Personalized Alzheimer’s Progression Predicted by Model

A model was developed to predict cognitive decline in individuals with mild cognitive impairment or mild dementia due to Alzheimer’s disease.

From the Amsterdam Dementia Cohort, researchers selected amyloid-beta (Aβ) positive participants with mild cognitive impairment or mild dementia and at least two longitudinal Mini-Mental State Examination (MMSE) measurements to predict MMSE scores over time. MMSE scores range from 0 to 30; scores lower than 10 indicate severe dementia.

A hypothetical Alzheimer’s patient with mild cognitive impairment, a baseline MMSE score of 28, and cerebrospinal fluid (CSF) Aβ1–42 of 925 pg/mL was predicted to reach an MMSE score of 20 after 6 years, reported Pieter van der Veere, MD, of Amsterdam University Medical Center in the Netherlands, and co-authors.

For a patient with mild dementia with a baseline MMSE of 20 and CSF Aβ1–42 of 625 pg/mL, the predicted time to reach an MMSE score of 15 was 2.3 years, the researchers wrote in Neurology.

With a hypothetical intervention that could reduce decline by 30%, the predicted time to reach those thresholds would be 8.6 years for the patient with mild cognitive impairment, and 3.3 years for the patient with mild dementia.

“In this study, we make a first effort to model the decline on a cognitive test over time in individual patients with mild cognitive impairment or mild dementia due to Alzheimer’s disease,” said co-author Wiesje van der Flier, PhD, also of Amsterdam University. “This could provide an answer to the questions many patients ask — namely, ‘what can I expect, what is my prognosis?'”

The personalized predictions can give patients a general indication of their Alzheimer’s course, the researchers said.

“The model provides a prediction, but is by no means perfect,” van der Flier told MedPage Today. “There remains uncertainty.” A graphical display visualizes this uncertainty, allowing a conversation about it, she added.

“We are currently building this into an online tool, which allows doctors to use it in the context of research,” van der Flier said.

“It is not yet available for clinical practice, because it needs further refinement,” she continued. “The online tool also includes patient-facing information about the diagnosis, diagnostic tests, and prognosis.”

The model was based on data from 961 patients with Alzheimer’s disease, including 310 people with mild cognitive impairment and 651 with mild dementia. It incorporated age, sex, cognitive test scores, APOE4 status, amyloid PET or CSF biomarker data, and MRI total brain and hippocampal volume.

Mean age of participants was 65, and 49% were women. The mild cognitive impairment group had a median follow-up of 3 years with an average of four MMSE measurements. The mild dementia group had a median follow-up of 2 years with an average of three MMSE measurements.

Cognitive decline rates increased over time in both groups. In the mild cognitive impairment group, mean MMSE scores dropped from 26.4 to 25.8 after 18 months, to 24.2 after 3 years, and to 21.0 after 5 years. In the mild dementia group, mean MMSE scores fell from 22.4 to 19.8, 15.3, and 7.8, respectively.

In the mild cognitive impairment group, the mean R² and median absolute deviation in internal cross-validation were 0.17 and 2.05, respectively. In the mild dementia group, these values were 0.26 and 2.83. “This means that in half of the predictions made for patients with mild cognitive impairment, the observed MMSE deviated by less than 2 points from the predicted MMSE,” van der Veere and co-authors wrote. “Correspondingly, the deviation was less than approximately 3 points in mild dementia.”

External validation in a cohort from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) showed comparable performance.

Overall performance indicated that a substantial amount of variation in MMSE decline could be explained by age, sex, baseline MMSE, and time since baseline, van der Veere and colleagues noted.

“Additional information on MRI volumetric and CSF Aβ1–42 and phosphorylated tau biomarkers, representing etiologic disease characteristics, aided in the prediction of MMSE decline in our amyloid-positive sample,” the researchers wrote.

“However, further increasing model complexity by adding other clinical and vascular risk factors did not improve predictive performance despite their known association with Alzheimer’s disease dementia,” they pointed out. “Potentially, tau PET information could improve predictive performance because of the association with Alzheimer’s disease-associated symptom severity, but we could not incorporate this because of a lack of data.”

The model has other limitations, the researchers acknowledged. It relied on MMSE, which may have intra-individual variation. It also was built for use in memory clinics and may not apply to the general population.

Please enable JavaScript to view the

comments powered by Disqus.