New Models Predict Time From Mild Cognitive Impairment to Dementia

Pauline Anderson

Using a large, real-world population, researchers have developed models that predict cognitive decline in amyloid-positive patients with either mild cognitive impairment (MCI) or mild dementia.

While the models may help clinicians better answer common questions from their patients about their rate of cognitive decline, the investigators, led by Pieter J. van der Veere, MD, Alzheimer Center and Department of Neurology, Amsterdam Neuroscience, VU University Medical Center, Amsterdam, the Netherlands, noted.

The findings were published online on July 10 in Neurology.

Easy-to-Use Prototype

On average, it takes 4 years for MCI to progress to dementia. While new disease-modifying drugs targeting amyloid may slow progression, whether this effect is clinically meaningful is debatable, the investigators noted.

Earlier published models predicting cognitive decline either are limited to patients with MCI or haven't been developed for easy clinical use, they added.

For the single-center study, researchers selected 961 amyloid-positive patients, mean age 65 years, who had at least two longitudinal Mini-Mental State Examinations (MMSEs). Of these, 310 had MCI, and 651 had mild dementia; 48% were women, and over 90% were White.

Researchers used linear mixed modeling to predict MMSE over time. They included age, sex, baseline MMSE, apolipoprotein E ε4 status, cerebrospinal fluid (CSF) beta-amyloid (Aβ) 1-42 and plasma phosphorylated-tau markers, and MRI total brain and hippocampal volume measures in the various models, including the final biomarker prediction models.

At follow-up, investigators found that the yearly decline in MMSEs increased in patients with both MCI and mild dementia. In MCI, the average MMSE declined from 26.4 (95% CI, 26.2-26.7) at baseline to 21.0 (95% CI, 20.2-21.7) after 5 years.

In mild dementia, the average MMSE declined from 22.4 (95% CI, 22.0-22.7) to 7.8 (95% CI, 6.8-8.9) at 5 years.

The predicted mean time to reach an MMSE of 20 (indicating mild dementia) for a hypothetical patient with MCI and a baseline MMSE of 28 and CSF Aβ 1-42 of 925 pg/mL was 6 years (95% CI, 5.4-6.7 years).

However, with a hypothetical drug treatment that reduces the rate of decline by 30%, the patient would not reach the stage of moderate dementia for 8.6 years.

For a hypothetical patient with mild dementia with a baseline MMSE of 20 and CSF Aβ 1-42 of 625 pg/mL, the predicted mean time to reach an MMSE of 15 was 2.3 years (95% CI, 2.1-2.5), or 3.3 years if decline is reduced by 30% with drug treatment.

External validation of the prediction models using data from the Alzheimer's Disease Neuroimaging Initiative, a longitudinal cohort of patients not cognitively impaired or with MCI or dementia, showed comparable performance between the model-building approaches.

Researchers have incorporated the models in an easy-to-use calculator as a prototype tool that physicians can use to discuss prognosis, the uncertainty surrounding the predictions, and the impact of intervention strategies with patients.

Future prediction models may be able to predict patient-reported outcomes such as quality of life and daily functioning, the researchers noted.

"Until then, there is an important role for clinicians in translating the observed and predicted cognitive functions," they wrote.

Compared with other studies predicting the MMSE decline using different statistical techniques, these new models showed similar or even better predictive performance while requiring less or similar information, the investigators noted.

The study used MMSE as a measure of cognition, but there may be intraindividual variation in these measures among cognitively normal patients, and those with cognitive decline may score lower if measurements are taken later in the day. Another study limitation was that the models were built for use in memory clinics, so generalizability to the general population could be limited.

The study was supported by Eisai, ZonMW, and Health~Holland Top Sector Life Sciences & Health. See paper for financial disclosures.

 

TOP PICKS FOR YOU
Recommendations

3090D553-9492-4563-8681-AD288FA52ACE