Approaches, challenges, and opportunities in the application of genetic epidemiological techniques to lifecourse epidemiology

Abstract

Adverse health conditions that emerge in later life are often shaped by processes experienced throughout life, including in utero. Better understanding how exposures at different life stages influence health outcomes is key for developing effective disease prevention and treatment strategies. Despite advancements in lifecourse epidemiology, the adoption of robust causal inference methods has been slow, particularly in addressing unmeasured confounding. In response to this, I identify and utilise existing approaches as well as develop novel advances in the application of genetic techniques to lifecourse epidemiology.

Mendelian randomisation (MR) is a genetic epidemiological method, that, under certain assumptions, estimates causal effects of modifiable exposures on outcomes. In this thesis, I systematically identify MR methods that have previously been developed and apply selected methods to interrogate and explain techniques that may be used successfully in a lifecourse research context. I start by using inverse variance weighted (IVW) univariable MR and multivariable MR (MVMR) to estimate effects of a phenotype (body size) at specified periods across the lifecourse on several later life outcomes (including cardiovascular disease and fracture risk). Next, I evaluate IVW-MVMR alongside a structural mean modelling MR approach. Shifting to a multigenerational setting, I estimate effects of a parental phenotype (smoking) on adult offspring cardiometabolic outcomes using intergenerational MR. I then apply a novel combination of epidemiological methods to explore the effect of early-life exposures and mediating mechanisms (body size and body image dissatisfaction) on self-harm and disordered eating in adolescence.

In this thesis, I highlight the importance of underlying methodological assumptions in the application of MR to lifecourse research questions and show how output from analytical frameworks that rely on different assumptions should be interpreted. Finally, I consolidate the key insights from the thesis. I discuss the challenges and limitations of this work, the wider implications of the findings and potential avenues for future research.

Date of Award	1 Oct 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	George Davey Smith (Supervisor), Jon E Heron (Supervisor), Tom G Richardson (Supervisor), Jessica Tyrrell (Supervisor) & Timothy M Frayling (Supervisor)