CSF biomarker variability in the Alzheimer's Association quality control program

Abstract

The cerebrospinal fluid (CSF) biomarkers amyloid-β 1–42 (Aβ42), total-tau (T-tau), and phosphorylated-tau (P-tau) are increasingly used for Alzheimer's disease (AD) research and patient management. However, in addition to significant differences between platform-specific commercial assay kit results, there are still large variations in biomarker measurements between and within laboratories. The main sources of this variability remain undefined. Here data was used from the Alzheimer's Association quality control (QC) program to specifically identify sources of analytical variability. Data from the first nine rounds of the Alzheimer's Association QC program was used. In each round, three CSF samples (aliquots of pooled human CSF) prepared at the Clinical Neurochemistry Laboratory at the University of Gothenburg, Sweden, were analyzed by participating laboratories for tau and Aβ proteins by single analyte enzyme-linked immunosorbent assays (ELISA), a multiplexing xMAP assay or an immunoassay with electrochemiluminescence detection (Meso Scale Discovery; MSD). Moreover, 5 laboratories with extensive experience in performing these tests served as reference laboratories and analyzed all samples in 6 runs. A total of 84 laboratories participated. Coefficients of variation (CV) between the laboratories were around 20–30%, while within-run CV (in each laboratory) was smaller (< 5–10%). Longitudinal within-laboratory CV was 5–19%. Interestingly, longitudinal within-laboratory CVs differed considerably between biomarkers at individual laboratories, suggesting that a component of the variability was assay-dependent. Despite the measurement variability, the overall consistency between laboratories in classification of samples (using common pre-hoc derived cutoffs for AD) was high (>90% between-laboratory consistency in 15/18 samples for ELISA and in 12/18 samples for xMAP). The overall variability of CSF AD biomarker measurements remains too high (target variability may be below 10–15%) to allow assignment of universal cutoff values for a specific intended use, even for laboratories using the same commercially available assay. Thus, each laboratory must ensure longitudinal stability (lot consistency) in their measurements and use internally qualified cutoff levels. Kit lot-dependent effects have a significant influence on variability, especially for Aβ42. Further standardization of laboratory procedures and improvement of kit performance (including adoption of a universal reference standard) will likely increase the usefulness of CSF AD biomarkers for researchers and clinicians.