This project aims to help people with diabetes get optimal drug treatment and is an extension of a large MRC-funded project for which we have recently received 3-years follow on funding. Diabetes is a condition where the body’s blood sugars are too high. This can lead to serious problems affecting the eyes, kidney, and nerves and a higher risk of heart attacks and strokes. Diabetes can be treated with tablets and/or insulin injections, but determining which treatment is best for individual people is challenging. We have been seeking the best way to determine which treatment gives a patient the best response. Our research to date, largely carried out in CPRD (previous protocol references 13_177, 18_062, 22_002000; >12 publications), has shown that simple clinical features, available in a patient’s GP record (e.g. age, sex, weight, blood test results) can be combined by a computerised calculator to help inform treatment decisions for 4 of the main diabetes drugs. We are now ready to expand this work to help develop a clinical calculator that will help with treatment decisions for all diabetes drugs, based on a range of outcomes (e.g., risk of side effects/cardiovascular outcomes, as well as effect on blood sugars) that is valid across different time points (e.g., at time of diagnosis and later on in the disease), and different patient groups (e.g., across different ethnicities). By the end of our project, we aim to have developed a calculator ready to start testing in real-life GP practice computer systems.
Background: In Exeter, through our MRC-funded MASTERMIND precision medicine programme, we have been at the forefront of research showing clear, robust, clinical criteria that can help better inform treatment decisions in type 2 diabetes, with much work already carried out in CPRD (ISAC protocols 13_177, 18_062, 22_002000; >12 publications). Following the success of this original project, we have been awarded further MRC funding to extend our work to develop full treatment selection models for diabetes (MR/W003988/1).
Aim: To develop treatment selection algorithms that provide detailed personalised information on the potential risks and benefits of all glucose-lowering therapies for individuals with diabetes.
Study population: All patients with diabetes from 2004 to date.
Primary outcomes/exposures: Outcomes: Glycaemic response, drug tolerability, side-effects, cardiorenal outcomes. Exposures: Clinical features routinely recorded in primary care e.g. demographics, blood test results, comorbidities
Data sources: CPRD GOLD, Aurum, Hospital-episode statistics
Methods: We will develop statistical models focussed on “heterogeneity of treatment effect” (HTE) i.e. for treatment choice, it is more important to quantify the difference in outcomes between drugs, dependent on a person’s characteristics, rather than directly predicting individual drug outcomes. Multiple statistical approaches will be applied with the aim of optimising performance of the treatment selection model. Additional methodological approaches will be explored to account for potential confounding and missing data. Final models will be validated in out-of-sample datasets, using both CPRD and external data sources.
Intended public health benefit: There is currently huge variation in type 2 diabetes prescribing patterns across the UK and current approaches for determining the best treatment for a patient are largely “trial and error”. The treatment selection model we propose developing will enable a more personalised discussion between the clinician and patient of the potential benefits and risks of the different glucose-lowering drugs for them based on their clinical features.
1) Short-term outcomes
• Glycaemic response – change in HbA1c after initiating glucose-lowering treatment (primary care)
• Treatment tolerability – discontinuation of glucose-lowering treatment (primary care)
• Side-effects – glucose-lowering treatment side-effects (primary or secondary care)
2) Long-term complications
• Microvascular complications of diabetes (primary or secondary care)
• Cardiovascular complications of diabetes (primary or secondary care)
• Renal complications of diabetes (primary or secondary care)
• Cardiovascular and all-cause mortality (ONS death registration data)
Beverley Shields - Chief Investigator - University of Exeter
John Dennis - Corresponding Applicant - University of Exeter
Andrew Farmer - Collaborator - University of Oxford
Andrew Hattersley - Collaborator - University of Exeter
Andrew McGovern - Collaborator - University of Exeter
Angus Jones - Collaborator - University of Exeter
Ethan de Villiers - Collaborator - University of Exeter
Ewan Pearson - Collaborator - University of Dundee
Jack Bowden - Collaborator - University of Exeter
John Bassom - Collaborator - University of Exeter
Joseph Ramage - Collaborator - University of Exeter
Katherine Young - Collaborator - University of Exeter
Laura Guedemann - Collaborator - University of Exeter
Long Him Li - Collaborator - University of Exeter
Naveed Sattar - Collaborator - University of Glasgow
Pedro Cardoso - Collaborator - University of Exeter
Rhian Hopkins - Collaborator - University of Exeter
Rury Holman - Collaborator - University of Oxford
Thijs Jansz - Collaborator - University of Exeter
Trevelyan McKinley - Collaborator - University of Exeter
William Henley - Collaborator - University of Exeter
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Townsend Index