A semi-mechanistic model for estimating fat free mass in children
Hesham S Al-Sallami (1), Ailsa Goulding (2), Rachel W Taylor (2), Andrea M Grant (2), Sheila M Williams (2), Stephen B Duffull (1)
(1) School of Pharmacy, University of Otago, Dunedin, New Zealand; (2) School of Medicine, University of Otago, Dunedin, New Zealand
Objectives: Body size correlates with clearance and can be used to scale drug doses. Lean body weight (LBW) has been proposed to be a better size descriptor than other measures of weight. Mathematical models for estimating LBW (approximated by fat free mass, FFM) have been developed in adults. There are currently no models available to predict in children. The aim of this project is to develop a semi-mechanistic model to quantify FFM in paediatric patients.
Methods:
- A general model for maturation was developed for FFM using NONMEM VI. An index dataset (496 females and 515 males) containing demographic data and body composition measurements was used to estimate model parameters. Missing data were imputed.
- An empirical model for FFM was developed using STATA 11.
- The predictive ability of the adult model (Janmahasatian et al, 2005) and the general maturation model were evaluated with respect to the empirical model using the mean squared error (MSE).
- A test dataset (90 females and 86 males) was used to evaluate the general maturation model.
Results:
- A semi-mechanistic sigmoid Emax maturation model was developed:
FFMchildren = FFMbaseline + AGE^GAMMA/(AGE^GAMMA + AGE50^GAMMA) x (FFMadults - FFMbaseline)
- An empirical model with 9 terms (including interactions) was developed using mixed-effect linear regression.
- Using the index dataset, the adult model had a variance of 15 kg2 whereas the maturation model had a variance of 12 kg2. The increment in MSE using the adult model in relation to the empirical model (which had a variance of 6 kg2) was 146%; the increment in MSE using the maturation model in relation to the empirical model was 99%.
- Using the test dataset, the adult model had a variance of 16.5 kg2 whereas the maturation model had a variance of 12.2 kg2. The increment in MSE using the adult model in relation to the empirical model (which had a variance of 8.5 kg2) was 94%; the increment in MSE using the maturation model in relation to the empirical model was 44%.
Conclusions: The adult model provided an unbiased descriptor of FFM in children. The general model for maturation for FFM provided a more precise estimate of FFM in children than the adult model. The loss of predictive performance was significantly less for the general model for maturation compared to the adult model for both internal and external evaluation.
References:
[1] Janmahasatian et al. Clin Pharmacokinet 2005; 44(10): 1051-1065.