How should we determine normal echocardiographic right ventricle function reference values in pediatrics?

Several models have been used to establish pediatric Z-scores for echocardiographic right ventricle (RV) variables. This study aimed to determine the most appropriate mathematical model for RV function reference values in pediatrics. Prospective cross-sectional study among 314 healthy children ages 2 days to 18 years (46% female, 88 infants, 26 neonates, 226 children). RV parameters (S', E', A' waves, TEI index, TAPSE) were modelized with four explanatory variables: age, height, weight and body surface area. Four mathematical models were applied: a linear model, a polynomial model of degree 2, a linear regression model by spline with free knot, and a polynomial regression model of degree 2 by spline with a free knot. The choice of the best method used the adjusted coefficient of determination (aR2). The modelization of RV variables did not follow a linear model. A single explanatory variable could not determine all Z-scores, as specific and independent variables were required for each parameter. The quadratic spline model best adjusted the modelization of RV variables. The S' wave was best modelized by the spline model with the weight (knot at 6.86 kg). The E' wave was best modelized by the spline model with the age (knot at 0.29 years). The A' wave was best modelized by the model with the height (knot at 81 cm). The TAPSE was best modelized by the spline model with the weight (knot at 9.04 kg). The spline regression models best applies to echocardiographic RV reference values in pediatrics.