This study comprehensively examined associations between the school food environment and student consumption, and in turn, associations with BMI. We found that the availability of SSBs and nutritional practices at school were associated with consumption of SSBs, but no associations with consumption of other less healthful foods. We also found that both the availability of SSBs and less healthful foods were associated with student BMI; although these associations differed by BMI category (overweight versus obese). Specifically, the association with SSBs was only observed among the obese adolescents, whereas an association with less healthful foods was only observed among the overweight adolescents. Our findings add to the limited and inconsistent findings in this area  and provide further support for improving the school food environment to reduce childhood obesity.
Similar to previous studies [18, 19, 28], the availability of SSBs and less healthful nutritional practices at school were both associated with greater SSB consumption highlighting the importance of schools in promoting healthy dietary habits. In contrast, Taber et al. reported that reduced access to SSBs at school reduced purchasing but not overall consumption of SSB . Unlike previous studies [18, 19, 28, 30], associations in this study were observed for SSB consumption only and no associations were found between the other school food environment variables (policies, programs, resources, support, availability) and the Food Consumption Index. Our findings may indicate a need for attention to specificity in food environment measures or reflect limitations in how the Food Consumption and Food Availability Indices were measured.
Access to SSBs at school and their consumption were both associated with obesity providing further support for targeting schools to help address adolescent obesity [20, 31]. While the association between SSB consumption and BMI is supported by a recent review , the association was present for obese but not overweight adolescents in our study. It is possible that access to SSBs in the school setting may disproportionately affect students who come from a less healthy home environment as they likely consume SSBs both at home and school.
Interestingly, we found an association with our Food Consumption Index and weight, but only for overweight adolescents compared to normal weight adolescents. This association was somewhat expected as a review by Perez-Escamilla  supports an association between energy density and BMI in children and adolescents; however, it is somewhat inconsistent that we observed this relationship only among those who were overweight and that consumption of SSBs appeared most related to BMI among obese adolescents. This difference may relate to the limitations of our methodology and is discussed further below.
With respect to the covariates we included in our model, similar to other studies boys were more likely to consume SSBs [34, 35], consume energy dense foods , and to be overweight and obese than girls [1, 35]. Boys are thus a vulnerable group that may benefit from changes to the school food environment to a greater extent than girls might. Similar to other studies, we found that adolescents from more disadvantaged neighborhoods also consumed more SSBs [34, 35] and energy dense food  and that students in suburban and rural schools were more likely to be overweight and obese [37, 38]. These findings highlight groups of adolescents who may be differentially affected by creating healthier eating environments at school.
Our findings should be interpreted in light of the context in which the data were collected. Unlike the US, Canada does not have a national breakfast or school lunch program that is subsidized by the federal government . In BC, while subsidies for school lunch targeting students in need can be obtained from the provincial government, guidelines to regulate the school food environment were first written in 2005, but full implementation was only expected in 2008, after the data were collected for this paper. Interestingly, even in such a varied context and with a guideline in place (albeit without an accountability mechanism) many of our findings echo what others have found in the US [28, 30]. Given that the school food environment of Canadian schools was quite different than that of US schools, it remains important to understand whether research from the US context translates into other jurisdictional contexts.
Finally, the study limitations are important to consider when interpreting our results. First, the cross-sectional nature of the data limits our ability to make causal inferences. Second, we used self-report to measure student consumption and BMI, and although commonly used in large studies, are known to be associated with measurement errors that can mask or dampen existing associations. Third, many principals did not complete the survey and half the schools required written parental consent resulting in lower student participation; therefore, we do not know how non-response bias may have affected the results. Fourth, although we evaluated the psychometric properties of our school food environment measures, they were developed or adapted from other measures to fit the BC context. Fifth, we utilized an established measure for the availability of food and beverages at school;  however, the measure did not identify if healthier versions of specific food were offered. An unpublished government review suggests that little change in the school food environment had occurred before the full implementation of the first food guidelines were expected in schools which was after we collected the data. Sixth, we highlighted any effects that were significant at a p < .05 but some effects might be less stable as they were not significant at a p < .01 as well (i.e., the association between less healthful foods and overweight adolescents). Given the exploratory nature of our analyses, all of our findings should be replicated. Seventh, each province and territory in Canada has different policies/guidelines affecting the food environment of public schools, and without a federally subsidized school breakfast/lunch program, the generalizability of our findings to other jurisdictions is limited. Finally, measuring consumption over the entire day limits our ability to determine associations with school-specific consumption. This is important as recent data among US children showed a shift in the amount of energy intake obtained from school sources to fast food places . A better understanding of where students’ food purchases occur may shed further light on these findings.