Three estimates of the association between linear growth failure and cognitive ability

Trop Med Int Health. 2009 Sep;14(9):1020-4. doi: 10.1111/j.1365-3156.2009.02321.x. Epub 2009 Jun 22.


Objectives: To compare three estimators of association between growth stunting as measured by height-for-age Z-score and cognitive ability in children, and to examine the extent statistical adjustment for covariates is useful for removing confounding due to socio-economic status.

Methods: Three estimators, namely random-effects, within- and between-cluster estimators, for panel data were used to estimate the association in a survey of 1105 pairs of siblings who were assessed for anthropometry and cognition. Furthermore, a 'combined' model was formulated to simultaneously provide the within- and between-cluster estimates.

Results: Random-effects and between-cluster estimators showed strong association between linear growth and cognitive ability, even after adjustment for a range of socio-economic variables. In contrast, the within-cluster estimator showed a much more modest association: For every increase of one Z-score in linear growth, cognitive ability increased by about 0.08 standard deviation (P < 0.001). The combined model verified that the between-cluster estimate was significantly larger than the within-cluster estimate (P = 0.004).

Conclusion: Residual confounding by socio-economic situations may explain a substantial proportion of the observed association between linear growth and cognition in studies that attempt to control the confounding by means of multivariable regression analysis. The within-cluster estimator provides more convincing and modest results about the strength of association.

Publication types

  • Comparative Study

MeSH terms

  • Child
  • Child Development / physiology*
  • Cognition / physiology*
  • Epidemiologic Methods
  • Female
  • Growth Disorders / epidemiology
  • Growth Disorders / psychology*
  • Humans
  • Male
  • Philippines
  • Psychometrics
  • Socioeconomic Factors