Clinical, Diagnostic Biomarkers, Preclinical

Identification of prodromal, shared and discriminatory features in multiple system atrophy (MSA)

In the early stages, it can be difficult to distinguish Multiple System Atrophy (MSA) with other parkinsonian disorders such as Parkinson’s Disease (PD) with up to a 30% misdiagnosis rate. Furthermore, progression to death for MSA is unpredictable, with some patients progressing rapidly (<5 years) whilst others can live on for >20 years. These factors make running treatment trials challenging in terms of patient recruitment and result interpretation.

Traditional statistical methods have shown that certain clinical features can be used to differentiate between the conditions, as well as predict disease progression, but with limited accuracy. Machine learning has been shown to be able to accurately predict outcomes by determining statistical patterns not immediately obvious to the human eye. This study seeks to see whether accurate machine learning models can differentiate MSA as well as predict rate of disease progression.

In the first instance, we will extract clinical information from patients with post-mortem confirmed MSA and PD to create an initial machine learning training cohort. This will be used to develop a machine learning model to distinguish between the two conditions. We will combine the best current fluid biomarker, neurofilament light (NfL) to investigate if this further enhances diagnosis. This model then undergoes repeated tuning until the best possible model is determined. Subsequent validation will be performed using clinical information from larger cohorts of patients.

This project opens the potential to advance MSA diagnosis and prediction of progression, allowing patients to enter clinical trials earlier (before they lose function) and that clinical trials can be more specific about patient selection, so as not to confuse natural disease progression with treatment impact. If successful, we will be able to develop an online application that everyone will be able to use to help differentiate between MSA and PD and help predict disease progression.