Predicting dementia outcomes using simple, non-invasive assessments: a prospective population-based study

Background:
Around 670,000 people in the UK are currently living with dementia, and this number is expected to double over the next twenty years. Despite many years of research, we still do not have a treatment that prevents or cures this devastating condition. We now understand that the damage causing dementia begins many years before someone develops symptoms, and so it is possible that treatments do not work because the condition is too severe by the time we give them to patients. We therefore need to find a way of identifying people who are currently healthy but are at risk of getting dementia in the future. We ideally need to do this using only the sort of information that is available to GPs, to avoid doing invasive and expensive tests on lots of healthy people.

Project aim:
In this study I will use data from a large study called UK Biobank, to create a model that uses simple information to predict who is most at risk of developing dementia over a 5-10 year period.

Where this research will be performed:
This research represents a collaboration of several groups of researchers across several sites at the University of Edinburgh. I will perform the analyses with support, guidance and training from experts in the field.

How the predictive model will be created:
UK Biobank (UKB) is a very large population-based cohort study of 503,000 middle-aged people. During recruitment participants were extensively evaluated and took brief, electronic thinking tests. In 2014-2015 over 118,000 participants responded to a repeat online memory test, making this the biggest study of repeat cognitive testing ever.

The participants in UKB are followed up using routine NHS datasets. When patients are diagnosed with conditions such as dementia either by their GP, in hospital or after they have died, this is recorded in these datasets. The participants in UKB have consented to let UKB access these records so they can learn about their health. Conservative predictions have shown there is likely to be around 4000 dementia cases in the cohort by mid-2017, which would make this by far the largest ever study to create a dementia prediction model. It has also given me relevant research experience and an appreciation of the issues involved when working with data from cohort studies and with healthcare datasets from England, Scotland and Wales.

I will apply to access UKB data that includes the information obtained at recruitment, during the repeat online tests and in the routine NHS datasets. I will then investigate which simple characteristics can best predict who is likely to get dementia. These are likely to be things such as age, smoking status, educational level and family history. I will also look at how physical health problems (such as diabetes, heart disease and stroke) might impact on a person's mental health, by seeing whether having one of these conditions increases the risk of getting dementia. I will also use the brief thinking tests that participants took at recruitment and during follow up to see if changes in these can predict who will get dementia before they have obvious symptoms. I will then combine the most predictive characteristics into one model.

After creating the model, the next, important stage will be to test it. To do this I will use data from a Scottish study called Generation Scotland (GS). GS has many similarities to UKB in the way participants were recruited and tested. I will also test the model using real-life data from two very large sources of GP data from England and Wales.

Why this research matters:
We need to change the way we test new dementia treatments to increase the likelihood we find one that works. My goal is to build a prediction tool that can be used to identify people at risk of developing dementia, so they can be invited to participate in trials testing new treatments. If an effective treatment becomes available, doctors could also use this tool to identify who would benefit.

Project aim:
To develop a model that predicts the risk of developing dementia over a 5-10 year period in an asymptomatic population using simple, readily available, non-invasive measures and brief, touchscreen cognitive assessments

Objectives:
1. Obtain and thoroughly explore the UK Biobank (UKB) baseline and follow up data
2. Identify the variables that most strongly predict incident dementia diagnosis (all-cause dementia, Alzheimer's disease and vascular dementia)
3. Create multivariate risk prediction models by combining the independently predictive variables
4. Validate the model using surplus UKB outcomes and data from external sources

Methodology:
I will obtain the UKB baseline data for all 503,000 participants as well as coded data for the entire cohort from linked health datasets (hospital admissions, death registrations and primary care). I will perform univariate logistic regression to identify the variables that most strongly predict dementia over a 5-10 year period and then use multivariate logistic regression to combine independently predictive variables to create the prediction models. I will investigate the incremental value of the addition of the cognitive testing data in "high risk" participants.

Power calculation:
It is generally accepted that 10-20 outcomes are required per variable in a multivariate logistic regression analysis, meaning that 100-200 cases would be sufficient for a 10-variable model. Our scoping work has identified a predicted 4000 all-cause dementia cases by 2017.

Model validation:
I will comprehensively validate the models using surplus UKB outcomes (internal) and using data from Generation Scotland, CPRD and SAIL databases (external).

Outputs:
My research will fill a key gap in the pathway to development of better interventions for dementia prevention and treatment, and will equip me with research skills that will provide the foundations of an ongoing clinical academic career in dementia research.

Whereas the primary aim of this project is to advance scientific knowledge, from its inception I have considered ways in which I can maximise the impact of this research on our society.

In summary I anticipate this project will have a positive impact in the following areas:

1. Promoting healthy behaviours by highlighting the role of modifiable lifestyle factors that contribute to dementia risk

There is currently a great deal of public and media interest in dementia and dementia research, and I believe that in general public understanding of dementia has been improved as a result. More needs to be done however to explain the differences between normal cognitive ageing and dementia, and the role of lifestyle factors in significantly contributing to dementia risk. This is still an evolving scientific area, and by engaging with the public I anticipate that people would seek healthier lifestyle choices if they understood the role of these factors in increasing the risk of developing dementia. By developing a simple risk calculator that incorporates these factors, it will highlight the role they play in contributing to the development of dementia. Furthermore, this project will raise awareness about the relatively recent discovery that the processes underlying dementias begin decades before symptom onset, encouraging people in middle age to minimise their risk of developing dementia in older age by making lifestyle changes.

2. Highlighting the benefits of using health data in research

The use of health data has been a controversial issue over recent years. I anticipate that this research will highlight some of the many benefits of the secure collection and analysis of routinely coded health data. In particular I intend to promote the "Data Saves Lives" campaign throughout the project by using this research as an example of how these data can be of benefit to our society.

3. Promoting the UK as a world class place to do epidemiological research

The nature of NHS as a cohesive healthcare system and the long history of investment in cohort studies and recording of routine health data, means that the UK is world-leading in epidemiological research. I hope that this research, when presented to epidemiologists, clinicians and data scientists around the world, will promote the role of the UK in this regard, and encourage international researchers to invest in and apply for UK health data for their own purposes. In particular, as this project combines the use of data from two large cohort studies in UK Biobank and Generation Scotland as well as two very large primary care data resources (CPRD and SAIL), this will show the UK as a unique place to perform collaborative, "big data" epidemiological research into chronic diseases.