Normalizing Alzheimer's Biomarkers to Reference Proteins Dramatically Boosts Accuracy
Dividing CSF and plasma AD biomarkers by reference proteins like Aβ40 or np-tau sharply improves their agreement with brain PET imaging.
Summary
Researchers found that normalizing Alzheimer's disease fluid biomarkers—such as p-tau217 and MTBR-tau243—to reference proteins like amyloid-beta 40 (Aβ40) or non-phosphorylated tau (np-tau) substantially strengthens their correlation with tau-PET and amyloid-PET brain scans. Studying over 1,700 participants across four independent cohorts, the team showed that simple ratio calculations can boost biomarker accuracy metrics (R²) by up to 0.28 for amyloid pathology and up to 0.13 for tau pathology. The approach works for both CSF and plasma samples, works across multiple assay platforms, and reduces between-person variability without harming sensitivity to disease-related changes. These findings suggest that reference-protein normalization could make already strong AD fluid biomarkers even more reliable for diagnosis and clinical trial monitoring.
Detailed Summary
Accurate fluid biomarkers are critical for detecting and staging Alzheimer's disease (AD), especially as disease-modifying therapies become available. However, a recognized but underaddressed problem is that total protein abundance in CSF and plasma varies substantially between individuals for reasons unrelated to AD pathology—dilution differences, production rates, clearance—introducing noise that reduces biomarker precision. This study systematically tested whether dividing AD-relevant biomarkers by 'reference proteins' that track this non-AD variability could sharpen their agreement with gold-standard PET imaging.
The primary cohort was the Swedish BioFINDER-2 study (n=1,702; mean age 68.4 years), with replication in BioFINDER-1, the Knight Alzheimer Disease Research Center (ADRC), and the TRIAD cohort. Investigators also included an Italian multiple sclerosis cohort as a disease-specificity control. Biomarkers measured included CSF and plasma p-tau217, p-tau181, p-tau205, multiple non-phosphorylated tau (np-tau) fragments, Aβ42, Aβ40, CSF MTBR-tau243, SNAP-25, neurogranin, YKL-40, sTREM2, and plasma eMTBR-tau243, using both mass spectrometry and immunoassays. PET outcomes were tau-PET (flortaucipir) and Aβ-PET (florbetapir/florbetaben) standardized uptake value ratios. Univariate linear regression R² values were compared before and after normalization.
Normalizing CSF biomarkers to Aβ40 produced the most widespread improvements. The association of CSF MTBR-tau243 with tau-PET rose from R²=0.65 to R²=0.78 after Aβ40 normalization—one of the largest single-biomarker improvements observed. Synaptic markers (SNAP-25, neurogranin) and p-tau isoforms also gained substantially (ΔR²=0.064–0.24 for tau-PET; ΔR²=0.016–0.28 for Aβ-PET). Normalization to np-tau was most beneficial for amyloid-PET associations; the CSF p-tau217/np-tau ratio achieved R²=0.65 versus 0.46 for unnormalized p-tau217. In plasma, eMTBR-tau243/np-tau reached R²=0.72 for tau-PET (up from 0.60), and p-tau217/np-tau reached R²=0.62 for Aβ-PET (up from 0.53). Longitudinal analyses indicated that normalization primarily reduced inter-individual rather than intra-individual variability, explaining why it narrows noise without suppressing true disease-related signals over time.
Importantly, normalization did not improve inflammatory CSF biomarkers (YKL-40, sTREM2) in AD, and it did not improve biomarker associations in the multiple sclerosis cohort, demonstrating that the benefit is disease-pathway-specific rather than a blanket mathematical artifact. Results were replicated across three independent international cohorts and across multiple assay platforms.
These findings have immediate practical implications. Reference-protein normalization is computationally trivial—it requires only the additional measurement of Aβ40 or np-tau, which are already commonly included in AD biomarker panels. Implementing ratios could meaningfully reduce misclassification at borderline thresholds, improve clinical trial stratification, and strengthen the correlation between fluid tests and the PET imaging they are designed to replace. Caveats include the cross-sectional nature of most primary analyses, reliance on cohorts that skew toward memory clinic populations, and the need for further standardization of np-tau assays across laboratories before widespread clinical adoption.
Key Findings
- CSF MTBR-tau243/Aβ40 achieved R²=0.78 with tau-PET, up from R²=0.65 unnormalized—the largest improvement seen.
- CSF p-tau217/np-tau reached R²=0.65 with amyloid-PET, versus R²=0.46 for non-normalized p-tau217.
- Plasma eMTBR-tau243/np-tau improved tau-PET association from R²=0.60 to R²=0.72.
- Normalization reduced inter-individual variability over time without suppressing longitudinal disease signals.
- Benefits were AD-pathway-specific: inflammatory markers and MS cohort biomarkers showed no improvement.
Methodology
Cross-sectional and longitudinal univariate linear regression analyses in BioFINDER-2 (n=1,702) compared R² values of fluid biomarkers—measured by mass spectrometry and immunoassay—against tau-PET and Aβ-PET before and after normalization to Aβ40 or np-tau. Findings were replicated in three independent cohorts (BioFINDER-1, Knight ADRC, TRIAD) and tested for disease specificity in a multiple sclerosis cohort.
Study Limitations
Most primary analyses are cross-sectional, limiting causal inference about biomarker trajectories. Cohorts are predominantly memory clinic populations, which may not represent the general population. Np-tau assays lack full inter-laboratory standardization, which must be addressed before clinical implementation of np-tau-normalized ratios.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: