Umeå University's logo

Objective: Researches which studied the relation of dietary greenhouse gas emissions with health outcomes are few, inconsistent and most of them are modelling studies which have not investigated empiric dietary emission patterns. In this study, we employ a posteriori data dimension reduction method, treelet transform, to identify dietary and diet related emission patterns concurrently. We aim to evaluate if these patterns are correlated, if they areassociated with diabetes and if emission patterns can be used as a proxy for dietary patterns for assessment of association with diabetes.

Design: Food items from dietary questionnaire were aggregated to 34 food groups. GHGE was estimated by linking food intakes with life cycle assessment data on emission. Dietary and emission patterns were identified by employing treelet transform on food intake and corresponding greenhouse gas emission data. Multivariate logistic regression was performed to investigate associations between quintiles of dietary patterns and diabetes. Adjusted mean values of emission estimates were obtained for the identified dietary patterns. Adjusted proportions of diabetes across quintiles of emission patterns were computed.

Setting: Västerbotten Intervention Program

Subjects: women (n 38,118); men (n 36,042) between the age of 35 and 65 years

Results: Four dietary and four corresponding emission patterns in women, five dietary and five corresponding emission patterns in men were identified. Moderate to strong correlations were observed between dietary and corresponding emission patterns. Prudent dietary pattern (PP) in women was inversely associated with dysglycemia [ORQ5 vs. Q1 = 0.82 (95% CI 0.69—0.97, Ptrend =0.003)]. PP in women was also inversely associated with diabetes [ORQ5 vs.Q1 = 0.37 (95% CI 0.17—0.78, Ptrend = 0.002)]. However, adherence to this dietary pattern was associated with higher dietary emission. Finally, none of the corresponding emission patterns, were associated with adjusted proportions of either dysglycemia or diabetes.

Conclusion: Treelet transform produces correlated dietary and emission patterns which are sparse and easily interpretable. However, some differences in loading structures between dietary and emission patterns result in different conclusion regarding the association with diabetes, rendering the usage of emission patterns as proxies of dietary patterns inappropriate. Results from our study also show that healthy dietary patterns do not necessarily reduce greenhouse gas emission.