Table 4. Receiver Operating Curve (ROC) Analysis for Abnormal Aβ-PET in the Alzheimer Disease Neuroimaging Initiative^a.

Plasma assay	Aβ-PET, AUC (95% CI)b
Aβ+, No.	59
Aβ−, No.	63
Aβ42/40 IP-MS-WashU	0.845 (0.772-0.917)
Composite IP-MS-Shim	0.821 (0.747-0.895)
Aβ42/40 IA-Elc	0.740 (0.651-0.829)c
Aβ42/40 IA-N4PE	0.685 (0.590-0.781)d
Aβ42/40 IP-MS-UGOT	0.662 (0.565-0.758)e
Aβ42/40 IA-Quan	0.634 (0.534-0.734)e

Abbreviations: Aβ, amyloid-β; AUC, area under the curve; IA-Elc, Elecsys immunoassay from Roche Diagnostics; IA-N4PE, N4PE Simoa immunoassay from Quanterix; IA-Quan, Simoa immunoassay from Quanterix; IP-MS-Shim, immunoprecipitation coupled mass spectrometry method developed by Shimadzu; IP-MS-UGOT, immunoprecipitation-coupled mass spectrometry method developed at the University of Gothenburg; IP-MS-WashU, immunoprecipitation-coupled mass spectrometry method developed at Washington University; ROC, receiver operating characteristic; PET, positron emission tomography.

^a In ADNI, CSF Aβ42 and Aβ40 measures at the time of plasma collection were only available in a small group of participants, and therefore we used Aβ PET as the outcome.

^b AUCs of 2 ROC curves were compared with DeLong test. Aβ-PET data was binarized using a previously described threshold of 1.11. ^14,15

^c P < .05, compared with IP-MS-WashU Aβ42/40.

^d P < .01, compared with IP-MS-WashU Aβ42/40.

^e P < .001, compared with Aβ42/40IP-MS-WashU.