• Body mass index;
  • cohort studies;
  • maternal smoking;
  • risk factors;
  • secular changes


What is already known about this subject
  • Factors associated with children's body mass index (BMI) include parents’ BMIs, birth weight, maternal smoking, sleep duration and television watching.
  • Few studies have attempted to quantify either changes in the association between risk factors and BMI or the contribution of changes in the risk factors to increases in BMI over a generation.
What this study adds
  • The magnitude of the association between most risk factors and children's BMIs has not changed over a 29-year period.
  • Increases in the population level of mothers’ body mass index (BMI) explains ∼20% of the increase in children's BMI whereas the smaller increase in fathers’ BMI contributes only 6%.
  • Maternal smoking, despite the decrease in prevalence, contributes ∼17%.


Using two cohorts born 29 years apart in Dunedin, New Zealand we aim to examine changes in risk factors and their associations with body mass index (BMI) at ages 3 and 7 years, and estimate their contribution to the secular changes in BMI at age 7 years.


Birth weight and anthropometric measures at ages 3, 5 and 7 years were obtained for 974 participants in the Dunedin Multidisciplinary Health and Development Study (DMHDS), born in 1972–1973, and 241 in the Family Lifestyle, Activity, Movement and Eating Study (FLAME), born in 2001–2002. Information about maternal age, education and smoking in pregnancy, as well as breastfeeding, children's television time and time in bed, was obtained by questionnaire.


The increase in BMI over the 29-year period was 0.84 (95% CI 0.61, 1.06) kg m−2 at age 7. A 1-unit difference in the mother's BMI was associated with a 0.06 (0.03, 0.08) kg m−2 difference in offspring in both studies; the 3.4 (2.8, 4.0) kg m−2 increase in the mothers’ BMIs accounts for a change of 0.19 kg m−2 in the children's BMI. The much smaller generational increase in fathers’ BMI (0.7 kg m−2) correspondingly had a more limited effect on change in child BMI over time (0.06 kg m−2). Although smoking in pregnancy decreased by 15% (8, 21) its association with BMI increased from 0.20 (−0.01, 0.42) in the DMHDS cohort to 1.24 (0.76, 1.71) kg m−2 in the FLAME cohort, contributing 0.18 kg m−2 to the increase in children's BMI.


Societal factors such as higher maternal BMI and smoking in pregnancy contribute most to the secular increase in BMI, with changes in behavioural factors, including sleep and television viewing, having little effect in this setting.