Analyse d'une publication en épidémiologie

The mean age of children at study entry was 6.6 years (+/- 0.65; range 4.5-8.9). Average BMI was 16.79 at study entry (+/- 2.81). By year 5 of the study BMI had increased approximately 2.6 units to 19.35 (+/- 4.21) with boys showing a slightly greater increase. Based on Centers for Disease Control percentiles between the 85th and 95th percentile, rates of overweight were 14.4%. Obesity rates measured as BMI scores equal to or greater than the 95th percentile were 15%.

Traffic density at 150 m radius had a positive, but borderline significant (p <0.1) association with the intercept and the slope of BMI growth curves of the children. Further evaluation of the traffic effects revealed that they were confounded by other variables, particularly whether the questionnaire had been completed in Spanish, suggesting the child was from a family of recent immigrants from Latin America. A final model included asthma status of the child, the language used to complete the questionnaire (Spanish or English), whether the child was exposed to second-hand smoke in the home, the parental level of education, the gamma index (a measure of the connectivity of the street network around the child's home which affects walking distances), the number of fast food outlets within 500 m of the child's home, greenness around the home as measured by the normalized difference vegetation index, the number of active recreational programs for children offered within 5 km of the home, and traffic density at 150 m. In this fully-adjusted model, the effect of traffic within 150 m remained positive on the slope, but was reduced by more than 20% by the confounders and was no longer borderline significant. Of note, we tested several variables measuring various aspects of physical activity or participation in sports, but none of these variables met our inclusion criteria for confounding.

In the screening of the air pollution variables, non-freeway NOx levels were significantly and positively associated with BMI at age 10 and the rate of growth over the four year follow-up period, while the freeway-related exposures were not associated with BMI growth, consistent with other previous studies on respiratory health. The association between BMI and non-freeway NOx was reduced but remained significantly elevated in models containing the same variables as those in the fully adjusted traffic density model described above and with those chosen specifically to confound NOx. Again none of the physical activity variables met the inclusion criteria as confounders. Interaction by gender was tested, but no significant evidence of difference in the effects on boys and girls was found.

Confounders at the community and school levels were further tested by including the average terms for each level and the individual deviations from the mean of the level. Neither community level crime nor poverty confounded the within-community effect of air pollution. The impact of the school level was then tested by including a fixed effect for school in the model, but air pollution remained significantly and positively associated with BMI growth with little change in the coefficient. This suggests that the school level variables did not confound the air pollution effect on BMI growth.