Background
Several coronary artery Z score models have been developed. However, a Z score model derived by the lambda-mu-sigma (LMS) method has not been established.
Methods
Echocardiographic measurements of the proximal right coronary artery, left main coronary artery, proximal left anterior descending coronary artery, and proximal left circumflex artery were prospectively collected in 3,851 healthy children ≤18 years of age and divided into developmental and validation data sets. In the developmental data set, smooth curves were fitted for each coronary artery using linear, logarithmic, square-root, and LMS methods for both sexes. The relative goodness of fit of these models was compared using the Bayesian information criterion. The best-fitting model was tested for reproducibility using the validation data set. The goodness of fit of the selected model was visually compared with that of the previously reported regression models using a Q-Q plot.
Results
Because the internal diameter of each coronary artery was not similar between sexes, sex-specific Z score models were developed. The LMS model with body surface area as the independent variable showed the best goodness of fit; therefore, the internal diameter of each coronary artery was transformed into a sex-specific Z score on the basis of body surface area using the LMS method. In the validation data set, a Q-Q plot of each model indicated that the distribution of Z scores in the LMS models was closer to the normal distribution compared with previously reported regression models. Finally, the final models for each coronary artery in both sexes were developed using the developmental and validation data sets. A Microsoft Excel–based Z score calculator was also created, which is freely available online ( http://raise.umin.jp/zsp/calculator/ ).
Conclusions
Novel LMS models with which to estimate the sex-specific Z score of each internal coronary artery diameter were generated and validated using a large pediatric population.
Highlights
- •
Echocardiographic data were collected from 3,851 healthy children.
- •
A sex-specific Z -score model of each coronary artery was generated and validated.
- •
The LMS model with BSA as the independent variable showed the best goodness of fit.
Kawasaki disease (KD) is a childhood systemic vasculitis syndrome that mainly affects the coronary arteries. Although recent progress in the treatment of KD has decreased the incidence of echocardiographically detectable coronary artery abnormalities (CAAs) to <3% at 28 days after onset, KD remains the most common cause of pediatric acquired heart disease in developed countries. The occurrence of giant aneurysms is indeed the most severe outcome in KD; however, recognition of mild or transient dilatation of the coronary arteries is generally agreed as important in the diagnosis, management, and long-term follow-up of patients with KD.
Two-dimensional echocardiography is commonly used to identify the presence of CAAs. The initial definition of CAA was given in the Japanese Ministry of Health criteria, which dichotomously define abnormalities as a maximum absolute internal diameter of ≥3 mm in children <5 years of age, a segment 1.5 times greater than an adjacent segment, or the presence of luminal irregularity. The Japanese Ministry of Health criteria were widely used because of their simple and quick assessment of CAA formation. However, the criteria have several flaws, such as a lack of adjustment for body size and a lack of a standard specific for each major coronary artery.
Quantitative assessment of cardiac and vascular dimensions is essential for the evaluation and management of cardiovascular disorders. However, the estimation of accurate standard values for the pediatric population remains a substantial challenge because the dimensions of the heart and vessels depend on body size. To adjust for these differences in body size, a standardized score ( Z score) was proposed and is now widely used in the clinical setting. Regrettably, recent critical and systematic reviews have pointed out that echocardiographic nomograms or reference values of cardiovascular structures in the pediatric population have many problems, including a limited number of healthy subjects, poor differentiation among age subgroups, and methodologic flaws. The previously reported Z score for each coronary artery also reflected these limitations; especially, these models did not have validated accuracy.
In 1992, Cole and Green proposed the lambda-mu-sigma (LMS) method with penalized likelihood to generate accurate normal references, which is extremely flexible and widely applicable. The LMS method has been increasingly used in recent years in standards such as the Centers for Disease Control and Prevention’s growth charts in the United States and the World Health Organization’s worldwide child growth standards. Z score curves for pediatric coronary arteries derived using the LMS method might provide more accurate standard values.
Against this background, this study was undertaken to develop a new pediatric Z score for coronary arterial internal diameter on the basis of a large sample using a statistical approach based on the LMS method. We also validated the reproducibility of the LMS models and compared the LMS models with previously reported models.
