Drawing on available data on people with mild cognitive impairment (MCI) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), researchers at McGill University in Canada had trained a machine-learning algorithm to identify which patients will go on to develop dementia within a 24 month period with 84% accuracy.
The data used was in the form of PET scans done on volunteers with MCI to measure the concentration of beta-amyloid protein – considered to be a precursor to dementia – in their brain tissue.
According to the research team behind the study, their main goal was to develop a tool for predicting which patients with MCI are likely to develop dementia within a short observation period, and which of them are likely to remain stable.
The necessity for such a tool is grounded in the unreliable nature of beta-amyloid as a biomarker – in other words, not all patients with mild cognitive impairment go on to develop Alzheimer’s disease.
In the future, as more (and more reliable) biomarkers for dementia are discovered, the researchers plan to incorporate them into the algorithm, thereby increasing its predictive capabilities.
By using this tool, clinical trials could focus only on individuals with a higher likelihood of progressing to dementia within the timeframe of a particular study, which might reduce the cost and time necessary to conduct them.
However, given the complicated nature of the quest for (and application of) disease biomarkers in medicine (such as the controversy that surrounds early screening of breast and prostate cancer), before the new algorithm is rolled out, it will have to undergo further extensive study, culminating in certification by relevant authorities.
As history has shown, medical screening, conducted on the basis of biomarkers, often leads to high rates of false positives and negatives, which harm individuals not only by subjecting them to unnecessary treatment, but also by taking a toll on their mental health.
In addition, it is not obviously clear that early intervention would necessarily lead to better clinical outcomes – this will have to be established experimentally.
Regardless of these hurdles, the researchers are optimistic and claim their study is “an example [of] how big data and open science brings tangible benefits to patient care”.
The authors are currently conducting further testing to validate the algorithm in different patient cohorts.