Defining Left Ventricular Hypertrophy in Children on Peritoneal Dialysis

Discussion

In this study we utilized the largest series of echocardiographies collated in dialyzed children to date to examine the effect of using different reference systems on the apparent prevalence of LVH in this population.

Current consensus guidelines recommend indexing LVM to height to the power of 2.7 in pediatric subjects (13). These guidelines are based on studies performed in the early 1990s, when obesity was still less prevalent and different echocardiographic equipment and techniques were utilized.

In analogy to the findings of Foster et al. and Khoury et al. in healthy children (14,15), we observed a significant inverse relationship of this LVMI with body size in children shorter than 130 cm. Although we cannot rule out the possibility of more marked LVH in young as compared with older children on dialysis, it is most likely that the height exponent of 2.7 does not adequately reflect the biologic relationship between heart size and body size in small children with CKD as in healthy children. The reason for the incomplete age normalization might be given by the fact that the inadequate normalization achieved by indexing to height to the power of 2.7 in very small individuals was apparently ignored in the original publication, which encompassed 241 adults and 444 infants, children, and adolescents (11). LVM is likely to scale best to lean body mass (17,18), of which height, even if used with allometric correction factors, is a rather imperfect surrogate. Whereas a precise assessment of lean body mass requires elaborate technologies such as densitometry or isotope dilution studies, allometric equations utilizing height and weight allow for estimation of lean body mass. One approximation is body surface area, which is closely correlated with basal metabolic rate and incorporates the product of height to the power of 0.73 and weight to the power of 0.43. Recently, an allometric pediatric estimation equation for lean body mass utilizing slightly modified allometric exponents to weight and height has been proposed (19). However, the approach to normalize LVM to anthropometric estimates of lean body mass is inherently limited by the predictive power of allometric equations based on height and weight. The body weight component in the equation may introduce bias related to variable inclusion of obese subjects in the reference and study cohorts. In diseased populations such as children with CKD, additional inaccuracies may occur because of simultaneous malnutrition and fluid retention, which may result in normal weight for height despite reduced lean body mass. It would be of interest to utilize more direct measures of lean body mass such as multifrequency bioimpedance and other emerging noninvasive technologies.

While awaiting further fundamental research in pediatric echocardiography, current studies in diseased populations have to utilize the existing body of reference information (Table 1). Here, we applied four reference systems to our study cohort and compared the results with respect to the calculated prevalence of LVM abnormalities in children on dialysis.

The most widely used definition of LVH is an LVMI exceeding 38.6 g/m^2.7, which corresponds to the 95th percentile of LVM distribution in a study of 192 healthy children aged 6 to 17 years living in Cincinnati, Ohio, in the early 1990s (20). Major advances in echocardiographic methodology and changes in anthropometric indices at the population level have occurred since then, creating a need for updated reference standards.

Recently, Foster et al. studied 440 healthy children and 239 children at risk of LVH (14). This single-center study was carefully standardized and a single cardiologist performed all measurements. All participants were normotensive at the time of the study and 1 year later. The authors utilized the LMS method to provide height-specific coefficients describing the distribution of LVM in the reference population and allowing quantitation of an individual's value by standardized SDS. Being particularly advantageous for parameters with non-Gaussian distribution, this approach is the current method of choice in cardiovascular auxology, applied, for instance, to express BMI, ABPM, and carotid intima-media thickness (21–23).

Khoury et al. recently published reference percentile tables for LVM and LVMI (g/m^2.7) obtained in 2273 healthy children, the largest reference population collected to date (15). Details regarding the standardization of echocardiographic and anthropometric measurements were not given. The authors chose to categorize by gender and age (0 to 6 months, 6 to 18 months, and thereafter at 2-year intervals). The relatively broad age categories could introduce errors, mainly in the youngest, smallest children, among whom even a small difference in age may translate into a big difference in height and consequently in normal LVMI. Moreover, the reference to chronological age in the Khoury charts creates a problem of applicability in populations in which the normal relations among age, height, and sexual maturation are disturbed, as is the case in children with CKD. An approximative solution to this problem is the use of height age instead of chronological age, assuming that the body composition, cardiac mass, and cardiac output of a growth-retarded child should match that of a child with the same height who is at the 50th percentile for age. Although one could speculate whether the lean mass of a very short, but fully sexually mature, adolescent should be appropriate for pubertal stage rather than matching that of a late prepubertal child of the same height, this consideration is largely irrelevant for LVMI, which attains constant values in the second decade of life. Hence, we propose to use height age when referencing a growth-retarded population to the Khoury tables. In this analysis, we compared the effects of referencing to chronological and height age, respectively. Although the overall difference in LVH prevalence was small, referencing to height age eliminated the apparent effect of growth retardation on LVH prevalence that was observed when referencing to chronological age.

