## Background

Echocardiographic measurement of left ventricular (LV) mass is routinely performed in pediatric patients with elevated cardiovascular risk. The complex relationship between heart growth and body growth in children requires normalization of LV mass to determine its appropriateness relative to body size. LV mass is strongly determined by lean body mass (LBM). Using new LBM predictive equations, the investigators generated sex-specific LV mass-for-LBM centile curves for children 5 to 18 years of age.

## Methods

This retrospective study used M-mode echocardiographic data collected from 1995 through 2003 from 939 boys and 771 girls between 5 and 18 years of age (body mass index < 85th percentile for sex and age) to create smoothed sex-specific LV mass-for-LBM reference centile curves using the Lamda Mu Sigma method. The newly developed reference centiles were applied to children with essential hypertension and with chronic kidney disease, groups known to be at high risk for LV hypertrophy (LVH). The identification of LVH using two different normalization approaches was compared: LV mass-for-LBM and LV mass index-for-age percentiles.

## Results

Among 231 children at risk for LVH, on average, relative LV mass was higher using the LV mass index-for-age percentile method than the LV mass-for-LBM percentile method. LVH was more likely to be diagnosed among overweight children and less likely among thin children.

## Conclusions

This study provides new LV mass reference centiles expressing LV mass relative to LBM, the strongest determinant of LV mass. These reference centiles may allow more accurate stratification of cardiovascular risk in children.

The presence of left ventricular (LV) hypertrophy (LVH) is a strong predictor of adverse cardiovascular outcomes in adults and is considered an intermediate outcome of clinical importance in children. To determine whether LVH is present, LV mass must be measured and then normalized to determine its appropriateness relative to body size. Currently, the most accepted method of normalizing LV mass is to calculate the LV mass index (LVMI; LV mass [g]/height [m] ^{2.7} ) and express it as a percentile relative to age. However, this method has limitations, resulting in controversy regarding both the best method of normalizing LV mass and the best body size variable against which to normalize LV mass measurements.

One of the most flexible and practical methods of normalizing is the Lamda Mu Sigma (LMS) method. This is the same technique that was used to create the child growth curves and to create reference centiles for ambulatory blood pressure in children. We have previously demonstrated the usefulness of this approach in normalizing LV mass for body size.

The choice of LV mass scaling variable is best based on physiology. Many studies in children and adults have shown that LV mass is strongly determined by lean body mass (LBM). More of the variability in LV mass is explained by LBM than by either height or weight alone. Height generally serves as a surrogate for LBM when it is used as the scaling variable. LVMI-for-age percentiles improve on raw LVMI by including the effects of age on LBM (and therefore on LV mass). Accounting for age ensures that the effects on LBM of different body proportions among infants, and of puberty among adolescents, are captured. However, height and age are not the only determinants of LBM ; weight, physical training, and race also contribute. Thus, using LBM to normalize LV mass in children would be a preferred method to generate normative pediatric references. One of the barriers to the development and use of such centiles has been the difficulty in measuring LBM in clinical practice. Foster *et al* . recently developed and validated predictive equations to estimate LBM among individuals 5 to 21 years of age.

This study had two aims. First, we aimed to generate new, smoothed LV mass-for-LBM centile curves for children and adolescents between 5 and 18 years of age using the new equations for LBM and the same reference data set previously used to generate normative pediatric values of LVMI, wherein LVMI was expressed as a percentile relative to age. Second, we applied the newly developed reference centiles to children with essential hypertension and with chronic kidney disease, groups known to be at high risk for LVH, and compared the identification of LVH using the different normalization approaches.

## Methods

This retrospective study was conducted using echocardiograms of children previously participating in the development of normative values for LV mass at Cincinnati Children’s Hospital Medical Center (CCHMC). This is the largest and most commonly used reference data set. We also used the echocardiograms of children previously participating in prior research protocols conducted at CCHMC as well as studies done for clinical purposes at the Montreal Children’s Hospital. All research protocols were approved by the institutional review boards at the relevant institutions, and all subjects or their guardians gave written informed consent when required.

