Intended for healthcare professionals


Early life risk factors for obesity in childhood: Early feeding is crucial target for preventing obesity in children

BMJ 2005; 331 doi: (Published 18 August 2005) Cite this as: BMJ 2005;331:453
  1. Iain E Buchan, senior lecturer (buchan{at},
  2. Richard F Heller, professor,
  3. Peter Clayton, professor,
  4. Peter E Bundred, reader,
  5. Tim J Cole, professor
  1. Medical School, University of Manchester, Manchester M13 9PT
  2. Medical School, University of Manchester, Manchester M13 9PT
  3. Department of Primary Care, University of Liverpool, Liverpool L69 3GB
  4. Institute of Child Health, University College London, London WC1N 1EH

    EDITOR—Reilly et al present their findings on some early life risk factors for obesity at age 7 in a cohort of 7758 children born in the early 1990s.1 We believe that their conclusions about breast feeding and time of weaning are flawed.

    The investigators infer that these early feeding factors are unimportant because they are not independently associated with the obesity outcome in a statistical model that includes more proximal, parental, and prenatal factors. If early feeding factors exert their influence on childhood obesity largely through a causal pathway that includes other factors in the investigators' statistical models then it is not surprising that the early feeding factors do not emerge as independent predictors: it is a mistake to infer that these factors are therefore unimportant. The central public health issue is to identify the modifiable risk factors that can reasonably be assumed to be causal for obesity in children.

    We believe that there is sufficient bio-behavioural and epidemiological evidence to support the idea that infant feeding practice directly affects early weight gain, and that trials of early feeding interventions to prevent later obesity in children are justified.25 The investigators' univariate findings that early feeding factors predict later obesity add to this evidence, and their multivariate results should not be used to weaken the evidence on grounds of specious confounding. These data could be reanalysed with more attention paid to plausibly causal pathways.


    • Competing interests None declared.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.