Dementia is a disease characterised by progressively worsening memory and cognitive decline, with loss of functional independence, with the patients being ultimately unable to perform everyday activities. Alzheimers disease (AD) accounts for the vast majority of dementia cases. Currently there is no treatment that stops or slows dementia. Many risk factors and comorbidities are known to be associated with dementia including hypertension, diabetes, cerebrovascular disease, and systemic inflammatory disorders. As treatments for many of these biological mechanisms and conditions already exist, we hypothesise that existing drugs, currently used to treat other conditions, may be also used to treat dementia. This is called drug repurposing. In order to test this, we will use data from the Clinical Practice Research Datalink and aim to analyse this dataset as close as possible to a clinical trial design with two highly matched groups exposed to the drugs of interest and to a control drug. The list of drugs that will be examined will include drugs chosen on the basis of robust scientific evidence from previous studies, either related to evidence from existing literature on mechanisms of action that are relevant to dementia or their primary indications that are linked to dementia.
This study aims to examine whether already approved drugs for other diseases can be used to lower the risk of dementia and AD and/or slow disease progression. We will conduct a retrospective longitudinal cohort analysis with data on drugs (e.g., metformin) and the diagnosis/progress of dementia. Our outcome of interest will include participants with a diagnosis of dementia or prescriptions to dementia specific medications. We will examine associations between drugs of interest and risk of dementia using several statistical approaches such as Cox regression. For each candidate drug, we will implement a robust synthetic matching with distance measures based on propensity scoring estimated using relevant confounders including: age at initial prescription, gender, deprivation index, smoking status, body mass index, and comorbidities (e.g., cardiovascular diseases, depression, cancer, COPD, chronic kidney disease). We plan to refine matching by using modern methods of dealing with high-dimensional data such as Lasso and other penalized methods. Our goal is to balance the need to obtain observed treatment and control groups that have similar covariate distributions with the fact that, in some situations, one or both of the observed treatment and control groups may not allow matching using all available covariates. We will investigate the impact of missing values and different approaches to dealing with missing values including imputation methods and deep learning approaches.
The outcome of interest is the incidence and progression of any dementia or dementia attributable to Alzheimers disease. We will sequentially add definite or probable dementia cases using the list of diagnoses from the CPRD dataset, then complementing it with data from the Hospital Episode Statistics (HES), and potentially using additional algorithm-derived cases.
For instance, using CPRD data, patients will be considered to have dementia if they had a dementia diagnosis based on Read codes or if they were prescribed with anti-dementia drugs (donepezil, galantamine, rivastigmine, or memantine). Additional information from HES and other linked datasets on mental tests, neuroimaging assessments, and referrals to neurological specialists will also be used to ascertain the suspected cases.
The event date of dementia onset will be defined as the first-time diagnosis date or the first-time prescription date, whichever came first. The time period of dementia progression will be defined by the time elapsed between the timing of AD onset and the timing of the first prescription in the anti-psychotic class.
Ioanna Tzoulaki - Chief Investigator - Imperial College London
Ioanna Tzoulaki - Corresponding Applicant - Imperial College London
Bang Zheng - Collaborator - Imperial College London
Bowen Su - Collaborator - Imperial College London
Lefkos Middleton - Collaborator - Imperial College London
Mahsa Mazidi - Collaborator - Imperial College London
Roger Newson - Collaborator - Imperial College London
Roy Welsch - Collaborator - Massachusetts Institute of Technology
Stan Neil Finkelstein - Collaborator - Massachusetts Institute of Technology
HES Accident and Emergency;HES Admitted Patient Care;HES Outpatient;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation