Differences in Tsimane children's growth outcomes and associated determinants as estimated by WHO standards vs. within-population references.

Anthropometric measures are commonly converted to age stratified z-scores to examine variation in growth outcomes in mixed-age and sex samples. For many study populations, z-scores will differ if calculated from World Health Organization (WHO) growth standards or within-population references. The specific growth reference used may influence statistical estimates of growth outcomes and their determinants, with implications for biological inference. We examined factors associated with growth outcomes in a sample of 152 Tsimane children aged 0-36 months. The Tsimane are a subsistence-scale population in the Bolivian Amazon with high rates of infectious disease and growth faltering. To examine the influence of growth reference on statistical inferences, we constructed multiple plausible models from available infant, maternal, and household attributes. We then ran identical models for height-for-age (HAZ), weight-for-age (WAZ), and weight-for-height (WHZ), with z-scores alternately calculated from WHO and robust Tsimane Lambda-Mu-Sigma growth curves. The distribution of WHO relative to Tsimane HAZ scores was negatively skewed, reflecting age-related increases in lower HAZ. Standardized coefficients and significance levels generally agreed across WHO and Tsimane models, although the strength and significance of specific terms varied in some models. Age was strongly, negatively associated with HAZ and WAZ in nearly all WHO, but not Tsimane models, resulting in consistently higher R2 estimates. Age and weaning effects were confounded in WHO models. Biased estimates of determinants associated with WHO HAZ may be more extreme in small samples and for variables that are strongly age-patterned. Additional methodological considerations may be warranted when applying WHO standards to within-population studies, particularly for populations with growth patterns known to systematically deviate from those of the WHO reference sample.