. Interobserver percentage error was 2.8 ± 2.3% for the right CCA and 2.9 ± 2.5% for the left CCA. Intraobserver percentage error for CIMT measurements was 2.4 ± 1.7% and 3.2 ± 3.3% for the right CCA and 2.1 ± 1.8% and 5.2 ± 4.6% for the left CCA for each sonographer, respectively. The percentage error for both inter- and intraobserver measurements was not significantly associated with gender; lipid profile; body surface area; body mass index Z score; systolic, diastolic, or mean blood pressure Z scores; or whether the subject belonged in the case or control group.
|n||ICC (95% CI)||Percentage error, mean ± SD||n||ICC (95% CI)||Percentage error, mean ± SD|
|“Optimal resolution” method|
|Right CCA||13||0.83 (0.41–0.95)||3.2 ± 3.3||25||0.86 (0.69–0.94)||2.8 ± 2.3|
|12||0.88 (0.62–0.96)||2.4 ± 1.7|
|Left CCA||12||0.94 (0.81–0.98)||2.1 ± 1.8||24||0.86 (0.67–0.94)||2.9 ± 2.5|
|12||0.78 (0.22–0.94)||5.2 ± 4.6|
|Right CCA||9||0.91 (0.69–0.98)||1.9 ± 2.2||24||0.85 (0.65–0.93)||2.2 ± 2.9|
|Left CCA||9||0.99 (0.98–1.00)||0.6 ± 0.8||24||0.95 (0.87–0.98)||2.1 ± 1.7|