Dementia Risk Prediction Model: Development and Validation

NCT ID: NCT03943641

Last Updated: 2024-12-10

Basic Information

• Recruitment Status: COMPLETED
• Total Enrollment: 400000 participants
• Study Classification: OBSERVATIONAL
• Study Start Date: 2019-05-08
• Study Completion Date: 2024-12-05

Brief Summary

At present, there is no treatment for dementia that changes the course of the disease. However, it is now understood that the proteins in dementias such as Alzheimer's disease are present years before someone develops symptoms of dementia. Studies may therefore need to give potential treatments to patients before they develop symptoms of dementia. To do this, researchers need a way of predicting who will go on to develop dementia in the future.

There are several ways of doing this, however, many of these methods are costly and difficult to implement at a population level - such as brain imaging, lumbar punctures or psychological tests. In this study, the investigators aim to develop a method of predicting who will go on to develop dementia (and dementia due to Alzheimer's disease) using only the sort of information that a general practitioner would have available to them.

To do this, the investigators will develop a dementia prediction model using data from the Secure Anonymised Information Linkage (SAIL) Databank, which contains anonymised primary care, hospital admissions and mortality data for the population of Wales, United Kingdom (UK). They will then go on to test how well it performs in an external dataset, such as the UK's Clinical Practice Research Datalink (CPRD).

Detailed Description

To date, no dementia drugs have shown a disease-modifying effect in clinical trials. It is now understood that the pathology underlying Alzheimer's disease is present decades before symptoms become apparent. Starting an intervention only when a patient develops cognitive symptoms, and therefore when there is substantial disease burden, may reduce the chance of any disease-modifying effect. Instead, targeting interventions earlier, when the pathological burden is lower, may increase the likelihood of preventing or delaying dementia onset.

Consequently, there is a need for a method that identifies patients who are at an increased risk of developing dementia. This requires the development of a risk prediction model, which utilises multiple predictors in combination to produce individualised estimates of the risk of developing dementia risk over time.

An ideal risk prediction model for a population-based application would need to use predictors that are already available to, or readily obtainable by, general practitioners (GPs). Such a predictive tool could be used as a low cost, scalable method of recruiting an 'at risk' group of participants to future trials of risk modification strategies or preventative therapies. Once an effective disease-modifying intervention is identified, clinicians could use the same model to identify at-risk patients who may benefit most from undergoing the intervention.

An ideal dementia risk prediction tool would contain only information that is readily available to, or easily obtainable by, clinicians such as General Practitioners (GPs).

The investigators aim to develop two 10-year risk prediction models: one to predict all-cause dementia and one to predict Alzheimer's disease dementia, in UK adults aged 60-79 years, using only predictors that are routinely available to GPs. They will develop the model using data from the Secure Anonymised Information Linkage (SAIL) Databank, which is composed of anonymised, linked primary care, hospital admissions and mortality data for the population of Wales, UK.

The investigators will then go on to externally validate their dementia risk prediction models in an external dataset, such as the UK's Clinical Practice Research Datalink (CPRD). They will also validate an existing, published study using data from the The Health Improvement Network (THIN) (Walters et al. 2016) using this external dataset, allowing us to compare the performance of the models.

Conditions

Dementia Alzheimers

Study Design

• Observational Model Type: COHORT
• Study Time Perspective: RETROSPECTIVE

Study Groups

Population-based

Population-based cohort of participants registered with a SAIL-contributing practice.

This is not an intervention study

Intervention Type: OTHER

This study is based on retrospective analysis of linked routinely-collected healthcare data

Interventions

This is not an intervention study

This study is based on retrospective analysis of linked routinely-collected healthcare data

Intervention Type: OTHER

Eligibility Criteria

Inclusion Criteria

• Registered with a SAIL practice before 2008
• Aged between 60-79 during the study window (1st January 2008 to 31st December 2017)
• Aged 60-79 years by January 2008

Exclusion Criteria

• Deprivation quintile missing for the start of follow-up (deprivation scores will probably not be missing at random)
• All-cause dementia code in any dataset prior to 1st January 2008 (i.e. dementia diagnosis at baseline)

• Minimum Eligible Age: 60 Years
• Maximum Eligible Age: 79 Years
• Eligible Sex: ALL
• Accepts Healthy Volunteers: Yes

Sponsors

Keele University

OTHER

Sponsor Role: collaborator

University of Edinburgh

OTHER

Sponsor Role: lead

Responsible Party

Responsibility Role: SPONSOR

Locations

Usher Institute, University of Edinburgh

Edinburgh, Midlothian, United Kingdom

Countries

United Kingdom

Other Identifiers

AC19049

Identifier Type: -

Identifier Source: org_study_id