Precision medicine in diabetes: Pharmacogenetic studies of large randomised controlled trials of diabetes therapies

In the treatment of type 2 diabetes, there are five non-insulin therapies available for use after metformin. These are called sulphonylureas (SU), thiazolidinediones, dipeptidyl-peptidase 4 inhibitors (DPP4i), Sodium Glucose Transporter 2 inhibitors (SGLT2i) and GLP-1 Receptor Agonists (GLP-1RA). The latest guidelines aim to guide treatment choice based upon risks (e.g. of low blood sugar or weight gain), cost (e.g. SU are cheap and could be used in low and middle income countries), and benefit (e.g. reduction in risk of heart failure or heart attack with GLP-1RA and SGLT2i) however there is considerable variation in who benefits, and who is harmed, from any of these treatments. We have established that blood sugar response to diabetes treatments is highly "genetic" and have identified examples where genetic variants alter how well the diabetes drugs work. As we move to a time in the near future when genetic information will be available for a patient when the doctor prescribes a drug, it will be possible to take into account this genetic information when choosing the best medication for a patient with diabetes.

To date, genetic studies of drug response in diabetes have been limited to observational studies. These studies are limited in their ability to study newer drugs, and despite including genetics, are still prone to bias and noise seen in real-world studies. The present UK-Canada collaboration brings together, for the first time, genetic data on randomised controlled trials (RCT) of the newer diabetes agents, including trials that established the cardiovascular benefit of SGLT2i and GLP-1RA. Currently we have access to data from participants to SGLT2i trials (Dapagliflozin and Empagliflozin, n=10,943), GLP-1RA trials (Albiglutide, Lixisenatide, Dulaglutide n=16,596), DPP-4 inhibitor trials (Saxagliptin, n=3048). These unique resources are only available to the principal applicants and provide considerable power to identify genetic variants that alter blood sugar response, side effects and cardiovascular outcome of these drugs that are expensive and increasingly used in the UK and Canada.

We will undertake studies of multiple drug related responses including glycaemic response, weight reduction, blood pressure reduction, nausea/vomiting, thrush, renal outcomes and cardiovascular outcomes. First, we will undertake genetic studies looking at nearly 5 million genetic variants for each trial participant to identify associations that will provide mechanistic insight into drug action as well as variants that may be used in the clinic to predict who will respond well or poorly to medication. Second, we will then look at combinations of genetic variants that increase risk of diabetes due to a common underlying cause - called partitioned polygenic scores (pPS) - and investigate how different pPS alter response and outcome to the diabetes drugs. Finally, we will use the genetic information to ask questions in relation to how the newer diabetes drugs work (via what mechanism) and in particular how they improve cardiovascular outcomes.

By identifying genetic variants or pPS that alter response and outcomes of SGLT2i, DPP-4i and GLP-1RA, and undertaking more complex mechanistic studies in large highly powered clinical trials, we will provide a major advance in precision medicine in diabetes. This will enable better targeting of treatment for better patient outcomes at reduced cost to the health care system.

There are now many treatment options in type 2 diabetes, with an increasing trend to using newer agents earlier due to their added cardiovascular benefit. This will greatly increase treatment costs, yet we know that some drugs are ineffective or cause adverse effects in some patients. We have established that glycaemic response to metformin is moderately heritable and it is likely that other response traits to other diabetes drugs are also heritable. We now propose that understanding this genetic variation will provide insights into drug mechanisms and the potential to select patient groups who will benefit most to specific treatment options.

To date, pharmacogenomics in diabetes has been limited to older drugs and to observational data. We have published the first metformin GWAS on just 1000 patients. In the present proposal we have assembled over 30,000 patients from randomised trials of the newer diabetes drugs (SGLT2i, GLP1RA and DPP4i) with completed genome-wide data. Both SGLT2i and GLP1RA have shown reduction in cardiovascular (CV) risk, yet the mechanism for this remains uncertain.

We propose to use genomic data for SGLT2i, GLP-1RA and DPP4i to: 1) run GWAS of 9 response traits (endophenotypes), 5 safety outcomes and 5 CV outcomes. 2) To use partitioned polygenic scores of diabetes risk to investigate what genetically defined aetiological processes alter response to these drugs. 3) To investigate tissue site of action using PrediXcan and pathways associated with response traits from Aim 1. 4) To investigate shared genetic aetiology between response endophenotypes and CV outcomes within and between different drugs using bivariate GREML, and to investigate causal associations between endophenotypes and CV and renal outcome.

The identification of genetic variants that predict response and outcomes with SGLT2i, DPP-4i and GLP-1RA drugs will provide a major advance in personalized treatments for diabetes.

Diabetes is associated with considerable mortality and morbidity, consuming over 11% of the global health budget (up to 60% in some developed countries), estimated at $673 bn worldwide. There is therefore a clear need to reduce complications of diabetes. Current well established approaches are to lower blood sugar as measured by HbA1c and target cardiovascular risk factors by treating aggressively with statins and anti-hypertensives. Yet despite this the rates of complications remain high and the standardised mortality rate is increased in both Type 1 and Type 2 diabetes.

The newer diabetes therapies such as GLP-1RA, SGLT2i and DPP4i are increasingly used early in treatment protocols. The joint EASD/ADA guidelines now recommend the use of GLP-1RA or SGLT2i after metformin in patients with previous CV disease. Whilst these new drugs have been established to reduce CV risk and should result reduced mortality, morbidity and improved patient outcomes. Yet these treatments are expensive and increasing use ahead of the very cheap drugs such as metformin, sulphonylureas and pioglitazone will result in increasing costs to the health care systems. The outputs of this proposal will identify who are the patients who most and least benefit from these expensive drugs, and will provide evidence to help guide drug choice at the point of prescribing. This should further optimize patient outcome and reduce treatment costs.

Impact will be seen by many stakeholders. The main intent is that patients with diabetes will be the main beneficiaries, due to improved patient care resulting from increased understanding of who responds well or poorly to a diabetes treatment, enabling better targeting of treatment. Insights into mechanism may also help improve targeting of treatment or development of novel therapies or repurposing of existing therapies for patient benefit.

Other beneficiaries will include health care providers such as the NHS or payers in Canada - better targeting of treatment will result in substantial cost savings and reduced waiting times; the pharmaceutical industry - insight into the biological processes underlying benefit and harm from their drugs will enable better targeting of therapy, potential for repurposing and potentially novel therapies; diabetes and cardiovascular researchers - our research will lead to novel biological insights at the cellular, animal model and human physiology level.