## Reference Subjects

Healthy, nonoverweight (body mass index [BMI] < 85th percentile for age, on the basis of US Centers for Disease Control and Prevention reference values ) children 5 to 18 years of age, with structurally normal hearts and no systemic disease, were evaluated in the echocardiography laboratory at CCHMC (1995-2003). Exclusion criteria included hypertension (defined as a history of a diagnosis of hypertension, use of antihypertensive medications, or a referral for evaluation of hypertension), dysrhythmia, structural heart disease, myocardial dysfunction, and valve regurgitation of more than trivial severity. Generally, these were children referred to the echocardiography laboratory for evaluation of an innocent murmur or noncardiac chest pain who were then determined by echocardiography to have normal cardiac anatomy and physiology. Height was measured to the nearest 1 cm and weight to the nearest 0.1 kg in all reference subjects.

## Subjects “At Risk” for LVH

Children at risk for LVH were evaluated at two centers. Seventy children with hypertension were studied in the echocardiography laboratory at the Montreal Children’s Hospital. At CCHMC, 79 children with chronic kidney disease stages 2 to 4, 33 children on maintenance dialysis, and 49 children with kidney transplants were studied. Height or length was measured to the nearest 1 mm and weight to the nearest 0.1 kg in all at-risk children.

## Echocardiography

Echocardiography was performed using commercially available cardiac ultrasound scanners according to the guidelines of the American Society of Echocardiography. Studies were recorded on VHS videocassettes. All studies were reviewed by a single pediatric cardiologist at each of the sites. Studies performed for clinical purposes were reexamined for the purposes of this study. Echocardiograms were analyzed offline. Three cardiac cycles were measured by one of two observers, and the indices were averaged from those three measurements. Prior studies have suggested that intraobserver and interobserver variability is low ( *r *= 0.96 and *r *= 0.84, respectively). Patients were examined in the left lateral decubitus position, resting comfortably. All echocardiographic studies were performed with the patient in a quiet, resting state. Two-dimensionally guided M-mode echocardiography was performed from a parasternal long-axis view. Using the leading edge–to–leading edge technique, the LV cavity dimension, posterior wall thickness, and interventricular septal thickness were measured at end-diastole at the level just below the mitral valve leaflets. LVM (grams) was calculated using the Devereux equation: LVM = 0.8{1.04[(LV cavity dimension + posterior wall thickness + interventricular septal thickness) ^{3 }− (LV cavity dimension) ^{3 }]} + 0.6.

## Anthropometric Measures

BMI was calculated and expressed as a *Z *score relative to age and sex, using the Centers for Disease Control and Prevention, National Centers for Health Statistics 2000 growth curves. Body surface area (BSA) was calculated using the Mosteller equation: BSA = √{[height (cm) × weight (kg)]/3,600}. We estimated LBM using validated sex-specific predictive equations. These predictive equations were developed using LBM measured by dual-energy x-ray absorptiometry in children between 5 and 21 years of age and validated in two independent samples. The equations predict LBM within 5% of measured LBM, with no evidence of bias by age or level of adiposity. The equation for male subjects is ln(LBM) = −2.8990 + 0.8064 × ln(height) + 0.5674 × ln(weight) + 0.0000185 weight ^{2 }− 0.0153 × (BMI *Z *score) ^{2 }+ 0.0132 × age. The equation for female subjects is ln(LBM) = −3.8345 + 0.954 × ln(height) + 0.6515 × ln(weight) − 0.0102 × (BMI *Z *score) ^{2 }. LBM was predicted only for children 5 to 18 years of age, because the LBM predictive equations have not been validated in younger children.

## Statistical Analysis

Analyses were performed using Stata version 12.0 (StataCorp LP, College Station, Texas) and LMS Chartmaker Pro (Institute for Child Health, London, United Kingdom).

## Generating LV Mass Centiles and *Z *Scores

Using data from the reference population, we constructed smoothed LV mass-for-LBM reference centile curves using the LMS method, as previously described. We initially generated a single set of reference centiles including both boys and girls, because prior studies suggested that sex differences in LVM were eliminated when LVM was expressed relative to LBM. However, we noted significant differences in relative LV mass between healthy reference boys and girls when data were combined, suggesting that the relationship between LV mass and LBM may differ for boys and girls. Therefore, we assessed the relation between LV mass and LBM in boys and girls using linear regression, with an interaction term to consider the possibility that the association differed by sex. A regression including a term for sex allows for the possibility that there was a difference of fixed magnitude between boys and girls across the whole range of LBM; an interaction term considers the possibility that the slope of the relation between LV mass and LBM differed in girls compared with boys (i.e., that the difference in LV mass between boys and girls differed for different values of LBM). On the basis of the results, we constructed sex-specific reference centiles.

