Table 3. Classification performance of combinations of AD-MCI trending miRNA biomarkers, Aβ₄₂:T-Tau, and APOEε4 status.

The boost in the area under the receiver operating characteristic curve (AUC) from adding the trending miRNAs to combinations of the existing predictors is shown as a change in AUC. The miRNAs dramatically outperformed APOEε4 for either diagnosis and added value to an Aβ₄₂:T-Tau prediction of AD. The boost was even more robust for MCI, where Aβ₄₂:T-Tau levels were less predictive of MCI than AD. APOEε4 added little or nothing to models already containing Aβ₄₂:T-Tau information. Omnibus tests of the one-sided hypotheses showed that the trending miRNAs uniformly improved prediction of disease status (i.e. positive change for all 3 models considered, accounting for dependencies among the models) return p-values of 0.024 for MCI and 0.039 for AD.

Marker Combinations					MCI vs. NC		AD vs. NC
					AUC	Change	AUC	Change
				Trending miRNAs	0.705		0.770
		APOEε4			0.639		0.620
		APOEε4	+	Trending miRNAs	0.713	+0.074 (p=0.176)	0.772	+0.152 (p=0.037)
Aβ42:T-Tau					0.758		0.867
Aβ42:T-Tau	+			Trending miRNAs	0.813	+0.055 (p=0.051)	0.903	+0.036 (p=0.113)
Aβ42:T-Tau		APOEε4			0.756		0.881
Aβ42:T-Tau	+	APOEε4	+	Trending miRNAs	0.819	+0.063 (p=0.027)	0.898	+0.017 (p=0.414)