Logistics & Quality Control
Over 20 years of experience in handling thousands of samples on a monthly basis has materialized into optimized procedures for sample handling and software for data integration and quality control. Our procedures and logistics are set up to avoid human errors, minimize preanalytical variation and assay drift, and maintain adequate assay performance in terms of accuracy, precision and low sample consumption.
Sample handling and data integration
The raw data from the separate platforms and analysis sets are integrated, handled and controlled by a specially designed database. The system generates printouts that match the trays installed into the robots carrying out sample handling (Figure 1). This procedure secures correct location of the separate samples in the assay tray. The data files from the separate platforms are merged by name and stored by the system, which calculates spread between parallel runs, and calculates summary statistics. Outliers are flagged by macros, and all import and export functions are automated by scripting.
Validation across platforms
Some stable metabolites are measured at two or three platforms (methods), but against the same calibrators. These metabolites are total homocysteine (tHcy), total cysteine (tCys), cystathionine (Cysta), methionine (Met), tryptophan (Trp), kynurenine (Kyn), histidine (His), ornithine (Orn) and trimethylamine N-oxide (TMAO) (Figure 2). For these analytes, the above-mentioned software calculates the ratios between concentrations obtained by different platforms. Analysis on each platform requires separate pipetting of sample and reagents and separate organization of vials into the sample tray. Therefore, analytical errors related to sample identification, handling, pipetting and organization as well as instrumental performance are likely to be discovered as metabolite ratio(s) different from 1. This ensures adequate sample handling and logistics but also minimizes the possibility of assay interference.
Each set of 96 vials contains 6 vials with calibrators, 3 with control plasma samples with known biomarker concentrations and one vial without biomarker (blank, to control for carry over). The calibrators are diagonally located (from upper left to lover right corner) across the sample tray to verify positioning of the tray in the autosampler. Large stock solutions of plasma calibrator and control plasma are prepared in sufficient amount to last for years, to minimize chance of assay drift over time. These stocks are aliquoted and stored at -80 °C. New stock are calibrated by comparison with the previous validated stock solution by analysing about 1000 parallel samples over one month. BEVITAL participates in external quality control programs for total homocysteine, MMA, cystatin C (DEKS), vitamin D (DEQAS), 24 amino acids, 3-hydroxybutyrate, creatinine (ERNDIM), vitamin K (KEQAS), 5-methyltetrahydrofolate, cobalamin and hsCRP (LABQUALITY).
Minimizing sample consumption and analyte degradation
Logistics have been established to minimize consumption of samples to be analyzed on more than one platform and to avoid preanalytical analyte degradation during and after sample thawing.
Long-term assay drift (LAD)
LAD refers to systematic rather than random changes in measured analyte concentrations that may take place over months or years. This may result in slight differences in central estimates and/or distribution when comparing values from repeated analyses carried out years apart. LAD most likely occurs when first measurements are carried out shortly after the method has been launched (new method), is detected in large sample series, and is occasionally observed for biobank samples in spite of no detectable changes in levels for longitudinal control samples. Possible explanations include column replacement, altered instrument performance, and method optimization. To avoid bias from LAD, BEVITAL measures sample series for each project within a short time period, when no deliberate changes or optimization in method design are undertaken. BEVITAL strongly discourages composing data sets where samples from comparison groups are analyzed at different time points. Possible bias from LAD may be reduced but not totally avoided by value correction based on overlapping samples or control plasmas.