AD is pathologically characterized by the accumulation of amyloid-β (Aβ) in extracellular plaques and hyperphosphorylated tau (p-tau) in intracellular neurofibrillary tangles. These pathological changes can be detected using reference standard methods, namely positron emission tomography (PET) imaging of amyloid and tau pathology or measurement of Aβ and tau proteins in cerebrospinal fluid (CSF)¹. In recent years, blood-based biomarkers for amyloid accumulation (Aβ40 and Aβ42), tau pathology (different Tau phosphorylated forms, such as p-tau181, p-tau217 and p-tau231), neurodegeneration (Neurofilament Light Chain, NfL) and astroglial reactivity (glial fibrillary acidic protein, GFAP) have become a less invasive and cost-effective alternative to reference standard methods. Their levels vary along the Alzheimer’s continuum, making them valuable for AD early diagnosis^2–4. These blood-based biomarkers are widely used in research settings and are being adopted in clinical practice as supportive tools for AD diagnosis^2,5–7. Recent developments of new therapies have prompted their implementation in clinical trials to assess participant eligibility and to monitor treatment effectiveness.

In preclinical AD, biomarker changes over time are subtle compared to symptomatic stages. As a result, preanalytical and other confounding factors can influence measured levels⁸, potentially masking true disease-related effects and hampering the findings reproducibility. This underscores the need to evaluate factors during sample collection, processing, and storage that may affect biomarker measurements. Empirical results obtained from different laboratories contribute to consensus guidelines and standard operating procedures for blood sample handling^7,9. Still, data on the impact of some factors, such as fasting conditions before sample collection or long-term storage at -80ºC, are scarce and should be analysed from different perspectives due to their complexity. This study aims to assess the impact of fasting conditions and long-term storage at -80ºC on plasma biomarkers for AD pathology (Aβ40, Aβ42, Aβ42/Aβ40 ratio, p-tau181, p-tau231), glial reactivity (GFAP) and neurodegeneration (NfL) using samples from the ALFA cohort^10,11.

Aliquots were thawed at RT for 1 hour, vortexed and centrifuged for 10 minutes at 4000g at RT. To reduce experimental variability, all samples were randomized before the measurements and analysed at the same time.

Biomarker concentrations were quantified using the SIMOA HD-X Analyzer at the Fluid Biomarker Facility of BBRC. Aβ40 and Aβ42, GFAP and NfL were analysed with the Neurology 4-PLEX E (N4PE) Advantage (103670, Quanterix) and p-tau181 with the p-tau-181 Advantage v2.1 (104112, Quanterix) commercial kits according to the manufacturer’s instructions. Plasma pTa231 was analysed using a validated homebrew assay througly described in previous studies^4,12. Quality controls (QCs) were included in every analysis to ensure the reliability of the measurements. The coefficient of variation (CV) intraplate and interplate of the QCs analysed along the experiments are below 10% and 15%, respectively, which are within the accepted range.

To assess the impact of fasting on plasma biomarker levels, paired samples from 16 individuals were collected under two conditions: fasting, after a minimum fasting period of 8 hours, and non-fasting, when no medical indication for fasting was present (mean age 64 years old; 69% female, 50% amyloid positive; Table 1). Fasting was confirmed by serum glucose levels (< 115 mg/dl). The average time between paired samples collection was 54 days.

For the storage time analysis, samples from different individuals collected along the 10 years were chosen. The final number of samples selected was 623, from which 211 samples were stored for up to 10 years, and the rest for 2-6.5 years (Table 2). Samples at different time points were balanced in terms of age, sex and APOE-ε4 carriership using MatchIt package (4.7.2 version). All biomarkers were measured in samples subjected to less than 3 freeze-thaw cycles (FTC). For the storage time, 91% of the samples in p-tau181 and N4PE had 2 FTC and 9% had 3 FTC; for p-tau231, 99% had 2 and 1% had 3 FTC, respectively.

