Emeric Chassaing1, Philippe Vergnaud1, Erik Stoops2, Hugo Vanderstichele2 and Tanja Schubert1
(1) BioClinica LAB, Lyon, France (2) ADx NeuroSciences, Gent, Belgium
The pharmaceutical industry is searching for biochemical markers that can predict a cognitive decline in pre-symptomatic stages of Alzheimer's disease (AD). Promising biomarkers are cerebrospinal fluid (CSF) Beta-Amyloid 1-40 (Aß1-40), Beta-Amyloid 1-42 (Aß1-42) and Total Tau.
Although CSF biomarkers may predict the stage of the disease, some performance issues of the first generation of assays hampered their widespread use. As an independent, College of American Pathologists (CAP) accredited laboratory, we investigated the analytical performance of the aforementioned second generation assays from EUROIMMUN (CE registered) and developed by ADx Neurosciences.
Apply our efficient quality-driven validation procedure to evaluate the performances of a commercial in-vitro diagnostic manual immuno-assays suitable for patient sample evaluation in clinical trials
VALIDATION PROCEDURE & SAMPLES SELECTION
BioClinica LAB analytical validation method is designed to ensure the accuracy and reliability of the results. The validation process, the criteria applied, and the parameters chosen are adapted based on the recommendations from different guidelines, clients' feedback and their internal requirements, and field experience.
The BioClinica LAB team assesses several parameters which give information regarding calibration, accuracy and precision, spiking recovery, linearity of the dilution, working range, and biomarker stability.
The samples used for the validation are cerebrospinal fluids (CSF) provided by 4 different centers. Two centers treated the samples with additives, while the two others centers provided CSF from healthy patients or patients diagnosed with Alzheimer's disease. Due to the limited number of samples used for the validation (n=30), no reference ranges have been established for the different populations and sampling conditions. All CSF samples are stored at -70°C and thawed at +4°C before use. One determination of one sample is always performed in double determination (duplicate). When additional tubes are required, pre-rinsed LoBind tubes from Eppendorf are used.
Note: although they share the same name, please note that for each test, each markers has its own set of samples.
Visual comparison of the calibration curve with (blue) and without (red) intermediate points as initially indicated by the manufacturer. BioClinica LAB added intermediate points in order to have a minimum of six, validated, duplicate non-zero calibrator concentrations covering the entire range as indicated by the FDA guidance (2013). The variation on the results (control and samples) between the two calibration curves is below 6% for all three markers.
Five Quality Control (QC) samples were determined for each assay according to their concentration. QC 1-2-3-4-5 should be respectively near the LLOQ, 2 to 3 times the LLOQ, mid-range, high range and very high range, respectively.
Five QC samples have been performed 6 to 12 times within one run. All of the coefficients of variation (CV%) are below 10%.
The five same QC samples from the repeatability study are performed in at least 4 different assays. Almost all of the samples have a CV% below 15% except for the QC1 in Total Tau (19.4%) which can be explained by a concentration approaching the lower limit of quantitation (QC1: 114pg/mL ; LLOQ manufacturer: 85 pg/mL)
Three different samples with mid-high to high values on the three biomarkers have been diluted subsequently until 1:8 in the sample buffer provided in the kit. The recovery percentages are then calculated as followed:
(measured value X dilution factor) / expected value X 100.
All the recovery percentages are within the acceptable range for Aβ1-40 & Aβ1-42, indicating that CSF samples can be diluted until at least 1:8 from the neat (or after predilution for Aβ1-40). Total Tau can be diluted until 1:4 in sample buffer.
Two samples for Aβ and three samples for Total Tau with different concentrations have been spiked by three different concentrations of calibrators in a 75:25 ratio (sample : calibrator). The expected value is calculated from the sample value and the calibrator quantified separately. Then the assay value of the mix is compared to the expected value.
The recovery percentages are within the acceptable range for most of the samples for the three markers. With regards to consistency, recovery between 70% and 80% for Aβ1-40 and Total Tau are considered acceptable.
QC1 with two very low samples for Aβ1-42 and Total Tau and three very low samples for Aβ1-40 have been run at least 10 times within the same run. The mean concentration of the last sample with a CV% between 15 and 20% should be considered as the lower limit of quantitation. The LLOQ should also be higher than the first non-zero calibrator (Bioanalytical Method Validation FDA guidance, 2013). For Aβ1-40, Aβ 1-42, the LLOQ determination criteria did not fit the data obtained; the value of the first calibrator has thus been chosen as the LLOQ (Aβ1-40: 21.08 pg/mL or 442.7 corrected by predilution factor; Aβ1-42: 27.08 pg/mL). For Total Tau, the low sample 01 has a CV% at 15.5% and a concentration above the first non-zero calibrator (70pg/mL). The LLOQ determined for Tau is 89.03 pg/mL.
For each assay the sample buffer was run at least 16 times within the same run. The Limit of detection is calculated by adding two or three times the standard deviation (SD) to the mean concentration or the optical density when the concentration is undetectable. In this case, the LOD is back-calculated with the formula of the curve.
Note on CSF stability: Despite the relative good results on stability testing, the manufacturer recommends to avoid more than two freeze/thaw cycles and to store the CSF samples at -70°C after use. Long term storage stability is tested at this moment and will give indication about the stability of the CSF samples when stored several months at -70°C .
The three assays (Aß1-40, Aß1-42, Total Tau) meet the internal acceptance criteria in our laboratory for clinical trial sample testing. A centralized, batch-mode measurement should be favored. In our laboratory, the assays are fully compliant with the fit-for-purpose application in clinical trials.