Emulating randomized trials in observational data to determine ethnic differences in the effects of type 2 diabetes treatments

Applications are invited from graduates with an MSc (Distinction) or equivalent to work within the Wolfson Institute of Population Health. This 4-year studentship will commence in Autumn 2023 and will be based at the Whitechapel Campus of QMUL. This is an exciting opportunity for a graduate from disciplines related to epidemiology, medical statistics, or health data science.

Project description

Background: Type 2 diabetes is a heterogenous condition comprising a broad range of subtypes (termed phenotypes) with differing aetiologies and clinical characteristics. Previous studies have identified south Asian phenotypes, characterized by high intra-abdominal fat and high insulin resistance, and African phenotypes, characterized by lean body size and reduced pancreatic insulin secretion. In high income countries, people of Asian and African ethnicity have over twice the prevalence of type 2 diabetes and 2-3 times the risk of vascular complications and premature death than people of white ethnicity. Current type 2 diabetes guidelines are informed by trial evidence from studies of predominantly white European populations. There is widespread concern that existing guidelines are not tailored to the management of diabetes in different ethnic populations. While randomized interventional studies are considered the gold standard for estimating causal effects, non-interventional methods fill an important evidence gap as they can generate valid estimates of treatment effectiveness in populations excluded from or under-represented in clinical trials. This is achieved via trial emulation which creates trial-analogous cohorts in observational data sources.

Research aims:

The PhD project will address 3 main aims:

1)    To validate non-interventional methods against trial results by emulating a recent diabetes trial in UK based observational data

2)    To estimate treatment effects in ethnic groups underrepresented in the original trial

3)    To estimate the relative influences of phenotype, environment, lifestyle, and socio-cultural factors on ethnic differences in treatment response

Methods: The DURATION-2 trial compared the glycaemic effects of Exenatide (a GLP-1 agonist), Sitagliptin (a DPP4i) and Pioglitazone (a TZD) in individuals with type 2 diabetes treated with metformin. In the CPRD, adults with type 2 diabetes treated with metformin and initiating GLP-1 agonists, DPP4i, or TZD treatment will be identified, resulting in a pool of CPRD participants comparable to DURATION-2 participants. Cox proportional-hazards regression will be used to estimate differences in time to primary treatment failure (HbA1c>=7.5%/58 mmol/mol) and onset of vascular complications by treatment class. The results obtained will then be validated against the DURATION-2 trial results (aim 1). If the findings from the validation study in CPRD match those of the DURATION-2 trial, the CPRD study will be extended to examine treatment response in people of south Asian, east Asian, and African ethnicity (aim 2). Causal mediation analysis will be used to estimate the relative influences of phenotype, lifestyle and socio-cultural factors such as the healthcare setting, language, and socio-economic status on ethnic differences in antidiabetic treatment response (aim 3).

Timetable:

Year 1: Ethical approval for CPRD data access and request individual participant data from DURATION-2 trial. Complete systematic review and begin data management for aim 1

Year 2: Complete analyses for aim 1 and write up for aim 1 and draft paper for publication. Attend an international conference to present findings from aim 1.

Year 3: Complete analyses and write up for aim 2 and draft paper for publication. Attend an international conference and present findings from aim 1.

Year 4: complete aim 3 and draft paper for publication.  Write-up thesis and submit.

About the candidate: This PhD would be suitable for a candidate with a background in epidemiology, medical statistics, computer science, or related areas with an interest in big data, health analytics, electronic health records and diabetes. The successful candidate will be keen to undertake interdisciplinary work using pharmacoepidemiology and causal inference methods.

Informal enquiries can be made via email to: Professor Rohini Mathur, r.mathur@qmul.ac.uk

How to apply

Your application should consist of a CV and contact details of two academic referees. You must also include a personal statement (1,000 words maximum) describing your suitability for the selected project including how your research experience and interests relate to the project.

Please submit your application to: Patrick Mullan (p.mullan@qmul.ac.uk).