Genetic Evaluation of Multimorbidity towards INdividualisation of Interventions - GEMINI

1. To identify disease clusters using a combined genetic and observational approach; and to test the hypotheses that: a) using multiple data sources and genetic correlations will provide more robust estimates of clustering than has been so far possible, and that b) patterns of clustering differ between the sexes and ethnic minorities.
2. To improve reproducibility of multimorbidity research, and working with other collaborative bids where relevant, to develop STATA and R packages to standardize the MRC Reference: MR/V005359/1 Page 2 of 16 SPF Tackling Multimorbidity at Scale: Case for Support process of disease coding across GP databases, and to make these tools publicly accessible.
3. To develop and apply genetic and non-genetic "instruments" and causal inference methods to identify shared biological and non-biological determinants of disease clustering. These instruments will include clusters of genetic variants specific to potential shared disease processes and be randomized to disease incidence and adverse outcomes.
4. To identify a subset of emerging disease clusters for further, in depth study; to validate these clusters in additional datasets with different ascertainment criteria; and to test the hypothesis, using longitudinal data, that the timing sequences of the different disease components to the identified clusters differ from chance (life course).
5. To identify the cluster-associated outcomes of most impact, as informed by frequency, severity and our PPI advisors; to estimate the proportion of patients in each cluster who develop the most impactful outcomes and identify potential modifiable risk factors.
6. To identify the disease clusters that result in clinically meaningful differences in outcomes compared to patients with only one of the constituent diseases.

Researching the co-existence of multiple chronic conditions in a single individual (multimorbidity) is challenging using conventional study designs. Confounding, bias and reverse causality are often complex and severe and may partly explain apparently paradoxical associations. People with type 2 diabetes and additional conditions, for example, tend to have lower HbA1c than those with diabetes alone, and we and others have shown that there is marked weight loss and declining blood pressure for a decade before diagnosis of dementia. Our vision is to address these challenges by combining genetic and conventional approaches and using large-scale data resources from the UK, Spain, US and Canada, including 3 multi-million patient GP data sources. We will identify clusters of disease, use novel causal inference methodology to identify shared biological determinants, and study in-depth a set of disease clusters. By understanding biological determinants of multimorbidity clustering and identifying which are associated with markedly altered clinical outcomes, we will help clarify which multi-morbidity combinations are of most clinical importance to understand. We will define multimorbidity as the presence of 2 or more chronic conditions(1) but focus on those each occurring in >1% of men or women aged 40 plus and that are genetically correlated with other conditions. To address inequalities of multimorbidity we will study the excess burden in women and ethnic minorities. A multi-modal, data driven approach will be critical. Clustering that is consistent across genetic and observational data will be more reflective of shared determinants. Genetic approaches provide a test of lifelong exposure to risk factors and provide strong causal inferences. The widespread availability of genome wide information also means that we can study shared risk factors that are not measured in many studies (e.g. insulin resistance) and calculate disease clustering between, as well as within, databases(2). To achieve our vision we have formed a new multi-disciplinary research team (supported by outstanding external advisors), including researchers with extensive experience in multimorbidity in three GP databases; physical and mental decline in the elderly; specialists in key diseases (diabetes, vascular, dementia, musculo-skeletal) and with expertise in genetics and causal inference.