Genetic Evaluation of Multimorbidity towards INdividualisation of Interventions (GEMINI)

More than 50% of people over the age of 65 are living with more than one long term condition (multimorbidity). Despite this, people with multimorbidity are often excluded from clinical trials and there has been limited research into identifying the causes of multimorbidity. For example, we often do not know if two common long-term conditions occur together by chance as we get older, whether one leads to the other, or if they share a risk factor. This problem is partly because health care professionals and researchers tend of focus on one condition at a time. For example, there has been a lot of research into the causes and consequences of osteoarthritis but not why people with osteoarthritis have a higher frequency of asthma, even when accounting for sex, age and obesity.

The aim of our research is to uncover new links between long term conditions that could lead to improved interventions including drug treatments or other more focused treatments. These new links could include a better understanding of which cells in the body are most critical to the presence of two conditions in the same patient.

To achieve our aims, we have formed a partnership called the GEMINI (Genetic Evaluation of Multimorbidity towards INdividualisation of Interventions) collaborative. This team includes two people with multimorbidity, health care professionals including those in primary care and experts in statistics and genetics. In GEMINI we will study the causes of multimorbidity with a new approach. We will use existing databases of DNA sequence information linked to diseases from 10,000s of people. Using this genetic approach our initial research has identified many new and interesting links between conditions that were not previously well known. For example, between Rheumatoid arthritis and stroke (but not Rheumatoid arthritis and heart disease), gastro-reflux disease and depression, and between asthma and osteoarthritis. We will complement the genetic approach with data from millions of patients in primary care. These patients are representative of the UK as a whole and will allow us to study large numbers of people with combinations of conditions even if these combinations are quite rare.

Our research plans are divided into three parts. We will involve patients and carers in all stages to ensure we are using their data appropriately and to help us remain focused on the important conditions and outcomes of multimorbidity. First, we will use three sources of data from patients in primary care (GPs) to define the conditions we will study. We will start from all conditions that are long term and present in more than 1% of the people over 65 years. We will then use millions of DNA sequence changes - the genetic information we inherit from our parents - to identify which conditions share broad biological mechanisms. Second, we will use a similar number of genetic variants to identify the specific mechanisms involved. These techniques are based on the principle that inherited DNA sequence changes are fixed for life and so provide us with a way of assessing the causal direction of associated risk factors and diseases. For example, we will use genetics to test whether one disease leads to a second disease, or whether a shared risk factor leads to both. These risk factors will include well known risks such as obesity and more detailed measures of biology, such as how genes are switched on and off in different cells and tissues. Third, we will study in more depth patients with the conditions highlighted in the first two steps using primary care databases. We will hold workshops with patients and carers to understand in depth the most important outcomes of these conditions, for example is reduced lifespan more or less important than risk of frequent hospitalisation? We will then study patients with new combinations of conditions to see if they suffer from worse outcomes.

We will test the hypothesis that there are unrecognised combinations of long term conditions (LTCs) which arise due to shared biological pathways, and that these can be discovered using human genetic methods. Recent studies indicate that this is an exciting and feasible approach to advance the understanding of multimorbidity, with human genetic and genomic data leading to new discoveries about shared mechanisms between LTCs.

Our research is divided into three complementary work packages (WPs). In WP1, we will select LTCs that are common and use genetics to identify those that are likely to share biological pathways. We will use data from three primary care databases consisting of millions of patients to identify LTCs present in >1% of people aged >65 yrs. We will then use data from large (10,000s of cases) genetic studies to identify LTCs that are genetically correlated with each other.

In WP2, we will identify some of the specific mechanisms underlying the LTC combinations identified in WP1. We will test if known modifiable risk factors, such as BMI, account for some of the genetic correlations. We will then use more specific sets of genetic variants to identify potential causes. These genetic variants will include those representing genes expressed in specific cell types, those representing potential drug targets, and those that can be used in Mendelian randomization tests of specific risk factors. We will develop new methods to ensure our causal inferences are robust, including the use of genetic variants to test whether one condition causes a second condition.

In WP3, we will obtain patients' perspectives on the new LTC combinations. We will then use the primary care datasets to test the hypothesis that the new LTC combinations represent distinct clinical entities, defined as patients aged >40 yrs with >1 condition having worse outcomes than expected. Our results will inform future translational studies aiming to reduce the burden of multimorbidity