Response to Letter Regarding Article, “A Novel Method of Expressing Left Ventricular Mass Relative to Body Size in Children”
We thank Bocelli et al for their thoughtful comments on our article in which we proposed a new method of normalizing left ventricular (LV) mass for body size in children.1 They suggest that the inappropriate use of LV mass index (LVMI; in g/m2.7) in children <2 years of age was responsible for the flaws that we demonstrated with the LVMI. We agree that the younger the child is, the more problematic the LVMI is. However, residual dependence of LVMI on height extended beyond infancy, up to a height of ≈140 cm (the average height for a 10-year-old boy), indicating its limitations even in older children.
Bocelli et al propose raising height to the power of 2, rather than 2.7, in the calculation of LVMI in children and demonstrate that doing so improves the performance of the allometric scaling method. We agree that it may be possible to acceptably normalize LV mass to body size using allometric scaling, provided that the appropriate exponents are chosen. However, it is likely that the value of the exponent increases gradually with age, from a value <2.0 in neonates to 2.7 in taller children and adults. A small but significant residual dependence of LVMI on height remains in children <140 cm tall (R2=0.12, P<0.0001). An important advantage of the centiles method that we proposed is that it can be applied meaningfully from infancy to adulthood. A similar degree of resolution with allometric scaling would be possible only if the exponent used increased with increasing age.
Bocelli et al note that body surface area (BSA) has frequently been the preferred scaling variable for infants. We would argue that choice of scaling variable and method of normalization are 2 separate issues. As with height, the relation between LV mass and BSA is not the same throughout childhood.2 Therefore, successful normalization of LV mass to BSA with allometric scaling would also require that BSA be raised to different powers at different ages. Whether height or BSA is the best scaling variable for LV mass remains an unanswered question. The centiles method that we proposed would work equally well with BSA as the scaling variable.
Finally, we must disagree with the description by Bocelli et al of the centiles method as “complex.” Although the statistical methods underlying the development of the curves are indeed complex, practical use of the curves and calculation of z scores are very simple.