Storage time influenced plasma concentrations on Aβ peptides, with increased levels in samples stored for longer time (Aβ40: β = 0.0021, p-value = 0.0155; Aβ42: β = 0.1357, p-value = < 0.001). This effect was not observed with the ratio Aβ42/Aβ40 (β=-4.594-05, p-value = 0.7608, Fig. 2 and Suppl. Table 1). According to our regression model, the variability explained by the storage time (partial.R2) was 11.17% and 7.71% for Aβ40 and Aβ42, respectively, and only 0.02% for the ratio (Suppl. Table 1). Plasma p-tau181 and p-tau231 concentrations fluctuated over time, with significant lower levels in samples stored for 6 years (p-tau181: β = 0.0015, p-value = < 0.001; p-tau231: β = 0.0063, p-value = < 0.001). Plasma GFAP levels decreased in samples stored more than 6 years (β = 0.0088, p-value = < 0.001). However, the variability of the biomarker’s concentration explained by storage time was below 6% for all these biomarkers. We did not observe changes in NfL over time (β = 0.0010, p-value = 0.2199). Considering that some pre-analytical factors, such as the centrifugation settings and storage conditions differed in those samples stored for up to 10 years due to historical protocols, we performed a sensitivity analysis including only those following the same procedure (i.e., samples stored for up to 6.5 years; Suppl. Table 2 and Suppl. Figure 2). Among all the biomarkers, only p-tau231 (β=-0.0277, p-value = 0.0004; Supp. Table 3) and GFAP (β=-0.0499, p-value = < 0.001) showed decreased levels in samples stored for longer periods, but similarly, the variability on biomarker concentrations explained by the storage time was below 5%.

Understanding the impact of pre-analytical factors on fluid biomarker measurements is essential for their optimal implementation in clinical practice and trials^13–15. Here, we evaluated the impact of fasting conditions and long-term storage at -80ºC on plasma concentration of Aβ40, Aβ42, Aβ42/Aβ40, p-tau181, p-tau231, GFAP and NfL.

Given the variation in haematological analytes after a food intake^16,17, it is crucial to evaluate the impact of fasting conditions on plasma biomarker concentration. To this end, we analysed paired plasma samples from participants that were asked or not to fast before venipuncture. Blood samples under both conditions were collected within a short period of time (mean: 54 days) and the duration of blood processing for plasma separation was consistent across samples (less than one hour). Samples were analysed at the same time, reducing experimental variability. Our findings showed that plasma Aβ peptide concentrations were lower under fasting conditions which, in line with other studies, was mitigated by using Aβ42/Aβ40¹⁸. The use of the ratio has been previously recommended^17,18 as it minimizes the surface tube adsorption effect on Aβ42 levels and the interindividual variability^17,18. The other biomarkers were not by fasting status, neither at their absolute levels, nor on their performance to detect amyloid positivity, suggesting that fasting does not have a significant effect on these biomarker levels. Previous studies reported no changes on plasma Aβ40 and Aβ42 levels after and before food intake¹⁹, but a recent study demonstrated significant changes in all the biomarkers analysed (Aβ40, Aβ42, p-tau181, p-tau231, GFAP and NfL) in postprandial conditions²⁰. Of note, this study was conducted in a cohort of obese individuals following a specific diet, and thus under specific pathophysiological conditions. Our study analysed the impact of non-fasting conditions mimicking the routine practice, where there is no control of the patient’s diet and last food intake. We compared paired samples from the same individuals under different fasting conditions. The inclusion of different participants in each group (fasting/non-fasting) in previous studies^18,21 may also explain the discrepancies observed. Thus, our results suggest that fasting status is not relevant for the biomarkers’ performance, although some comorbidities such as obesity might be considered for biomarkers’ interpretation.

