This study is one of the first to examine associations between the neighbourhood physical activity environment and adiposity longitudinally, and to examine both children and adults simultaneously. Overall, few objectively assessed characteristics of the neighbourhood environment included in this study were associated with adiposity or changes in adiposity over three years among children in early and later primary school or among their female carers. Features for which there were associations generally differed by age group and there was little consistency between cross-sectional and longitudinal findings, and between the two neighbourhood scales (800 m and 2 km) adopted in this study. This suggests that different factors may be important during different life stages , and highlights the importance of careful consideration of the scale used for studies of different age groups and different behaviours. Nevertheless, in general, most associations were found for features within 800 m of home, rather than features within the wider definition of a neighbourhood (2 km), regardless of age.
Access to destinations for physical activity was found to be associated with adiposity among older children and female carers, but not younger children. The density of public open spaces designated specifically as sport/recreation was negatively associated with BMI z-scores among older children, suggesting that the provision of spaces specifically for physical activity may be important in efforts to prevent obesity in children. This finding is consistent with previous research linking access to recreational facilities with physical activity among children . Notably, the number of different sports that are popular among children that were available within the local neighbourhood was not associated with adiposity in any analyses. This measure, however, does not capture less organised forms of physical activity that may be important among children. These findings should be considered with caution since the cross-sectional finding regarding sport/recreation spaces was not replicated in the longitudinal analyses. This may reflect changes in leisure interests as children age, whereby these kinds of sports and recreation facilities may be less appropriate, relevant or desirable to mid-adolescents than to 10-12 year-olds. Consistent with the literature among adults , access to gyms/leisure centres and swimming pools within 2 km was associated with relatively smaller increases in BMI in longitudinal analyses among female carers. Provision of facilities for leisure activities popular among women, combined with consideration about the cost of using those facilities particularly in disadvantaged areas , may therefore be an important strategy for obesity prevention. However, cross-sectionally, length of walking/cycling tracks within 800 m of home was associated with higher BMI among women in this study - an unexpected finding that requires further exploration.
Road connectivity is thought to be an important component of the walkability of neighbourhoods because neighbourhoods that are better connected reduce travel times by foot or cycle and offer a greater diversity of routes to local destinations without adding substantially to travel times . In this study, intersection density and proportion of intersections that were at least 4-way, both indicators of connectivity , were longitudinally associated with relatively greater decreases in BMI z-scores over the three years among younger children, but not among older children. Similar findings have been found among children of a similar age in Canada , while Frank and colleagues  found intersection density to be a correlate of walking behaviour among adolescents but not children. Few studies have examined associations between similar measures of connectivity and physical activity behaviour among young children . More research is needed to better understand relationships between connectivity and adiposity in this age group and mediators of this relationship. Although connectivity is a well-known correlate of walking behaviour among adults , no associations were found with adiposity outcomes among female carers in this sample. Length of access paths is also a measure of connectivity, since these paths facilitate pedestrian travel by going under or over major roads, and making it quicker and less dangerous to reach destinations . In this study, access paths were negatively associated with BMI z-scores among both groups of children, suggesting that living in areas with many shortcuts or cut-throughs is important among children, possibly making it safer for children to walk locally. However, access paths were not associated with change in BMI z-score longitudinally in the younger children and became associated with greater relative increases in BMI z-scores among older children. This inconsistent finding needs further study.
Although the total amount of 'busy' roads within neighbourhoods was unrelated to adiposity outcomes in children, cross-sectionally greater length of local roads were associated with lower BMI z-scores among the older children. This finding is consistent with earlier cross-sectional findings in the same sample of children where the older children, but not the younger children, were more likely to be overweight or obese if their parents perceived there to be heavy traffic in their local streets and more likely to be obese if their parents agreed that road safety was a concern . Given that older children have greater independence, are more likely to walk or cycle in their neighbourhoods [16, 33] and are therefore more likely to be exposed to traffic, it is not surprising that associations were found for older children and not younger children who may be more likely to be accompanied by an adult or driven. These findings suggest that efforts to slow traffic in residential areas may be important for obesity prevention in children. However, again, this cross-sectional finding was not replicated longitudinally. One possible explanation is that as children age they gain greater competence as pedestrians or cyclists and traffic levels may become less relevant. They may also be more likely to use streets for travel rather than as a venue for play where traffic levels would be of greater importance. In contrast to the cross-sectional findings among children, length of busy roads was negatively associated with BMI among female carers. This may be because areas with more busy roads reflect a more connected street network, though there were no associations with other measures of connectivity in this study. Alternatively, the negative association between length of busy roads and BMI among female carers may be confounded by area-level socioeconomic status, whereby areas with busier roads may be more likely to be situated in inner-city areas and may potentially have higher levels of advantage. The relationship between socioeconomic status and obesity risk is well documented [35, 36].
The strengths of the present study include the use of a range of objective indices of environmental variables thought to influence physical activity, its prospective design, its use of unique neighbourhoods specific to individual residential addresses rather than administrative definitions , and the inclusion of children and adults exposed to the same neighbourhoods. The latter strength, however, also poses challenges in defining 'neighbourhoods' and thus two definitions were used here to accommodate both children and adults. It is possible that the 2 km radius may have been too large, even for adults, and may have attenuated any impact of the local environment on obesity, which may explain the null relationships between many of the attributes of the environment examined here and adiposity using this neighbourhood definition. In adults, for example, it has been suggested that the 'neighbourhood' be restricted to a 15 minute walk of 1.6 km . Given parental concern for children's safety, however, a neighbourhood that might encourage children's physical activity may be considerably smaller. Indeed, more associations were found for features measured within 800 m than within 2 km of home.
The study is limited by the relatively small sample size. Further, this study examined only physical features of the environment; it is possible that other features of the environment for which we do not have archival quantitative measures, such as the quality of destinations, may be stronger determinants of obesity. Moreover, perceptions of the environment may be stronger predictors of behaviour than objective variables . In addition, given that relatively few significant associations were found, it is possible that some associations may have been due to chance given the large number of statistical tests that were conducted for each age group. The results may also be masked by gender differences that were unable to be explored due to the small sample sizes. Although the longitudinal design is a strength, the lack of consistency between cross-sectional and longitudinal findings requires further exploration and may be due to changes in the saliency or importance of specific neighbourhood features as individuals age, mature and gain greater autonomy over what could be considered a substantial period of time among children. Alternatively, it is possible that the neighbourhood environment may have undergone change during the three years to follow-up. These issues should be considered in future prospective studies in order to establish causality. Finally, this study did not consider attributes of the environment that may be related to the other side of the energy balance equation, energy intake. Future studies should simultaneously consider the impact of neighbourhood physical activity and food environments on adiposity.