Association of Extracellular Vesicle Biomarkers With Alzheimer Disease in the Baltimore Longitudinal Study of Aging.
Level 4 - case-series / case-control
Case-control study nested within a longitudinal cohort with an independent validation cohort.
PubMed 31305918 · doi:10.1001/jamaneurol.2019.2462
What was done
This nested case-control study analyzed 887 longitudinal plasma samples from 350 participants in the Baltimore Longitudinal Study of Aging (BLSA), comparing 128 cognitively normal participants who later developed Alzheimer disease (AD) against 222 age- and sex-matched controls who remained cognitively normal (mean follow-up 3.5 years, range 0–9.73 years). Neuronal-enriched extracellular vesicles (nEVs) were isolated by immunoprecipitation to quantify tau (p-tau181, p-tau231, total tau), amyloid-beta (Aβ42), and phosphorylated insulin receptor substrate 1 (pSer312-IRS-1, pY-IRS-1), alongside nEV diameter and concentration. A predictive model was developed in a training set (n=161) and tested in a validation set (n=80). An independent cohort from the Johns Hopkins Alzheimer Disease Research Center (JHADRC; 35 AD cases, 29 controls; 128 samples) was also evaluated.
What was found
Preclinical AD cases in BLSA showed longitudinally higher p-tau181, p-tau231, pSer312-IRS-1, pY-IRS-1, and nEV diameter compared with controls, but showed similar Aβ42, total tau, TSG101, and nEV concentration. In the BLSA training set, a multimarker model achieved an area under the curve (AUC) of 89.6%, sensitivity of 81.8%, and specificity of 85.8%. In the BLSA test set, the model achieved an AUC of 80.0%, sensitivity of 55.6%, and specificity of 88.7%. In the JHADRC cohort, discrimination achieved an AUC of 98.9% (100% sensitivity, 94.7% specificity) in the training set and an AUC of 76.7% (91.7% sensitivity, 60.0% specificity) in the test set.
Why it matters
These findings suggest that longitudinal blood-based biomarkers measuring neuronal-derived extracellular vesicle phosphorylated tau and insulin signaling proteins can signal preclinical Alzheimer disease years before clinical symptom onset.
Limits
Performance dropped substantially in test sets, particularly test sensitivity in the BLSA cohort (55.6%) and test specificity in the JHADRC cohort (60.0%). The study relied on retrospective case-control matching with relatively small sample sizes in the validation sets, and nEV isolation requires specialized immunoprecipitation workflows not yet standardized for high-throughput clinical practice.
Cited by
- contradicts Blood biomarkers measuring inactive phosphorylated insulin receptor substrate 1 (IRS-1) can predict Alzheimer's disease diagnosis 10 years in advance with 100% accuracy.