Multiomics Blood-Based Biomarkers Predict Alzheimer's Predementia with High Specificity in a Multicentric Cohort...

Area of Science:

• Neurology
• Biomarker Discovery
• Medical Diagnostics

Background:

• Current Alzheimer's Disease (AD) diagnosis relies on cognitive tests and amyloid plaque detection, which have limited specificity, leading to potential misdiagnosis of mild cognitive impairment (MCI).
• Up to 25% of amyloid-positive individuals without dementia may be incorrectly diagnosed with MCI due to AD, highlighting the need for more precise diagnostic tools.
• The advent of anti-amyloid therapies necessitates accurate identification of patients who will truly benefit, particularly distinguishing AD from non-AD pathologies.

Purpose of the Study:

• To evaluate a predictive model using peripheral biomarkers for highly specific identification of MCI and mild dementia patients who will progress to AD dementia.
• To develop and validate a diagnostic tool capable of differentiating AD from non-AD brain disorders in a cognitively impaired population.

Main Methods:

• Peripheral biomarkers (45 proteins, 36 metabolites) identified in an animal model were targeted using mass spectrometry assays.
• A machine learning model, B-HEALED, was developed and trained on 70% of blood sample data from 345 participants across 7 cohorts, analyzing 81 biomarkers and age.
• The model's predictive accuracy was validated blindly on the remaining 30% of data and in external datasets, assessing its ability to distinguish AD patients from non-AD subjects.

Main Results:

• The B-HEALED model, using 19 biomarkers and age, achieved 93.0% specificity and 65.4% sensitivity (AUROC=81.9%) in internal validation for identifying individuals who will develop AD dementia.
• When combined with amyloid status (CSF or PET), the B-HEALED model achieved 100% specificity and 52.8% sensitivity in identifying AD.
• The model demonstrated consistent performance in blind external validation, confirming its reliability on independent datasets.

Conclusions:

• The B-HEALED test, employing multiomics blood biomarkers, offers high predictive specificity for AD detection in cognitively impaired individuals, significantly reducing false positives.
• When used with amyloid screening, B-HEALED effectively identifies a pure prodromal AD cohort, crucial for clinical trial recruitment and treatment selection.
• These findings support the optimization of clinical trial criteria, drug development, and the precise identification of patients most likely to benefit from disease-modifying AD therapies.