Background: A growing body of evidence suggests that neuroinflammation contributes actively to pathophysiology of Alzheimer's disease (AD) and promotes AD progression. The predictive value of neuroinflammatory biomarkers for disease-staging or estimating disease progression is not well understood. In this study, we investigate the diagnostic and prognostic utility of inflammatory biomarkers in combination with conventional AD biomarkers.
Methods: We included 258 participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) who had CSF biomarkers of β-Amyloid (Aβ), tau, and inflammation. The primary outcome of interest was clinically meaningful cognitive decline (CMCD) as defined by an increase of ≥4 on the Alzheimer's Disease Assessment Scale Cognitive Subscore 11 (ADAS-11, scores 0-70, higher scores indicate worse cognition). Predictor variables were categorized as demographics (D; age, sex, and education), genetic (APOE4 status (A)), inflammatory biomarkers (I), and classic (C) cerebrospinal fluid (CSF) biomarkers of Aβ and p-tau181. Simultaneous inclusion of eleven CSF inflammatory biomarkers as covariates in logistic regression models was examined to assess improvements in classifying baseline clinical diagnoses (cognitively normal (CN), mild cognitive impairment (MCI), Dementia) and predicting individuals with and without CMCD over one year of follow-up.
Results: At 1-year follow up, 27.1% of participants experienced CMCD. Inclusion of inflammatory biomarkers improved baseline classification of CN vs MCI as well as CN vs Dementia for models including D and A variables (DA; both p<0.001). Similarly, when classic CSF biomarkers of AD were included into the model (DAC model), inclusion of inflammatory markers improved classification of CN vs MCI (p<0.01) as well as CN vs Dementia (p<0.001). Addition of inflammatory biomarkers to both DA and DAC models improved predictive performance for CMCD in persons with baseline MCI and Dementia (all p<0.05), but not in the CN group. In addition, the predictive performance of the DAI model was superior to the DAC model in the MCI and Dementia groups (both p<0.05).
Conclusions: Addition of CSF inflammatory biomarkers to CSF biomarkers of AD can improve diagnostic accuracy of clinical disease stage at baseline and add incremental value to AD biomarkers for prediction of cognitive decline.
Keywords: Alzheimer’s disease; Predictive modeling; inflammatory biomarkers; machine learning.