Ultra-processed foods and cardiometabolic health

The overall aim of my thesis was to investigate the associations between ultra-processed food consumption (UPF) and cardiometabolic health at the individual and the population-level while adhering to methodological principles such as incorporating and presenting a multiverse of statistical results and interpreting statistical hypothesis testing in a non-dichotomous way. I addressed these aims in three studies: a systematic review and meta-analysis of 41 prospective cohort studies, an ecological, longitudinal cross-country comparison, and a prospective analysis in the UK EPIC-Norfolk cohort. The research presented in this dissertation revealed consistent associations between UPFs and adverse cardiometabolic health. The results of the systematic review and meta-analyses of chapter 2 provided the first systematic analysis of published nutritional epidemiology studies from the perspective of food processing. I defined UPF consumption more broadly and identified studies in which UPF consumption had been termed in several different but related ways, such as fast, convenience, or Western foods. I combined those diets and dietary pattern studies that were characterized by a higher relative intake of UPFs in non-linear and summary random-effects meta-analysis estimates. Higher intakes of UPFs were associated with both an increased risk of cardiovascular diseases (CVD) and type 2 diabetes mellitus (T2DM). Higher sales of UPF were associated with an increased risk of adiposity and diabetes mellitus in the country-level analysis in chapter 3, using data from 76 countries across all five continents over 16 years. The panel analysis demonstrated that a strong and consistent association existed between the sales of UPFs at the food system level and adiposity and diabetes prevalence in low-to-middle-income countries (LMICs) for children, adolescent, and adult populations, as well as for both sexes separately. This finding adds value to the literature because no previous study had systematically investigated these associations in countries in which often a lack of individual data exists and had estimated associations for LMICs and high-income countries (HICs) separately. However, the analysis did not establish an association between UPF and adiposity in HICs, which was surprising, given that previous studies indicating an association at the individual level were mostly from HICs. The lack of variability in UPF data from HICs during the study period were likely a key reason for the lack of estimated associations in HICs. As a consequence of the many combinations of data processing and analytical methods and the variability of point estimates and P-values, no particular set of point estimates was emphasized, but the consistency of findings indicate that the expansion of global adiposity and diabetes since 2000 can partially be attributed to the increased sales of UPFs, at the level of the food system or country. The associations that were found in the meta-analysis and the panel study were replicated in prospective analyses of detailed data from over 17,000 individuals in the EPIC-Norfolk cohort, yielding associations between increased UPF intake and adiposity as well as risk of T2DM and CVD. This study demonstrated that the way UPF intake is operationalized can fundamentally influence results. Previous UPF studies had only expressed UPFs as weight or the proportion of food weight, with the justification that energy measures would not capture the non-calorific components of UPFs with potentially adverse effects on health. Guided by previous nutritional epidemiological research to adjust for total energy intake, I modelled UPF and disease risks in five different ways. The analyses revealed that these different approaches affect the statistical results, but also that they affect the results differently for different diseases. For example, the differences between measures of UPF intake based on energy and weight were much more pronounced for T2DM as an outcome than for CVD, suggesting that potentially different mechanisms relating to dietary energy and other factors common to UPFs might be responsible for different outcomes. Chapter four was also the first research to comprehensively test the associations between UPF and three important cardiometabolic disease outcomes (adiposity, T2DM, and CVD) in a prospective cohort with a long follow-up period. The findings of an almost consistent positive association provide the strongest evidence to date that UPFs are positively associated with adverse risk of cardiometabolic health. The secondary analyses of eight different food groups and outcomes indicated that greater consumption of ultra-processed meat, fish, and eggs; fast foods; and SSBs are associated with an increased risk of T2DM and CVD, whereas consumption of ultra-processed fruits and vegetables, milk and diary, fats, and breads and cereals might not be associated with an increased risk of disease. Bodyweight is likely a very important mediator of the association between UPF and T2DM and CVD, while increased total energy intake through UPFs is likely the most important driver of the UPF-adiposity association.