A new study, published today in the Neurology journal, describes prediction models which can provide personalised information on future cognitive decline for patients with early Alzheimer’s disease. According to the study authors, these models could also inform discussions between doctors and patients on whether or not to start treatment with medicines including disease-modifying therapies such as lecanemab and donanemab.
Led by Pieter van der Veere and Wiesje van der Flier of Amsterdam University Medical Center, the study used data from the Amsterdam Dementia Cohort to construct prediction models of cognitive decline in amyloid-positive patients with mild cognitive impairment (MCI) or mild dementia due to Alzheimer’s disease. In total, 961 participants were included, 310 of whom had MCI, and 651 of whom had mild dementia. The main cognitive outcome for the study was the MMSE score, a measurement of cognitive function that assesses attention, memory, language, ability to follow instructions, and orientation.
The prediction models integrated many different data sources, including cognitive test scores, amyloid positivity (in cerebrospinal fluid or PET scans) and MRI measurements of brain volume and structure. Statistical models were used to calculate predicted decline patterns for individuals with MCI or mild dementia, with validation performed using data from the US-based Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. Finally, the models were embedded in a prototype shiny app and calculator, for ease of use by clinicians.
The resulting prediction models could be used to forecast the rate and speed of cognitive decline, predicting the time to reach a certain score on the MMSE or RAVLT cognitive tests. According to the authors, the overall performance of both models showed that variation between individual trajectories could be explained by factors such as age, sex, and baseline cognitive test score. They also highlighted the relatively small number of variables required, and the simplicity of the statistical modelling approach, which means that these models could be easier to apply in real-world clinical practice. Read the full article in Neurology:
https://www.neurology.org/doi/full/10.1212/WNL.0000000000209605
