Determinants for physical growth patterns in low- and middle-income countries
E. Niclas Jonsson (1), Joakim Nyberg (1), Jonas Häggström (2), Lifecycle Auxology & Neurocognitive Development team (3), representing the Healthy Birth, Growth and Development knowledge integration (HBGDki) community.
(1) Pharmetheus AB, Uppsala, Sweden, (2) HDC, Stockholm Sweden, (3)
Objectives: The objective of growth monitoring in children is early detection of growth failure, to allow timely remedial interventions and prevention of further growth failure. Our goal was to identify predictors for physical growth between 0-19 years of age in low and middle-income countries.
Methods: Data from low- and middle-income countries (LMIC) was used to build a longitudinal non-linear mixed effects model for height-for-age Z-scores (HAZ) (heights which are age normalized to WHO standard heights) between 0 to 19 years of age. Approximately 135000 observations from 20500 subjects spread across five different countries were used in the analysis.
An inverse Bateman model with a time changing baseline was used to describe the general trend of initially declining HAZ scores followed by a longer-term recovery phase. Between site differences and between individual differences were accounted for by site specific fixed effects and individual specific random effects, respectively. The influence of the 26 covariates on the five model parameters was investigated using a modified SCM  search algorithm supported by PsN . The relevance of the covariates in the final model was quantified in terms of their impact on phenotypic features of the predicted age-HAZ responses, for example maximum attained growth failure, ie., minimum HAZ (nadir), time of nadir and rate of growth recovery.
Results: The mean HAZ score at birth for a typical child across all sites was estimated to be -0.21. The half-life of decline to the nadir of -2.2 at age 2.6 years was estimated to 1 year. The half-life of recovery from nadir to a HAZ score of -1.16 at age 15 was estimated to be 3 years.
The final model included 17 parameter-covariate relationships (11 unique covariates) and the covariates with the largest impact on the predicted growth curves were birth weight, length of maternal education and maternal height. For example, the predicted nadir for a child to a tall mother (163 cm) compared a child to a short mother (146 cm) was 26% smaller. Other covariates in the final model included degree healthcare utilization, social class and birth order.
Conclusions: The model successfully described the growth trajectories across the five sites. The identified covariates can be used to better understand the determinants for physical growth development in low- and middle-income countries, and the model can be used to support the design of future studies.
Sponsored by the Bill & Melinda Gates Foundation, Healthy birth, growth and development initiative
 Automated covariate model building within NONMEM. Jonsson EN, Karlsson MO. Pharm Res. 1998 Sep;15(9):1463-8.
 Lindbom L, Ribbing J, Jonsson EN. Perl-speaks-NONMEM (PsN)--a Perl module for NONMEM related programming. Comput Methods Programs Biomed. 2004 Aug;75(2):85-94.