Type 2 diabetes (T2D) is a complex metabolic disease which affects more than 500 million people worldwide, imposing enormous burdens on affected individuals and their families, healthcare systems and societies. Healthy diets play a crucial role in preventing T2D and meat has been reported to be associated with an increased risk of T2D. However, it is unclear whether different types of meat are all associated with increased risk and whether associations are the same in all individuals and in all populations. Finally, there are uncertainties about the causal nature of the association and the mechanisms that may underlie it. In my PhD, I aimed to investigate these uncertainties in analyses of epidemiological studies.

The initial elements of my work focused on refining measures of the exposure (meat intake) and the outcome (incident T2D) in the EPIC-Norfolk study, a population-based cohort study of over 25,000 participants. I worked on improving case ascertainment of T2D as the primary outcome in my analyses. I updated T2D case ascertainment in EPIC-Norfolk by linkage of multiple external data sources, including diabetic eye screening data and clinical biochemistry data. I identified over 2,000 additional incident diabetes cases. I then reported the association of self-reported intake of different types of meat with T2D in EPIC-Norfolk.

Dietary biomarkers can provide complementary information about diet-disease associations. I used untargeted metabolomics profiling to derive metabolite scores to quantify the consumption of red meat, processed meat and poultry based on 781 circulating metabolites and 7-day diet diary data in 11,432 participants in EPIC-Norfolk. The best performing score was for red meat, comprising 139 metabolites and accounting for 17% of the explained variance of red meat consumption. Eleven top-ranking metabolites that were included in the red meat score were validated in a trial conducted by collaborators in Lyon, France. These metabolites were mainly classified into groups of lipids, amino acids, and xenobiotics, such as plasmalogens, trimethylamine N-oxide, and stearoylcarnitine. I then showed that this red meat metabolite score was strongly associated with T2D incidence in EPIC-Norfolk.

I then investigated the potential causal roles of these eleven red meat-related metabolites in T2D incidence by conducting Mendelian randomisation (MR) analyses. I observed weak evidence of possible causal associations between meat-related metabolites and incident T2D, possibly due to limited power and weak genetic instruments.

In an analysis in two large studies (EPIC-InterAct and UK Biobank), I evaluated whether the association between meat consumption and T2D incidence differed in sub-populations with varying genetic and clinical baseline risks. I found that meat intake was associated with incident T2D independently of genetic and clinical predisposition to T2D. This suggests that there are benefits of reducing meat intake on T2D burden in the entire population.

Finally, I examined associations between types of meat intake (red meat, processed meat and poultry) and T2D risk based on a federated platform in the InterConnect, which enabled harmonised data analysis of 1.5 million individuals from 23 studies across the world. This meta-analysis of individual participant data provided unique evidence of meat-T2D associations in previously understudied populations, such as those in East Asia, and East Mediterranean. I included over 60,000 new-onset T2D cases with a median of 13 years of follow-up showing that consumption of red meat, processed meat and poultry were each individually associated with increased risk of T2D.

In summary, my work provides strong evidence on the consistency of the association of meat consumption with T2D risk in sub-groups within European populations and also across heterogeneous populations worldwide. This has implications for public health approaches to T2D prevention.