Comparing metabolomic and pathologic biomarkers alone and in combination for discriminating Alzheimer’s disease from normal cognitive aging

Table 2 Summary measures of model fit for each resulting model for discriminating between AD vs. controls in the full dataset

Model variables			Model variable abbreviations	AUC	Sensitivity	Specificity
E	M	P
X			E	0.96	1.00*	0.90
	X		M	0.70	0.68	0.67
		X	P	0.92	0.97	0.83
X		X	P\|E	0.90	0.93	0.89
	X	X	P\|M	0.89	0.88	0.88
X	X	X	P\|M\|E	0.90*	0.92	0.89

These results are the average measures of model fit in the testing (not training) sets across all cross validation intervals. The variables included in the modeling process are highlighted in indicator columns, along with their abbreviations. The data types are Luminex amyloid and tau-related proteins (P), GC-TOF mass spectrometry metabolites (M) and LC-ECA metabolites (E).
Abbreviations: AD Alzheimer’s disease, AUC Area under the curve. *rounded.

ISSN: 2051-5960