Optimizing plasma biomarkers combinations for the detection of amyloid pathology in cognitively unimpaired individuals

Abstract Background Some Alzheimer’s disease (AD) plasma biomarkers are altered in preclinical stages. However, the optimal combination of plasma biomarkers for detecting preclinical AD has not yet been established. Method We investigated 361 cognitively unimpaired individuals of the ALFA+ cohort with plasma biomarkers: Aβ42/Aβ40, p‐tau181, p‐tau231, GFAP, and NfL (Simoa‐based), and p‐tau217 and t‐tau (MSD‐based). All participants had CSF Aβ42/Aβ40 levels and 328 had [18F]flutemetamol amyloid‐PET. We tested different plasma biomarkers combinations ( z ‐score), with sex and age, in logistic regression models, using Ab‐status (Ab‐positive defined as CSF Aβ42/Aβ40<0.071 or amyloid‐PET>12CL) as dependent variable. For each combination, we selected a “best model” (lowest AIC; BM) and a “parsimonious model” (fewer variables, ΔAIC<2, similar AUC; PM). Biomarkers with p>0.10 were removed in PM. We used DeLong tests to compare AUCs between all the possible combinations in PM and BM, PM, plasma biomarkers alone or age and sex. The statistical threshold was set at p<0.05. Result Table 1 shows sample characteristics. When using CSF Aβ42/Aβ40 to define Ab‐status, BM included the combination of plasma Aβ42/Aβ40, GFAP, p‐tau181, p‐tau217, p‐tau231 and t‐tau (AUC = 0.85), while PM included Aβ42/Aβ40, GFAP, p‐tau217 and p‐tau231 (AUC = 0.84). Plasma Aβ42/Aβ40, GFAP and ptau231 (AUC = 0.83) or Aβ42/Aβ40, ptau217 and ptau231 (AUC = 0.83) retained similar accuracy versus BM/PM, while plasma Aβ42/Aβ40 and ptau231 (AUC = 0.82) was similar versus PM. All these biomarkers’ combinations had a significantly higher AUC than any of the plasma biomarkers alone or age and sex (Table 2). When using amyloid‐PET to define Ab‐status, BM/PM included the combination of plasma Aβ42/Aβ40, GFAP, ptau181, ptau217 and age (AUC = 0.89). Models with similar performance included: Aβ42/Aβ40, GFAP, ptau181, and ptau217 (AUC = 0.88); Aβ42/Aβ40, GFAP, ptau181 and age (AUC = 0.88); Aβ42/Aβ40, ptau181, ptau217 and age (AUC = 0.88); Aβ42/Aβ40, GFAP and ptau181 (AUC = 0.87); Aβ42/Aβ40, ptau181 and ptau217 (AUC = 0.87); Aβ42/Aβ40, GFAP and ptau217 (AUC 0.86); Aβ42/Aβ40, GFAP, ptau217 and age (AUC 0.86). All these biomarkers’ combinations had a significantly higher AUC than any of the plasma biomarkers alone or age and sex, except Aβ42/Aβ40, GFAP and ptau217 versus ptau217 alone (Table 3). Conclusion Different combinations of plasma biomarkers can improve the accuracy to detect Ab‐pathology in cognitively unimpaired individuals and facilitate preclinical AD trials recruitment.