Surrogate Outcomes: Modelling Uptake in Clinical Trials and Deriving From Multiple Biomarkers

Alzheimer's disease (AD) is a chronic, progressive neurodegenerative disorder and the number one cause of dementia. As the global age of the population continues to increase, a greater number of people will be impacted by the disease. The threat of AD is a global crisis and research needs to address treatment and ultimately, prevention.

At present, there are no disease-modifying therapies (DMT) for Alzheimer's disease but characterisation of the pathophysiology underpinning AD suggests possible targets for potential DMTs. However, trials have mainly focused on targeting the amyloid-B (AB) pathway via anti-AB monoclonal antibodies. Yet, present research depicts the increasing need for treatments targeting non-amyloid AD pathologies in combination with anti-amyloid therapies.

The use of surrogate outcomes allows for a substitution for true clinical outcomes such as how a patient feels, functions or mortality. A surrogate outcome does not directly measure the clinical benefit of the therapy in and of itself, but is used to predict that clinical benefit.

Clinically, biomarkers have been used as surrogate outcomes for the efficacy of drug treatments in many other diseases, yet, there are currently no biomarkers that are definite surrogates of drug treatment efficacy in Alzheimer's disease.

Therefore, this project aims to:

1. Model trends in use of surrogate outcomes using data from the UKHRA and extracting meta-data automatically using natural language processing/machine learning techniques. Advanced statistics and analytical methods to improve and learn about the process of surrogate outcome evaluations. As well as the uptake of surrogate outcome methods.
2. Use data from publicly accessible data repositories as a resource to evaluate potential surrogate outcomes for a more feasible outcome using established statistical methods such as those based on information theory.
3. Combine data from wearables and clinical outcomes for at risk individuals and use dimension reduction via machine learning techniques, to enable creation of multivariable biomarkers that are highly predictive of the outcome of individuals' quality of life.
4. Looking forward to new ways of gathering data and in multiple dimensions.