Another relevant factor scarcely investigated is the impact of long-term storage at -80ºC on plasma biomarker levels^22,23. The analysis of biomarkers in samples stored for long periods is essential for increasing sample size and statistical power; and in longitudinal studies to identify prognostic and monitoring biomarkers²⁴. Few studies have been performed assessing the effect of storage time on plasma biomarkers concentration. There are several approaches described in literature, some of which measure samples from the same individuals at different time points^22–24. A study on CSF made predictions based on the Arrhenius Eq. 2⁵; another one performed repeated measurements of the same sample for 26 months²⁶. However, these approximations cannot differentiate the effect caused by the storage time from the age effect. We took another approach by analysing the distribution of samples from different individuals stored for up to 10 years within the same immunoassay, to discard any potential batch effect that may impact biomarker measurements²⁷. Individuals were matched for factors known to influence biomarker levels such as sex, age and APOE-ε4 genotype^28–31. We observed that samples stored for longer periods presented higher levels of Aβ40 and Aβ42 with a variability of 11.17 and 7.7% respectively, which was mitigated by using the amyloid ratio. There were also fluctuations over time on plasma p-tau181 and p-tau231 concentrations, with lower levels in samples stored for 6 years. However, such variability was below 6%, which is within the coefficient of variation of the immunoassay and may thus not have a major impact on biomarker interpretation. Similar effects were seen for plasma GFAP, with decreased levels in samples stored for 6 years or more. Plasma NfL was the only biomarker stable along time. Previous studies have shown long-term stability of plasma Aβ40, Aβ42, GFAP and NfL within periods ranging from 5 to 20 years^22–24. However, in line with our findings, the stability of p-tau181 appeared to be more variable and sensitive to storage conditions in CU individuals²⁴.To further investigate these discrepancies, we performed a sensitivity analysis using the samples stored for up to 6.5 years, as those had gone through the same collection procedures (centrifugation and storage). Plasma Aβ peptides showed stability for up to 6 years (Suppl. Figure 2); p-tau231 and GFAP still had lower levels in samples stored longer, but the variability observed was under 5%. The impact of such variability on biomarker interpretation depends on the analytical robustness of the assay, which is high for Simoa assays of plasma GFAP and p-tau181 (similar performance with simulated variations > 20%), but low for plasma Aβs³². Overall, our results suggest that whether significant changes on some plasma biomarkers over time were detected, these will likely have limited effect for cross-sectional studies using samples stored for long periods of time. For longitudinal studies, the variability detected should be considered in the interpretation of the results. For amyloid peptides, the use of the ratio rather than the single peptides would be recommended.

Despite this study characterized two preanalytical factors relevant for the implementation of AD plasma biomarkers, there are several limitations to consider: i) participants were not asked to be in a non-fasting condition, and thus we cannot guarantee a postprandial state. Still, this simulates the clinical diagnosis scenario in which the non-fasting status is not specifically required for AD plasma biomarker analysis; ii) samples had either undergone two or three FTC; however, previous investigations demonstrated Aβ peptides to be stable up to 3 FTC^9,13, and the other biomarkers up to 4 FTC^9,33,34 and, therefore, it is not expected to influence our results; and iii) protocols for plasma collection and storage, including specific material, have historically changed across centers. In our case, centrifugation settings and storage conditions of samples stored longer were different. Despite a prior study suggests that the centrifugation settings (speed, temperature) do not affect the stability of biomarkers⁹, smaller surface/volume ratios could lead to smaller concentrations of AD biomarkers^35–37. However, differences on the surface/volume ratio have only been studied in CSF, with an effect on Aβ levels^35,36, whereas p-tau181 remained stable³⁵.

In summary, our findings suggest that fasting status does not impact biomarker measurements, although comorbid conditions, including obesity, might be considered in the interpretation of plasma biomarker levels. Even storage at -80ºC for up to 10 years may influence the Aβ peptides, p-tau181, p-tau231 and GFAP levels, its effect was below 6%. Plasma amyloid peptides showed the strongest effect, but this was overcome using the amyloid ratio. Since some biomarkers levels changed across storage time, this factor should be considered in the analysis and interpretation of the results, especially in longitudinal studies.