Work

P2‐055: Refinement and expansion of the Texas Alzheimer's research consortium serum‐based diagnostic algorithm for Alzheimer's disease

Year: 2011

Type: article

Abstract: We previously created a serum-based algorithm that yielded excellent diagnostic accuracy in distinguishing AD from normal controls. The current project was designed to refine that algorithm by reducing the number of serum proteins and inclusion of additional clinical variables. Additionally, we sought to determine if the serum-biomarker algorithm was significantly related to neuropsychological performance. We analyzed serum-protein multiplex biomarker data from 197 patients diagnosed with AD and 200 cognitively normal controls from the Texas Alzheimer's Research & Care Consortium (TARCC) longitudinal cohort. The biomarker algorithm was restricted to only the top 30 markers from the original publication. Additional clinical variables added to the algorithm included cholesterol, triglycerides, HDL, LDL, LpPLA2, homocysteine and C-peptide. The biomarker risk score utilizing only the top 30 markers was generated from a training set containing one half of the AD cases and controls, which was then applied to the test sample. When the biomarker risk score from the top 30 markers was applied to the test sample, the sensitivity, specificity, and AUC were .69, .90, and .80 reflecting a decline from the risk score derived from the larger panel. However, when the demographic data (age, gender, education) and clinical data (see above in addition to APOE genotype) were combined with the biomarker risk score, the SN, SP, and AUC rose to .85, .87, and .94 which is comparable to the previously published algorithm. In linear regression models, the biomarker risk score was most strongly related to neuropsychological tests of language (COWAT and BNT) and memory (WMS LM I and II; Delayed Recall; WMS VR I and II). It was also a powerful predictor of global cognition (MMSE) and disease severity (CDR scores). These data suggest that, when combined with additional clinical lab data, the algorithm can be reduced from 100+ serum-based proteins to 30 and still retain excellent diagnostic accuracy. Additionally, the biomarker risk score is significantly related to neuropsychological test performance, most heavily with the domains of memory and language.

Source: Alzheimer s & Dementia

DOI: 10.1016/j.jalz.2011.05.944