A *Z *score was calculated for each reference and at-risk child using the following equation : *Z *score = [(LV mass/M) ^{L }− 1]/(L × S). The L, M, and S values at each LBM are available in Supplemental Tables 1 and 2 . The *Z *score is the difference, in SD units, between an individual’s LV mass and the mean for healthy reference children of the same LBM. *Z *scores were converted into percentiles using normal distribution tables. LVH was defined as LV mass-for-LBM > 95th percentile.

## Generating LVMI-for-Age *Z *Scores

We also generated LVMI-for-age *Z *scores and percentiles. LVMI was calculated as LV mass (g)/height (m) ^{2.7 }. Briefly, using the quantile regression methods previously described in detail, we created 99 curves representing the percentiles 1 through 99 using linear regression with LVMI as the dependent variable and age, √age × age, age ^{2 }, sex, and age × sex as the independent variables. This method allows for a set of noncrossing percentile curves for each sex. Then each subject’s LVMI-for-age percentile was estimated by interpolating to a value between the next lower and higher percentile cut points for that subject’s age. Percentiles were converted to *Z *scores using normal tables. A value above the 99th percentile was set to the 99th percentile. LVH was defined as LVMI-for-age > 95th percentile.

## Comparing Relative LV Mass and Identification of LVH on the Basis of LVMI-for-Age Percentile versus LV Mass-for-LBM

Children in the “at-risk” group were each assigned an LV mass-for-LBM *Z *score and percentile and an LVMI-for-age *Z *score and percentile and were classified as having LVH, or not, on the basis of LVMI-for-age > 95th percentile and again on the basis of LV mass-for-LBM > 95th percentile. *Z *scores using the two methods were compared graphically using a simple correlation plot, a Bland-Altman plot, and a modified Bland-Altman plot. The proportion of at-risk children with LVH (with 95% CI) was determined using each of the methods, and agreement between methods was assessed with 2 × 2 tables. The characteristics of subjects with discordant classification of LVH were evaluated.

## Results

## LV Mass-for-LBM Reference Centiles

There were 939 boys and 771 girls between 5 and 18 years of age in the healthy group used to create the LV mass-for-LBM reference centiles. The characteristics of these children are summarized in Table 1 . We initially created a single set of LV mass-for-LBM centile curves including both boys and girls. However, there were significant differences in the LV mass-for-LBM *Z *-score distributions for healthy reference boys and girls (mean, +0.06 ± 0.03 for boys and −0.08 ± 0.04 for girls; *P *= .004). A linear regression model to assess the relationship between LV mass and LBM showed that for every 1 kg higher the LBM, LV mass was 2.44 g (95% CI, 2.37-2.52 g) higher; this relation did not differ by sex (interaction *P *= .90). However, at any given LBM, the LV mass was 2.28 g (95% CI, 0.40-4.16 g) lower in girls than in boys. Therefore, we generated sex-specific centile curves. The LV mass-for-LBM centile curves describe the distribution of LV mass relative to LBM among healthy, nonoverweight children. By definition, the mean LV mass-for-LBM *Z *score for each reference sample (i.e., boys 5-18 years of age and girls 5-18 years of age) was zero, with an SD of 1.0. The reference centile curves are shown in Figures 1 A and 1B. Supplemental Figures 1A and 1B show the sex-specific plots of LV mass by LBM on which centile curves were based.

Variable | Male subjects | Female subjects |
---|---|---|

n |
939 | 771 |

Age (y) | 12.6 (8.9 to 15.0) | 12.5 (8.6 to 14.9) |

Height (cm) | 155 (132 to 171) | 152 (130 to 162) |

Height-for-age percentile | 55.6 (29.0 to 81.9) | 51.2 (26.5 to 77.0) |

Weight (kg) | 43.1 (27.8 to 59.0) | 42.3 (26.2 to 53.5) |

BMI (kg/m ^{2 }) |
17.9 (16.0 to 20.2) | 18.0 (15.9 to 20.6) |

BMI percentile | 50.7 (27.9 to 69.9) | 54.7 (30.5 to 70.3) |

BMI Z score |
0.02 (−0.58 to 0.52) | 0.12 (−0.51 to 0.53) |

LBM (kg) | 33.2 (21.1 to 45.2) | 30.0 (18.6 to 36.7) |

BSA (m ^{2 }) |
1.37 (1.01 to 1.67) | 1.34 (0.98 to 1.55) |

LV mass (g) | 88.6 (62.0 to 122.8) | 78.2 (54.0 to 100.8) |

LVMI (g/m ^{2.7 }) |
29.5 (25.3 to 34.4) | 27.4 (23.8 to 31.5) |

Interventricular septum thickness (cm) | 0.69 (0.59 to 0.80) | 0.65 (0.56 to 0.74) |

