... Evidence of effects for traffic density was found in unadjusted models. This effect was confounded in fully adjusted models, although the effects did remain elevated.

Comparing children in the highest 10% of traffic-related air pollution exposure to those in the lowest 10% of exposure yielded a 0.39 BMI unit increase in the attained BMI level at age 10. This translated into a 13.6% increase in the rate of average annual BMI growth. These effects may have large population impacts because traffic-related air pollution is a ubiquitous exposure that affects billions of people globally, and in many countries traffic is increasing at a higher rate than the rate of population growth.

Examining the effects at different times during the follow up helps to interpret the results. As the children get older, the effects accumulate, and the slope difference between the lowest and highest deciles becomes more pronounced. By age 10 or 11 the difference is about 0.4 – 0.5 of a BMI unit.

... Traffic-related air pollution nonetheless was not confounded by other variables, suggesting that air pollution exerted a stronger effect on BMI growth than traffic density. This result was insensitive to which individual and neighborhood built environment confounding variables were used in the model. Based on the sensitivity analyses, variables at the school and community level do not confound the association between BMI and traffic-related air pollution.

The findings here differ from the only other study that examined the impacts of traffic density on BMI growth, which was conducted among an older cohort of children in 10 of the same study communities and used similar statistical techniques. With the same metric of traffic density within 150 m around the home, the earlier study found significant effects that were not confounded by other individual or built environment variables or community-level variables such as poverty. This difference in findings from the two cohorts may have resulted from mobility differences by age. Most of the children in the present analysis were less than 10 years old for most of the follow up, and children of this age are less likely to walk on their own than older children who were followed for the earlier research.

Qualitative research suggests that parents of children aged less than 10–11 perceive many barriers to allowing children to move freely in urban areas, but the same study indicates that at this age, which corresponds to the end of primary school, parents do begin to afford increased license to engage in physical activity alone or more likely in groups of peers. Quantitative research using global positioning systems to track children supports the qualitative research, indicating that there is a large rise in the proportion of children allowed to range freely around the ages of 10–11. Therefore, the pathway of reduced physical activity from traffic danger in younger children may be less pronounced in older children, because fewer of the younger children exhibited independent mobility on average. The earlier study on traffic density did not test for associations with traffic-related air pollution.

Reliance on the CALINE4 dispersion model limited our ability to discern which elements of the traffic pollution mixture were most important. Although we used NOx as our indicator of traffic-related pollution, this molecular gas had strong correlations with CO, NO2, and PM2.5 estimates from the CALINE4 model, with correlations greater than 0.9. We found non-freeway NOx had the association, while freeway NOx was not robust to confounders. We interpret the lack of effect from the freeway NOx as resulting from a small proportion of the total cohort who lived in proximity to freeways, rather than an attribution to a specific source from a different type of roadway. While the results indicate that traffic-related pollution likely has a stronger effect than traffic density, we are unable to identify which specific components of the traffic mixture were responsible for the effects.

Another limitation of this study related to the lack of information on food intake. Food access was controlled in the models, but dietary factors could not be directly evaluated. Given what is known about the association between lower socioeconomic position and higher traffic-related pollution exposures in California, some of the effects observed here may be confounded by dietary variables that are also associated with lower socioeconomic status, such as intake of sugar and fats. Socioeconomic status in the home and neighborhood was controlled for, which reduces the chance of residual confounding relating to socioeconomic status, but confounding by food intake, which might be associated with air pollution through socioeconomic status, cannot be directly ruled out.

To address the concern about diet, information on dietary intake in an older cohort (ages 10–18) of nearly 2000 children in 10 of the same study communities as in the current study was used to generate variables on macronutrients including total caloric, protein, carbohydrate, saturated fat, mono unsaturated fat, and cholesterol intake. A statistical analysis that controlled for community of residence, race, sex, and parental education as a marker of SES was performed, and for a wide array of traffic or traffic-pollution indicators there was no association between the total caloric intake and the traffic-pollution estimates or traffic density measures. A weak, borderline significant association between daily grams of carbohydrate consumption and nitrogen dioxide from non-freeway sources was observed, but the coefficient was very small. Equivalent diet information on the specific cohort used in our paper is not available, but the relationships between the traffic or pollution variables and food intake should be similar in both cohorts. Given that there was no difference in total calories or in any other macronutrient categories, the chance of confounding by unmeasured diet variables is limited.

Although we cannot rule out self selection of potentially more health-conscience families into areas with lower pollution, our mixed effects modeling framework properly controls for baseline BMI. As a result, the influence of self-selection is accounted for with subject-to-subject variability due to baseline characteristics. While self-selection could influence the trajectory, control for baseline characteristics that is inherent to our modeling framework makes it more likely that our results are from an ongoing influence of the environment and not some other factors.

The effects of pollution are significant, and the temporal pattern is consistent with the hypothesis that the inflammatory effects of air pollution predispose children to obesity in a similar way to what has been observed in laboratory experiments. By analogy, this pattern is also corroborated by human epidemiological studies finding associations between metabolic disorders and air pollution. Another explanation is possible; in areas of high traffic, children and their parents may have a heightened sense of danger that reduces activity by restricting the mobility of families. In this cohort, however, traffic effects were not significantly associated with BMI growth or attained level after controlling for confounding variables. There are several other pathways from stress resulting from noise or from other obesogens, which could be leading to higher BMI growth in children, but we are unable to test such pathways directly. Future research may usefully address these other pathways along with traffic pollution exposures.

A la fin de ce travail sur cette publication, quelles conclusions en tireriez-vous ? Comment pourriez-vous en tirer des recommandations en termes d'aménagement, de politiques publiques en matière de santé, de transport... ?

En vous basant sur votre travail, quelle(s) expérience(s) ou étude(s) complémentaire(s) proposeriez-vous afin de contribuer à établir le lien entre pollution issue du trafic et obésité ?