Analytical evaluation of the clinical chemistry analyzer Olympus AU 2700 plus

Background: The objective of this study was to perform the analytical evaluation of the clinical chemistry analyzer Olympus AU2700 plus. The evaluation was performed according to the guidelines of the European Committee for Clinical Laboratory Standards (ECCLS). Materials and methods: The evaluation consisted of determination of within-run and between-run imprecision, inaccuracy and comparison with Olympus AU2700. The tested analytes were: glucose, creatinine, urate, total bilirubin, cholesterol, tryglicerides, calcium, phosphate, iron, unsaturated iron binding capacity, aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, creatine kinase, alpha-amylase, alkaline phosphatase, potassium, sodium, chloride, C-reactive protein, antistreptolysin O and rheumatoid factor. Results: The results showed low within-run and between-run imprecision and acceptable inaccuracy for all analytes. High correlation with AU2700 analyzer was found. Intercept and slope (with 95% confi dence interval) met the preferences of Passing-Bablok regression for all of the analytes except alanine aminotransferase, sodium, rheumatoid factor, creatine kinase, total bilirubin and tryglicerides. Conclusions: Olympus AU2700 plus analyzer shows acceptable precision and accuracy for majority of analytes, with the exception of sodium and chloride. Instrument is fully comparable with Olympus AU2700 analyzer for all but several analytes (ALT, RF, sodium, CK, total bilirubin and tryglicerides), where only some minor deviations (constant and proportional diff erence) were observed.


Introduction
Olympus AU2700 plus (Olympus CO Ltd., Tokyo, Japan) is the new Olympus platform designed for clinical chemistry laboratories with large number of samples.Analytical evaluation of Olympus AU2700 plus analyzer was conducted upon the arrival of the analyzer at Clinical Department for Laboratory Diagnostic, Dubrava University Hospital, according to the protocol of the European Committee for Clinical Laboratory Standards (ECCLS) (1).
Analytical evaluation of the analyzer and methods should be done before the introduction of the new analyzer into the routine use in order to conrm declared speci cations of the analytical methods (2)(3)(4).
Olympus AU2700 is an automated, multi-channel, selective analyzer where the measurements are carried out using the spectrophotometry, turbidimetry and indirect potentiometry (using ion-selective electrode -ISE unit) in serum, urine, cerebrospinal uid, and other types of body uids.Analyzer can process 1600 tests per hour, including 270 ISE unit tests on the working temperature of 37 ± 0.3 °C.It is possible to program 99 di erent analyses at most.Analyzer can process 51 analyses simultaneously (48 di erent analysis and 3 electrolytes analysis on ISE unit).Olympus AU2700 plus has the same settings as the Olympus AU2700.Reagents are one-or two-component, lo-cated in two separate thermostated containers (4-12 °C).Analyzer uses the original reagents that are recognized using barcode readers or reagents from other manufacturers that are placed into speci c positions in the reagent containers.Type of reaction determines the number of reagents used for the speci c test.Samples are pipetted using a mechanical arm with two needles, which enables simultaneous pipetting in two parallel cuvettes.Analyzer has the possibility of continuous control monitoring, and statistical data processing obtained during operation (daily control and day-today control).
In this study we aimed to assess the analytical performance of 22 analytes determined on Olympus AU2700 plus analyzer (Beckman Coulter, Tokyo, Japan).For these analytes, a comparison with the Olympus AU2700 analyzer (Olympus CO Ltd.,Tokyo, Japan) was also done.

Analytical evaluation
Analytical evaluation of analyzer included the determination of within-run and between-run imprecision, inaccuracy (comparing to the declared values of the control sample) and methods comparison.Within-run and between-run imprecision were used to determine the extent of random error and accuracy (bias) was used to detect the extent of systematic error a ecting the measurement (5,6).
Between-run imprecision was determined measuring the concentration of analytes in the control sera of di erent concentration ranges (Level 1, Level 2, ITA 1, ITA 2) and pool serum in duplicate during the period of 30 days.Imprecision was expressed as the mean and the coe cient of variation (CV%).Within-run imprecision was determined in duplicate on 30 consecutive measurements of different analyte concentrations in control sera (Level 1, Level 2, ITA 1, ITA 2) and pool serum and also expressed as a coe cient of variation (CV%).
Inaccuracy of measurement of control samples was shown as bias (B%), percentage of deviation of the analyte mean value from the control sera declared mean value.To calculate the bias, measured values from day-to day imprecision were used.
Obtained values for precision and bias were assessed by comparing with the speci cations derived from biological variation (7).The desirable specications for imprecision and bias, derived from biological variation are presented in Table 2.
Comparison of results obtained on Olympus AU2700 plus and Olympus AU2700 analyzers was conducted on 50 sera samples with a wide range of values.The sera used for the comparison were collected at di erent hospital departments of Dubrava University Hospital due to the routine hospital check-up of the patients.The samples were randomly chosen for the comparison and were centrifuged immediately upon the arrival in Clinical Department for Laboratory Diagnostics, Dubrava University Hospital, at 1000 x g for 15 minutes in 35 R Rotina Hettich centrifuge (Tuttlingen, Germany).All the samples were collected in BD Vacutainer glass serum tubes.The commercial pool sera used were stored at -20°C until the analysis.