Methods
This was a multicenter, prospective, observational study in which 82 pediatric cardiologists and sonographers at 43 institutes participated. At each center, children referred for echocardiography were assessed for eligibility for this study. Inclusion criteria were (1) age ≤ 18 years, (2) benign symptoms or asymptomatic electrocardiographic abnormalities without anatomic or functional abnormalities (i.e., functional murmur, chest pain or syncope without cardiac cause, abnormal electrocardiographic results without symptoms, etc), and (3) mild cardiac valvular insufficiency without hemodynamic effects (i.e., trivial aortic or mitral valve regurgitation, mitral valve prolapse, mild aortic or pulmonary valve stenosis with flow velocity < 2.0 m/sec, or mild mitral or tricuspid valve stenosis with flow velocity < 1.5 m/sec). Exclusion criteria were (1) congenital heart defects, (2) history of KD or surgical or catheter intervention for heart disease, (3) myocardial disease, (4) chromosomal disorders (i.e., trisomy 21, Turner syndrome, Noonan syndrome, Marfan syndrome, Williams syndrome, 22q11.2 deletion, etc), (5) birth weight < 1,500 g, and (6) significant underlying disease (i.e., malignant tumor, leukemia, metabolic or endocrine disease, neuromuscular disease, etc.). We collected anonymous data regarding age, sex, body height, body weight, and coronary arterial internal diameter. From May 2010 to April 2011, we enrolled 2,567 participants who underwent echocardiographic studies. These data (developmental data set) were used to develop the LMS models. From May 2011 to December 2011, an additional 1,284 patients were collected to test the accuracy of the LMS and previously reported models; these data represented the validation data set. This prospective observational study was approved by the institutional review board of NTT East Japan Sapporo Hospital on March 16, 2010.
Echocardiography
Echocardiographic studies were performed at each center using the standard measurement methods for pediatric coronary arteries recommended by the Japanese Society of Kawasaki Disease. Briefly, the patient is examined in the supine or right decubitus position using a sector probe with a ≥5-MHz frequency. The focus depth should be set to the coronary artery, and the frame rate should be increased to raise the time resolution. The coronary artery is observed at the center of the monitor screen and zoomed in two to three times before the measurement. During the coronary artery diameter measurement, the gain should be lowered as much as possible to minimize the trailing echo from the intima-lumen interface (minimal gain setting). The coronary arterial internal diameters were measured from inner edge to inner edge. Coronary arteries were measured at four points: (1) proximal right coronary artery (RCA), (2) left main coronary artery (LMCA), (3) proximal left anterior descending coronary artery (LAD), and (4) proximal left circumflex coronary artery (LCX). The diameters of the RCA, LAD, and LCX were measured at 3 to 5 mm distal to their origins in the parasternal short-axis view. The diameter of the LMCA was measured at the midpoint between the ostium of the LMCA and the bifurcation of the LAD and LCX in the parasternal short-axis view.
Quality Control of Echocardiography
Before data collection at each hospital, an accuracy control survey of all echocardiographic equipment using phantom-based measurements of depth of penetration, beam profile, near field, axial and lateral resolution, and vertical and horizontal distance accuracy was performed, and it was confirmed that all equipment could perform measurements with an accuracy of <0.1 mm. Next, all echocardiography operators were assessed in terms of their measurement accuracy using sample pictures of eight coronary arteries. We confirmed that all operators had excellent measurement accuracy (mean intraobserver measurement error, 3.7 ± 1.5%; mean interobserver error, 3.1 ± 2.5%). Finally, the first echocardiographic pictures of each coronary artery were sent to the Z Score Project office, and the accuracy of their visualization was checked. After these quality control procedures had been completed, we enrolled the participants.
Statistical Analysis
All statistical analyses were performed using SPSS version 22.0J (IBM SPSS, Tokyo, Japan) or R. Continuous variables are presented as mean ± SD and/or medians with ranges or interquartile ranges and were tested using Mann-Whitney U tests. Categorical variables are expressed as frequencies and percentages and were tested using Fisher exact tests. P values < .05 were considered to indicate statistical significance in two-sided tests.
We analyzed the sex-specific data on age, body height, body surface area (BSA) calculated using the Haycock formula, and the internal diameters of the RCA, LMCA, LAD, and LCX. We fitted smooth curves for each coronary artery using the LMS method in the developmental data set and showed figures with lines indicating the −2.5, −2.0, −1.0, 0.0, 1.0, 2.0, and 2.5 SD boundaries (corresponding to the 0.6, 2.3, 15.9, 50.0, 84.1, 97.7, and 99.4 percentiles, respectively). Detailed methods for developing the LMS models are shown in the Supplemental Methods . We also generated three regression models (linear, logarithmic, and square root). The model with the lowest Bayesian information criterion (BIC) value was considered the best model, and BIC values were compared among explanatory variables (age, body height, body weight, and BSA) as well. In the validation data set, we visually tested the reproducibility of the LMS models and the previously reported regression models using Q-Q plots. We also calculated χ 2 value per degrees of freedom, derived from the χ 2 goodness-of-fit test. Finally, we developed the final model using the entire data set, shown in the tables and figures as well.