Posterior wall thickness (cm) | 0.65 (0.55 to 0.76) | 0.61 (0.52 to 0.70) |

LV diastolic dimension (cm) | 4.41 (3.92 to 4.90) | 4.24 (3.78 to 4.59) |

Relative wall thickness | 0.30 (0.26 to 0.34) | 0.29 (0.26 to 0.33) |

## At-Risk Children

There were 231 children 5 to 18 years of age in the group at risk for LVH. The characteristics of the at-risk group are presented in Table 2 . As shown in Table 2 , among at-risk children, relative LV mass was higher using the LVMI-for-age percentile method (median, 90.0) than when the LV mass-for-LBM percentile method was used (median, 81.8). Figure 2 shows the relation between LV mass-for-LBM and LVMI-for-age *Z *scores, as well as agreement in diagnosis of LVH. Supplemental Figure 2 shows a Bland-Altman plot to assess agreement between the measures, and Supplemental Figure 3 shows a plot of the difference in the *Z *scores by BMI *Z *score (assessing bias by adiposity).

Variable | Value |
---|---|

n |
231 |

Age (y) | 13.8 (10.5 to 16.7) |

Male | 150 (65%) |

Height (cm) | 151 (138 to 165) |

Height-for-age percentile | 29.4 (33.3 to 94.4) |

Weight (kg) | 50.0 (36.0 to 67.1) |

BMI (kg/m ^{2 }) |
20.6 (17.8 to 25.3) |

BMI percentile | 74.7 (33.3 to 94.4) |

BMI Z score |
0.66 (−0.43 to 1.59) |

LBM (kg) | 34.6 (24.6 to 44.8) |

BSA (m ^{2 }) |
1.48 (1.18 to 1.73) |

Disease category | |

CKD | 79 (34.2%) |

Dialysis | 33 (14.3%) |

Transplantation | 49 (21.2%) |

Hypertension | 70 (30.3%) |

LVMI-for-age Z score |
1.28 (0.55 to 2.33) |

LVMI-for-age percentile | 90.0 (71.0 to 99.0) |

LV mass-for-LBM Z score |
0.91 (0.18 to 2.00) |

LV mass-for-LBM percentile | 81.8 (57.0 to 97.7) |

Differences by normalization approach in the identification of LVH among at-risk children are highlighted by the comparisons in Table 3 and Figure 2 . Of the 231 at-risk children, LV mass-for-LBM *Z *scores could be determined in 227; of these, 88 (38.8%; 95% CI, 26.5%–39.1%) were classified as having LVH using the LVMI-for-age percentile method, compared with 74 (32.6%; 95% CI, 26.5%–39.1%) using LV mass-for-LBM percentile. Assuming the LV mass-for-LBM percentile method correctly classifies children with respect to LVH, there were 31 children (13.6%; 95% CI, 9.5%-18.8%) who were misclassified. Among those with no LVH on the basis of LV mass-for-LBM percentile, 14.4% (95% CI, 9.2%-21.0%) were classified as having LVH on the basis of LVMI for age; these children tended to be overweight, with a median BMI-for-age percentile of 95.8 (interquartile range, 90.3-98.7). Among those with LVH on the basis of LV mass-for-LBM percentile, 10.8% (95% CI, 4.8%-20.2%) were classified as having normal LV mass on the basis of LVMI-for-age percentile; these children were thin, with a median BMI-for-age percentile of 9.8 (interquartile range, 5.5-22.5).

