Combination of plasma amyloid beta(1-42/1-40) and glial fibrillary acidic protein strongly associates with cerebral amyloid pathology

Table 2 AUC and sensitivity and specificity at Youden’s cutoff to identify an abnormal amyloid PET scan in the total study population and in the non-demented subset

	AUC (95% CI)	Youden’s cut-point	Sensitivity (%)	Specificity (%)
Total population
Plasma Abeta_(1-42/1-40)	73% (66–81%)	0.16	70	76
Plasma GFAP	81% (75–87%)	125 pg/mL	73	79
Plasma NfL	71% (64–79%)	11.5 pg/mL	73	64
Panel*	*88% (83–93%)*	–	82	86
Non-demented subset (SCD + MCI)
Plasma Abeta_(1-42/1-40)	67% (57–78%)	0.16	72	65
Plasma GFAP	76% (67–85%)	108 pg/mL	75	69
Plasma NfL	63% (53–73%)	11.9 pg/mL	61	67
*Panel*^ǂ	*84% (76–92%)*	–	70	86

AUC with 95% confidence interval was calculated using receiver operator curve (ROC) analysis. Youden’s cut-point is at the coordinates of the ROC curve where a maximum sum of sensitivity and specificity is reached. For the single markers, this results in a useable cutoff thus presented here, whereas for the panels, this is a predicted value from the logistic regression model. The panels were established using an automated Wald’s backward selection procedure among plasma markers Abeta_(1-42/1-40), GFAP, NfL, age, sex, and APOE ε4 carriership. Predicted values of the logistic regression analysis are used for ROC analysis
Abeta amyloid beta, GFAP glial fibrillary acidic protein, NfL neurofilament light, SCD subjective cognitive decline, MCI mild cognitive impairment, AUC area under the curve, 95%CI 95% confidence interval
*For the total population, the panel includes plasma Abeta_(1-42/1-40), plasma GFAP, APOE ε4 carriership, and age
^ǂFor the non-demented subset (SCD and MCI), the panel includes Abeta_(1-42/1-40), plasma GFAP, and APOE ε4 carriership

ISSN: 1758-9193