Applied to our cohort, the Khoury and the Foster reference tables efficiently corrected the overestimation of LVH by the fixed LVMI criterion in young children, resulting in prevalence estimates independent of or even slightly increasing with age. However, the two reference systems yielded markedly different LVH prevalence figures. Whereas the 95th LVMI percentile of Khoury et al. approaches the previously established constant LVH criterion of 38.6 g/m^2.7 in school children and adolescents, the Foster charts show a much higher cutoff level and lower prevalence of LVH. The LVM SDS distribution plot further showed a large fraction of patients with apparent abnormally small heart size with the Foster charts that was not evident from the LVMI distribution within the Khoury charts. This suggests that the mu coefficients (distribution median) may have been overestimated and the sigma coefficients (distribution width) may have been underestimated in the relatively small validation cohort in the Foster study.

The differences in LVM distribution between the two recent reference cohorts might in part be explained by differences in body composition: Mean BMI was around the 58th percentile in the Foster study, whereas it was around the 45th percentile in the study of Khoury et al. These differences might reflect systematic differences in lean body mass that were not sufficiently accounted for by referencing to age or height. Furthermore, it is important to recognize the potential effect of methodological differences between the two studies. In the echocardiographic studies for the Foster paper, the wall and septum borders were hand-digitized using custom software, and thicknesses and diameters were calculated from the digitized borders as continuous variables throughout the cardiac cycle. End-diastole was defined as the time of maximum left ventricular dimension. The end-diastolic dimensions and wall thicknesses were calculated as the average of three successive cardiac cycles (14,24). Because this methodology of image analysis is not widely used, it may be the main reason for the difference in LVM estimates between the Foster and the Khoury papers.

Hence, the Foster and the Khoury studies represent significant advances in pediatric echocardiography, and both have drawbacks. The strength of the Khoury study is its very large cohort size and its partial concordance with previous reference data.

In subjects taller than 130 cm, the 95th LVMI percentile of Khoury approaches 38 g/m^2.7, the widely used—albeit not end point–validated—cutoff to define LVH in previous pediatric studies. The reference to age categorized at relatively broad intervals instead of a measure of body size is an important limitation of the Khoury tables that may be partially overcome by referencing to height age.

The weakness of the Foster reference group is its relatively small sample size and the fact that the measurement techniques used are not the same as those that are broadly applied in clinical practice.

On the other hand, the scaling of LVM to body size on a continuous scale is a clear advantage of the Foster study, as is the expression of LVM for height by standardized SDS. The latter allows using individual values from children of all ages for parametric statistical analysis in cross-sectional and longitudinal studies, a clear advantage over the traditional dichotomous LVH classification. With respect to the definition of LVH, the 97th percentile of the Foster reference data at the upper end of the height distribution corresponds with the LVMI cutoff of 51 g/m^2.7, which has been linked with adverse outcomes in adults (25).

In this study we validated existing definitions of LVH in a global registry-based population of children on CPD. The lack of standardization of echocardiographic technique and anthropometric measurements is an obvious limitation inherent to the data. Potential bias might also emerge from the application of reference systems validated in healthy children in case the underlying the relationship between LVM and body size differs fundamentally in children with ESRD. This general possibility is difficult to exclude and applies to any echocardiographic study in diseased pediatric populations.

In conclusion, recent advances in pediatric echocardiography allow a fresh look at abnormalities of cardiac adaptation in health and disease. However, further efforts will be required to develop optimal normalization techniques to express LVM and other morphologic and functional measures across the pediatric age range.