Improving causal inference in rheumatic diseases research by integrating observational and genetic epidemiology
Approved Research ID: 72723
Approval date: October 18th 2021
Rheumatic and musculoskeletal diseases (such as rheumatoid and osteoarthritis) are the leading causes of disability and loss of work productivity in developed countries. Their impact on individuals and societies will increase with ageing populations. Traditional research methods to identify risk factors for, and health-related consequences of, rheumatic diseases have inherent weaknesses. For example, body fat correlates with arthritis risk, but these research methods cannot tell us whether one causes the other and which way around. Untangling cause and effect is important because, for example, reducing body fat can only prevent arthritis if body fat causes arthritis, not if arthritis changes body fat. Inability to determine cause and effect is the same reason that the majority of molecules linked to arthritis do not translate into effective drug treatments. Newer scientific methods that borrow strength from genetic data can address these weaknesses and make existing research findings more relevant to public health policy and drug-development to prevent or treat rheumatic disease. We propose three objectives to address this overarching aim. We will investigate risk factors that can be modified to prevent rheumatic diseases (Aim 1) and, in reverse, whether rheumatic diseases influence risk factors or other diseases (Aim 2). We will incorporate genetic data to improve accuracy of disease definitions (Aim 3), which will help all future UK biobank studies of these diseases and is required to achieve Aims 1 and 2.
Genetic epidemiology can complement traditional observational study designs to improve causal inference. We propose to integrate genetic data (e.g., using Mendelian randomisation) to:
1) Investigate the causal roles of risk factors on rheumatic diseases. Specifically, we will focus on two groups of risk factors: a) environmental and lifestyle risk factors, such as body fat composition and smoking, that may be modifiable to prevent disease; b) plasma proteins, such as cytokines, that may have potential as therapeutic targets.
2) Investigate the causal roles of rheumatic diseases on health outcomes. These will again form two groups: a) diseases causing morbidity and mortality such as cardiovascular and malignant diseases; b) general health outcomes such as lifestyle factors and healthcare costs.
3) Refine rheumatic disease phenotypes. Disease populations identified using diagnostic codes or self-report can lack accuracy. By contrast, formal genome-wide association studies (GWAS) typically use more robust and homogenous disease definitions. Genetic data from the latter can be used to assess and improve robustness of phenotype definitions; for example, if genetic correlation between self-report and gold-standard disease definitions are poor, then the former may require refining before being used to identify patient populations.
Scope clarification: 1) By rheumatic diseases, we intended to refer to rheumatic, musculoskeletal and related immune mediated inflammatory diseases (IMIDs). For example, studying psoriatic arthritis without studying psoriasis (not a rheumatic disease) is incomplete since both belong to the umbrella of psoriatic disease; another example is spondyloarthritis and extra-skeletal manifestations of uveitis or inflammatory bowel disease. 2) When investigating potential protein drug targets, it is also necessary to examine potential adverse effects which would require extending such analyses to other relevant diseases. For example, studying interleukin-6 inhibition for polymyalgia rheumatica is incomplete without investigating potential harmful effects on cardiovascular or gastrointestinal diseases. Both points of clarification aim to harmonise Aims 1 and 2.