The study aims to determine whether diabetes impacts Alzheimer’s disease [AD] blood biomarkers in a manner that may contribute to false-positive indications of AD in diagnostic models.

Baseline data were analyzed among n=354 non-Hispanic Whites[NHW] (n=54 diabetic), n=293 Hispanics (n=92 diabetic), and n=729 non-Hispanic Blacks[NHB] (n=188 diabetic) participants involved in the Health and Aging Brain Study-Health Disparities (HABS-HD). Plasma biomarkers Aβ40, Aβ42, Total tau, neurofilament light chain (NfL) and ptau181 were assayed with single molecule array (SIMOA) technology. Support Vector Machine (SVM) models were conducted using plasma biomarkers to predict cognitive diagnosis (Cognitively Unimpaired[CU] or dementia). Discriminative analysis were used to identify false positives. Fisher’s Exact tests were used to explore whether the categorical variable of having diabetes was significantly associated with the false positive classification. Significance was set at p<0.05.

SVM models distinguishing CU from dementia yielded n=133 false positives (By race/ethnicity: NHB=54, Hispanic=51, NHW=28). The SVM performed at a sensitivity of 87.50%, specificity of 69.79%, and reached an area under the curve of 84.38%. Among false positive cases, there was statistical significance across all race/ethnic groups in those with diabetes (versus without) in the full and reduced models. For NHB and NHW participants, amyloid-only models showed statistical significance. For Hispanic participants, all biomarker models (full and reduced) showed statistical significance by diabetes status.

These findings demonstrate that diabetes was significantly related with false-positive classifications of AD blood biomarker-based SVM models—particularly among Hispanic and Black participants.