Results
During the study period, 3,851 participants who underwent echocardiographic studies were enrolled. We divided these participants into the developmental data set (2,567 participants) and the validation data set (1,284 participants). All participants were admitted to our hospital because of heart murmurs ( n = 1408), abnormal electrocardiographic results during school physical examinations in first, seventh, and 10th grades ( n = 915), screening for cardiac disorders (normal neonates, before minor surgery, or patients with cyanosis) ( n = 740), arrhythmia ( n = 440), chest pain ( n = 247), syncope ( n = 55), or abnormal chest x-ray findings during school physical examinations in first, seventh, and 10th grades ( n = 25). Twenty-one participants were volunteers with verbal consent from their guardians. No participants had any cardiac disorders, cardiac dysfunction, significant background disease, or family history of cardiomyopathy. Table 1 shows sex-specific distributions regarding number, mean ± SD, and median with range in the developmental and validation data sets. The internal diameters of all coronary arteries in male participants were significantly larger than those in female patients; therefore, we decided to generate sex-specific models.
| Variable | Developmental data set | Validation data set | ||||
|---|---|---|---|---|---|---|
| Male | Female | P | Male | Female | P | |
| Age (y) | .014 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 6.63 ± 5.17 | 6.06 ± 5.10 | 7.76 ± 5.16 | 6.62 ± 5.18 | ||
| Median (range) | 6.7 (0–18.6) | 5.9 (0–18.3) | 8.8 (0–17.7) | 6.6 (0–18.9) | ||
| Body height (cm) | <.001 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 111 ± 39.0 | 105 ± 37.5 | 118 ± 38.0 | 109 ± 37.4 | ||
| Median (range) | 117 (43.2–191) | 111 (40.0–170) | 126 (45.0–185) | 115 (46.0–172) | ||
| Body weight (kg) | <.001 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 24.4 ± 17.9 | 21.9 ± 16.0 | 27.3 ± 17.4 | 23.0 ± 16.2 | ||
| Median (range) | 21.0 (2.0–102) | 18.9 (1.7–72.4) | 26.0 (2.4–85.5) | 19.7 (2.3–93.8) | ||
| BSA (m 2 ) | <.001 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 0.85 ± 0.45 | 0.78 ± 0.42 | 0.93 ± 0.44 | 0.82 ± 0.42 | ||
| Median (range) | 0.82 (0.15–2.29) | 0.76 (0.13–1.79) | 0.95 (0.17–2.04) | 0.79 (0.17–2.12) | ||
| Internal diameter of RCA (mm) | <.001 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 2.11 ± 0.73 | 1.92 ± 0.63 | 2.24 ± 0.72 | 2.00 ± 0.62 | ||
| Median (range) | 2.1 (0.6–5.1) | 1.9 (0.7–4.2) | 2.3 (0.5–4.1) | 2.0 (0.5–4.1) | ||
| Internal diameter of LMCA (mm) | <.001 | <.001 | ||||
| n | 1,385 | 1,182 | 693 | 591 | ||
| Mean ± SD | 2.40 ± 0.76 | 2.19 ± 0.69 | 2.54 ± 0.75 | 2.27 ± 0.66 | ||
| Median (range) | 2.4 (0.9–4.8) | 2.2 (0.8–4.3) | 2.6 (0.9–4.8) | 2.3 (0.8–4.2) | ||
| nternal diameter of LAD (mm) | <.001 | <.001 | ||||
| n | 1,385 | 1,181 | 693 | 591 | ||
| Mean ± SD | 1.98 ± 0.67 | 1.81 ± 0.59 | 2.11 ± 0.67 | 1.91 ± 0.59 | ||
| Median (range) | 2.0 (0.7–4.8) | 1.8 (0.6–3.8) | 2.1 (0.7–4.1) | 1.9 (0.8–3.6) | ||
| Internal diameter of LCX (mm) | <.001 | <.001 | ||||
| n | 1,356 | 1,165 | 671 | 567 | ||
| Mean ± SD | 1.75 ± 0.61 | 1.61 ± 0.55 | 1.87 ± 0.64 | 1.71 ± 0.59 | ||
| Median (range) | 1.7 (0.6–4.9) | 1.6 (0.6–3.4) | 1.9 (0.5–4.5) | 1.7 (0.5–3.5) | ||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
