B-272 Performance of CSF Alzheimer’s biomarker assays across two laboratories: agreement, variability, and clinical implications

Abstract

Early and accurate diagnosis of Alzheimer’s disease (AD) is essential because disease modifying anti amyloid therapies are indicated for early symptomatic patients and require biomarker confirmation of amyloid pathology. Cerebrospinal fluid (CSF) biomarkers play a key clinical role, with the Aß42/Aß40 ratio showing strong concordance with amyloid PET imaging. However, implementing CSF beta amyloid (Aß) testing remains challenging due to its marked sensitivity to pre analytical factors such as collection materials, transfers, blood contamination, freeze-thaw cycles, and handling. Standardizing procedures across laboratories is therefore crucial to ensure reliable measurements, robustness and reproducibility in real world conditions. This study aimed to perform an interlaboratory validation of four CSF assays for AD biomarkers between two Brazilian laboratories, one serving as a support laboratory that provides pre analytical training to its network. Four EUROIMMUN immunossays were evaluated: Beta Amyloid 1 40, Beta Amyloid 1 42, pTau 181 and Total Tau. Fourteen CSF samples were analyzed in Lab A using an automated platform and in Lab B through manual processing. Results were calculated according to the manufacturers’ instructions. Biomarker classifications were assigned using the AT(N) system. Qualitative agreement was assessed using Cohen’s kappa (?). Quantitative agreement was evaluated using Pearson correlation coefficients (r) and linear regression (slope). Interlaboratory agreement was strong across biomarkers. Aß42 showed very good agreement (?=0.87), with the only mismatch attributed to a borderline sample. Agreement for the Aß42/Aß40 ratio, pTau, and Total Tau was good (?=0.71, ?=0.71, and ?=0.78, respectively). For Total Tau, two classification shifts were associated with borderline results. Two additional samples shifted from negative in Lab B to positive in Lab A for both the Aß42/Aß40 ratio and pTau. A slight trend toward higher values in Lab A was observed, particularly for Aß40. Because no borderline interval exists for these parameters, small variations near cutoffs may lead to different interpretations, consistent with the discordant samples. For the Aß42/Aß40 ratio, the discrepancy appeared predominantly driven by Aß40, which showed greater interlaboratory variation than Aß42. Quantitative correlations were high for all biomarkers: Aß40 (r=0.87; slope=0.98), Aß42 (r=0.97; slope=0.85), pTau (r=0.99; slope=1.2), and Total Tau (r = 0.95; slope = 0.94). Aß40 showed greater variability (R² = 0.75), reflected in classification shifts affecting the ratio. In the AT(N) evaluation, borderline values were treated as negative, resulting in one sample shifting from A+T-N- (Lab A) to A-T-N- (Lab B). Two samples near the pTau cutoff shifting from A+T-N- (Lab B) to A+T+N- (Lab A), although both remained within the Alzheimer’s continuum (A+). The overall AT(N) agreement rate was 92.9%. These results demonstrate good interlaboratory concordance, reproducibility, and consistent interpretation across biomarkers, despite minor discrepancies near clinical cutoffs, the influence of pre analytical factors, and differences in sample processing (automated versus manual). Altogether, the findings support the reliable use of these CSF AD assays in routine clinical practice and reinforce their ability to generate high quality results that confidently support early AD diagnosis.