Discordant results in CSF/PET Aβ biomarkers are dependent on APOE genotype and related to inter‐individual differences in CSF Aβ42, Aβ40 and Aβ38 isoforms

Background Recently, we provided evidence that discordance in CSF vs. PET Aβ‐biomarkers represents a natural phase in the progression of Aβ‐pathology, with either CSF or PET becoming abnormal first (Figure 1). The reasons why CSF‐Aβ42 becomes abnormal first in some subjects, while Aβ‐PET does in others, are unknown. Here, we aimed to investigate biomarker and genetic factors characterizing these alternative “CSF+ first” vs. “PET+ first” pathways of Aβ accumulation. Method We retrospectively selected N=867 cases from ADNI, ranging from cognitively normal to overtly demented, with [18F]Florbetapir‐PET and CSF‐Aβ42 measurements obtained within three months; a subset (N=289) also had two‐year longitudinal CSF/PET data. Each participant was classified as belonging to the “CSF+ first” or “PET+ first” pathway based on modelling of the continuous distribution of CSF‐Aβ42 and Aβ‐PET measures to a hyperbolic function, followed by residuals analysis, and avoiding the use of CSF or PET cut‐off values. This method allowed to assign each subject to a “CSF+ first” vs. “PET+ first” pathway and yielded highly stable results at two‐year follow‐up (Figure 2). Result Significant differences between participants belonging to the “CSF+ first” vs. “PET+ first” pathways were found in biomarkers for Aβ‐pathology (Figure 3) but not for tau or neurodegeneration, with CSF‐Aβ42 driving most of the difference (partial η 2 =0.67 [CSFAβ42]; partial η 2 =0.29 [Aβ‐PET]). Participants belonging to the “CSF+ first” pathway showed significantly lower levels of CSF‐Aβ42, Aβ40 and Aβ38 compared to their “PET+ first” counterparts, consistently across all concordant and discordant CSF/PET stages (Figure 4). At the genetic level, a significantly higher prevalence of homozygous carriers of APOE‐ε4 was found in the “CSF+ first” vs. “PET+ first” pathway (18.9% vs. 1.6%; Figure 5). No differences in modifiable risk/protective factors for Alzheimer’s disease were found. Conclusion This evidence argues against the claim that discordant Aβ CSF vs. PET findings systematically represent false positive/negative results. Instead, our findings suggest that CSF vs. PET discordance carries relevant information reflecting differences in Aβ clearance mechanisms, with genetic determinants partly leading each subject towards a CSF+ first or a PET+ first trajectory. Combination of CSF/PET Aβ biomarkers might help identifying subjects most likely to benefit from treatments targeting Aβ clearance.