Statistical analysis
Statistical methods included the calculation of mean, coe cient of variation (CV%) and bias (B%).Pearson correlation coe cient and Passing-Bablok regression, including the Cusum test for linearity were calculated for the method comparison.The level of signi cance was set at P < 0.01.Statistical analysis, including descriptive statistics, was performed using MedCalc 9.2.0.0 software (MedCalc, Mariakerke, Belgium).CVw -within-subject biologic variation; CVg -between-subject biologic variation; I = desirable speci cation for imprecision; B = desirable speci cation for inaccuracy

Results
Results of day-to-day imprecision are shown in Table 2. Coe cients of variation for day-to-day imprecision ranged as follows: in Level 1 from 1.46 % to 3.90%; in Level 2 from 0.97 to 2.52%; in ITA 1 from 1.27% to 2.90%; in ITA 2 from 0.83% to 1.27% and in pool serum from 0.56% to 7.07%.
The results of within-run imprecision are shown in Table 2. Coe cients of variation for within-run imprecision ranged as follows: in Level 1 from 1.15% to 2.51%; in Level 2 from 1.05% to 3.03%; in ITA 1 from 1.60% to 3.15%; in ITA 2 from 1.23% to 2.26% and for pool serum from 0.94 to 3.74%.
Test results for inaccuracy of control sera in two concentration levels are also shown in Table 2.For Level 1, bias ranged from 0.23% to 3.45%; for Level 2 from 0.01% to 1.86%; for ITA 1 0.35 to 0.69% and for ITA 2 from 1.43 to 1.86%.
Comparison of the results obtained on the Olympus AU2700 plus and Olympus AU2700 Olympus yielded the correlation coe cients between 0.961 and 1.000 (Table 3).The target values used for the calculation of inaccuracy were provided by the manufacturer of the commercial control sera.The slope and y-axis intercept as well as their respective 95% con dence intervals (8) obtained using Passing-Bablok regression for all analytes are presented in Table 3.
The Cusum linearity test performed on Passing-Bablok regression showed that there was no signi cant deviation from linearity for any analyte (P > 0.10) (9).

Discussion
The results of this analytical validation showed acceptable coe cients of variation for day-to-day imprecision, within-run imprecision, as well as a satisfactory degree of accuracy.For setting the criteria for acceptable imprecision and inaccuracy, we considered the biological variations of analytes (7,10,11).Depending on the measurement procedure, measuring instruments and compliance with a reference or de nitive method, we can assess whether the new method or analyzer is suitable for routine use, and whether they are of satisfactory accuracy (2,3,5,12).For most analytes tested in this study, the results of imprecision and bias were in the range of desirable speci cations for imprecision and bias, derived from biological variation (10,11).Bias and imprecision for sodium and chloride did not ful l required criteria.However, it is important to emphasize that these criteria are very strict: bias of 0.4% and 0.6% and imprecision of 0.3% and 0.5%, for sodium and chloride, respectively.The reason for not ful lling the desirable criteria for sodium and chloride is the low biological variation, as well as the lack of analytical method capable to follow that variation with the results that are precise and accurate enough.However, laboratories must be aware of this limitation, and ensure smaller measurement uncertainty using quality control tools (participation in external quality control assessment, control of preanalytical and postanalytical phase) (13,14).
Correlation analysis yielded high correlation coefcients proving high correlation for all the tested parameters.
Passing-Bablok regression showed that most analytes did not di er by a constant (95% CI of intercept includes 0), and that were well aligned, suggesting that there were no proportional di erences between methods (95% CI of slope includes 1) (9).Despite of high correlations, some analytes did not meet the preferences of Passing-Bablok regression.When following Passing-Bablok regression strict criteria, ALT, RF, and sodium di ered by a constant amount for both analyzers and the proportional di erence between methods was found for CK, total bilirubin and triglycerides (Table 3).Passing-Bablok regression yielded a wide range of 95% CI of intercept for sodium, and that should also be considered when de ning the analytical objectives.Unlike sodium, ALT and RF are a bit beyond the requested criteria (95% CI of intercept for ALT was 0.290-1.000and 0.677-1.377for RF), so in the analytical sense, those analytes follow the same linearity.Similar is the case with the CK, total bilirubin and triglycerides whose 95% CI of slope was very close to 1 (1.014-1.022,1.007-1.046and 0.944-0.985respectively), therefore the analysis can be considered acceptably aligned.

Conclusion
Olympus AU2700 plus analytical analyzer shows acceptable precision and accuracy for majority of analytes, with the exception of sodium and chloride.Instrument is fully comparable with Olympus AU2700 analytical analyzer for all but several analytes (ALT, RF, sodium, CK, total bilirubin and tryglicerides) where only some minor deviations (constant and proportional di erence) were observed.

TABLE 1 .
Methods and reagents used in the evaluation of Olympus AU2700 plus.

TABLE 2 .
Desirable speci cations for imprecision and bias derived from intraand inter-individual biologic variation for the tested analytes, day-to-day, within-run imprecision and inaccuracy.

TABLE 3 .
Correlation and Passing-Bablok regression data for Olympus AU2700 plus and Olympus AU2700 comparison.