Incremental Value of Baseline Cognition, APOE ε4, and Amyloid-beta for Predicting Progression to Mild Cognitive Impairment in Cognitively Normal Older Adults (21917)
Objective: Amyloid-beta (Aβ) deposition predicts cognitive decline in preclinical Alzheimer's disease; however, measuring Aβ levels is not always possible. In lieu of Aβ quantification, apolipoprotein (APOE) ε4 allele carriage and cognitive test scores, which are both associated with Aβ deposition, may be useful in the prediction of cognitive impairment. We aimed to evaluate the incremental value of cognition, APOE ε4, and baseline Aβ status to predict progression to mild cognitive impairment (MCI).
Methods: Baseline data of 1,964 cognitively unimpaired (CU) participants were obtained from the ADOPIC consortium. Cognitive composite scores were created for the domains of memory, executive function, and attention, derived from principal components analysis. We trained four random forest classifiers on 1,000 bootstrapped samples to predict progression to MCI at 8.35 years from baseline. The first uses age, sex, and education as predictors; the second adds the three cognitive composites; the third and fourth add APOE ε4 and baseline Aβ status, respectively. Performance was measured using ROC-AUC scores.
Results: At 8.35 years follow-up, 255 participants had progressed to MCI, while 1,709 remained CU (Table 1). Demographics as predictors resulted in a ROC-AUC of 0.57. Adding cognition, APOE ε4, and baseline Aβ status improved ROC-AUC scores to 0.68, 0.70, and 0.70 respectively (Figure 1). Although APOE ε4 carriage improved the model prediction, adding baseline Aβ status did not improve it further.
Conclusion: In the absence of biomarkers, neuropsychological assessment and demographic characteristics can provide an acceptable and accurate prediction of the future risk of MCI over 8.35 years.