## Results

## LV Mass-for-LBM Reference Centiles

There were 939 boys and 771 girls between 5 and 18 years of age in the healthy group used to create the LV mass-for-LBM reference centiles. The characteristics of these children are summarized in Table 1 . We initially created a single set of LV mass-for-LBM centile curves including both boys and girls. However, there were significant differences in the LV mass-for-LBM *Z *-score distributions for healthy reference boys and girls (mean, +0.06 ± 0.03 for boys and −0.08 ± 0.04 for girls; *P *= .004). A linear regression model to assess the relationship between LV mass and LBM showed that for every 1 kg higher the LBM, LV mass was 2.44 g (95% CI, 2.37-2.52 g) higher; this relation did not differ by sex (interaction *P *= .90). However, at any given LBM, the LV mass was 2.28 g (95% CI, 0.40-4.16 g) lower in girls than in boys. Therefore, we generated sex-specific centile curves. The LV mass-for-LBM centile curves describe the distribution of LV mass relative to LBM among healthy, nonoverweight children. By definition, the mean LV mass-for-LBM *Z *score for each reference sample (i.e., boys 5-18 years of age and girls 5-18 years of age) was zero, with an SD of 1.0. The reference centile curves are shown in Figures 1 A and 1B. Supplemental Figures 1A and 1B show the sex-specific plots of LV mass by LBM on which centile curves were based.

Variable | Male subjects | Female subjects |
---|---|---|

n |
939 | 771 |

Age (y) | 12.6 (8.9 to 15.0) | 12.5 (8.6 to 14.9) |

Height (cm) | 155 (132 to 171) | 152 (130 to 162) |

Height-for-age percentile | 55.6 (29.0 to 81.9) | 51.2 (26.5 to 77.0) |

Weight (kg) | 43.1 (27.8 to 59.0) | 42.3 (26.2 to 53.5) |

BMI (kg/m ^{2 }) |
17.9 (16.0 to 20.2) | 18.0 (15.9 to 20.6) |

BMI percentile | 50.7 (27.9 to 69.9) | 54.7 (30.5 to 70.3) |

BMI Z score |
0.02 (−0.58 to 0.52) | 0.12 (−0.51 to 0.53) |

LBM (kg) | 33.2 (21.1 to 45.2) | 30.0 (18.6 to 36.7) |

BSA (m ^{2 }) |
1.37 (1.01 to 1.67) | 1.34 (0.98 to 1.55) |

LV mass (g) | 88.6 (62.0 to 122.8) | 78.2 (54.0 to 100.8) |

LVMI (g/m ^{2.7 }) |
29.5 (25.3 to 34.4) | 27.4 (23.8 to 31.5) |

Interventricular septum thickness (cm) | 0.69 (0.59 to 0.80) | 0.65 (0.56 to 0.74) |

Posterior wall thickness (cm) | 0.65 (0.55 to 0.76) | 0.61 (0.52 to 0.70) |

LV diastolic dimension (cm) | 4.41 (3.92 to 4.90) | 4.24 (3.78 to 4.59) |

Relative wall thickness | 0.30 (0.26 to 0.34) | 0.29 (0.26 to 0.33) |

## At-Risk Children

There were 231 children 5 to 18 years of age in the group at risk for LVH. The characteristics of the at-risk group are presented in Table 2 . As shown in Table 2 , among at-risk children, relative LV mass was higher using the LVMI-for-age percentile method (median, 90.0) than when the LV mass-for-LBM percentile method was used (median, 81.8). Figure 2 shows the relation between LV mass-for-LBM and LVMI-for-age *Z *scores, as well as agreement in diagnosis of LVH. Supplemental Figure 2 shows a Bland-Altman plot to assess agreement between the measures, and Supplemental Figure 3 shows a plot of the difference in the *Z *scores by BMI *Z *score (assessing bias by adiposity).

Variable | Value |
---|---|

n |
231 |

Age (y) | 13.8 (10.5 to 16.7) |

Male | 150 (65%) |

Height (cm) | 151 (138 to 165) |

Height-for-age percentile | 29.4 (33.3 to 94.4) |

Weight (kg) | 50.0 (36.0 to 67.1) |

BMI (kg/m ^{2 }) |
20.6 (17.8 to 25.3) |

BMI percentile | 74.7 (33.3 to 94.4) |

BMI Z score |
0.66 (−0.43 to 1.59) |

LBM (kg) | 34.6 (24.6 to 44.8) |

BSA (m ^{2 }) |
1.48 (1.18 to 1.73) |

Disease category | |

CKD | 79 (34.2%) |

Dialysis | 33 (14.3%) |

Transplantation | 49 (21.2%) |

Hypertension | 70 (30.3%) |

LVMI-for-age Z score |
1.28 (0.55 to 2.33) |

LVMI-for-age percentile | 90.0 (71.0 to 99.0) |

LV mass-for-LBM Z score |
0.91 (0.18 to 2.00) |

LV mass-for-LBM percentile | 81.8 (57.0 to 97.7) |