Plasma glial fibrillary acidic protein detects Alzheimer pathology and predicts future conversion to Alzheimer dementia in patients with mild cognitive impairment

Table 3 Combined results from binary logistic regression models and ROC curve analysis for prediction of Aβ-positive status

	p	AIC	AUC	95% CI	Difference AUC (p)	Δ AIC with GFAP
GFAP	< 0.0001	177	0.79	0.72–0.86	–	–
Age	< 0.0001	208	0.64	0.56–0.73	–	–
APOE ε4 status	< 0.0001	187	0.73	0.65–0.81	–	–
GFAP and age	< 0.0001	179	0.78	0.71–0.86	ns	2
GFAP and APOE ε4 status	< 0.0001	152	0.86	0.80–0.92	0.02	−25
GFAP, age, and APOE ε4 status	< 0.0001	153	0.86	0.80–0.92	0.02	−24

Aβ-positive status was defined as CSF Aβ42/40 < 0.07. AIC Akaike information criterion, AUC area under the curve, 95% CI 95% confidence intervals, difference AUC (p) difference between AUCs measured with DeLong test (AUC of GFAP used as reference), Δ AIC difference in AIC (AIC of GFAP used as reference), ns not significant

ISSN: 1758-9193