Dietary patterns, global syndemic components and an exposome approach: a comprehensive data analysis of the CUME study

• Ano de defesa: 2024
• Autor(a) principal: Mattar, Jéssica Bevenuto
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Ciência da Nutrição
• Palavras-chave em Português: Obesidade; Desnutrição; Mudanças climáticas; Distúrbios da nutrição; Nutrição - Aspectos ambientais; Aprendizado do computador; CIENCIAS DA SAUDE::NUTRICAO::ANALISE NUTRICIONAL DE POPULACAO
• Link de acesso: https://locus.ufv.br/handle/123456789/32791; https://doi.org/10.47328/ufvbbt.2024.384

The pandemics of obesity, malnutrition and climate change make up the Global Syndemic, which has food systems as one of the main determinants in common. Current food systems have changed eating patterns based on natural and fresh foods to patterns rich in foods from animal source and ultra-processed foods. Researchers have debated the impact of different dietary patterns on human and environmental health, but no study has focused on the relationship between Brazilian dietary patterns and the three components of the global syndemic. Furthermore, the multiple and complex etiology of obesity, in addition to new evidence that points to relationships between climate change and obesity, reveals the need to study obesity using the exposome approach, which is the cumulative measure of environmental influences and associated biological responses throughout the lifespan. Therefore, the objective of this thesis was to evaluate the association between the dietary patterns of participants of the Cohort of Universities of Minas Gerais (CUME Study) and the components of the global syndemic, focusing on the relationship between obesity and climate change. To this end, cross-sectional and longitudinal analyzes were conducted using data from the CUME Study. Graduates from Minas Gerais universities participating in the CUME Study answered baseline (Q_0), two (Q_2), four (Q_4) and six (Q_6) year follow-up questionnaires in a virtual environment. Q_0 consists of 83 questions about lifestyle, sociodemographic, anthropometric, and clinical data, individual and family morbidity, in addition to a semi-quantitative Food Frequency Questionnaire (FFQ). The follow-up questionnaires aimed to identify changes in the data provided in Q_0. Dietary patterns were determined from FFQ data, through principal component analysis, and were associated, using the linear regression technique, with (1) undernutrition, assessed as insufficient intake of vitamins and minerals; (2) climate change, assessed by the carbon, water and ecological footprints of the food consumed; and (3)obesity, assessed by Body Mass index (BMI). Air pollution is closely related to climate change. Therefore, an assessment of the effect of long-term exposure to fine particulate matter (PM2.5) on longitudinal changes in BMI was conducted using multilevel linear mixed -effects models to assess the association between obesity and climate change. Furthermore, obesity was assessed using the exposome approach, using supervised classification models. Thus, five decision tree- based algorithms were conducted in RStudio® (4.3.3) to predict obesity using the Caret package. The algorithms performed were Random-Forest, Rpart/Cart, c5.0, Bagging and Boosting. Different partitions of the database were tested and accuracy was used as a criterion for determining the best model fit. Four dietary patterns were identified: (1)Unhealthy dietary pattern, which increased the odds of obesity and micronutrient inadequacy and presented the greatest environmental impact for the three parameters evaluated; (2)Brazilian dietary pattern, which increased the chance of obesity and micronutrient inadequacy and had the lowest environmental impact; (3)Healthy dietary pattern, which was protective against micronutrient inadequacy; and (4)Dairy dietary pattern, which had the greatest protective effect against Calcium inadequacy. No direct effect between PM 2.5 and BMI was found. However, an interaction effect between PM2.5 and physical activity on BMI was observed. In adjusted analyses, BMI was significantly higher among those physically inactive and exposed to 15-30 µg/m3 PM2.5 , compared to those who are insufficiently active. When compared with those who are physically active, those who are inactive had a significantly higher BMI when exposed to PM2.5 ≥ 15 µg/m3 . The Boosting model presented the best performance for predicting obesity. Then, a plot of the importance of variables was extracted from this model. The plot revealed that unhealthy dietary pattern is the most significant predictor of obesity, followed by age, maternal obesity, TV usage, per capita income of the census tract where the participants live, siblings’ obesity, exposure to PM2.5 , Brazilian dietary pattern, father’s obesity and individual income. This study highlights the importance of public policy actions to incorporate the costs of effects on human and planetary health, to align the Brazilian diet with the Brazilian Dietary Guideline, to control air pollution while encouraging increasing the practice of physical activity and improving the housing conditions of those living in regions of lower socioeconomic status. Keywords: Sustainable diet; Air pollution; Body Mass Index; Long-term effects; Machine learning.