Researchers have developed a model that findings show is able to predict dementia up to nine years before a clinical diagnosis. With greater than 80% accuracy, the test is potentially more reliable than current methods in common use for diagnosis.
In an article published in Nature Mental Health, the research team from Wolfson Institute of Population Health (WIPH) at Queen Mary University of London, led by Charles Marshall, detailed how they developed their neurobiological model and explained their findings.
Noting that "altered functional connectivity precedes structural brain changes and symptoms in dementia," they first analyzed over 1,100 functional MRI (fMRI) scans from the UK Biobank to estimate the effective connectivity between the 10 regions of the brain that constitute the default mode network (DMN), with the aim of being able to detect changes.
The researchers then assigned each patient a probability of dementia value based on the extent to which their effective connectivity pattern was more consistent with a pattern indicative of dementia or a pattern consistent with the control group. They then compared these predictions with the recorded UK Biobank medical data for each patient.
Their findings showed that the model they developed predicted the onset of dementia up to nine years before an official diagnosis was made with greater than 80% accuracy. In the cases where the volunteers had gone on to develop dementia, it was also found that the model was able to predict within two years the time until dementia diagnosis.
An analysis of the association of major risk factors for dementia with effective connectivity yielded the finding that both genetic risk for Alzheimer's disease and social isolation were associated with changes in the DMN.
The WIPH researchers acknowledged some limitations to their study. Among the limitations, they noted that fMRI is expensive, and its signal can be degraded in the presence of excessive head motion. Furthermore, they noted that the UK Biobank participants from which they sampled are generally healthier, socioeconomically privileged, and predominantly white, cautioning that the "generalizability of these results to a more representative sample needs to be assessed." In addition, they expressed uncertainty around how generalizable the effective connectivity-based models could be.
However, they concluded that their findings show that their test can be used as a noninvasive biomarker for predicting future dementia with a high degree of accuracy.
