Analytical performance of the selective multianalyser Olympus AU 5200

The analytical performance of a selective, automatic multianalyser- the Olympus A U5200 - was tested and assessed for practicability, following ECCLSguidelines. Twenty-two analytes were tested and compared with the Olympus A U5000 analyser. A Hitachi 747 analyser was also included in this survey in order to obtain correlation data for ISE measurements. The imprecision data, expressed as median CV values, were found to be below 2% in series for 21 parameters, and below 3% for 19 paramaters from day to day. Creatinine measured with the kinetic Jaffe method obtained a median CV value of 4% in series, creatine phosphokinase showed the worst imprecision from day to day with a CV of 9%. Slightly better precision values for the majority ofall tests were found on the Olympus AU5200 than on the AU 5000 analyser. The recovery of the assigned values in 32 commercial control sera was between 95% and 105% for 14 tests. Five of the remaining tests yielded recoveries with deviations between 5% and 10%, deviations above 10% showed albumine, alanine aminotransferase, aspartate aminotransferase and creatine phosphokinase. The accuracy of most test parameters was slightly better on the AU5200 analyser than on the comparison instrument. The range of linearity ofthe tested methods covered the range stated by the manufacturers; and no sample carry-over was detected. Most parameters tested yielded close correlation to those on the comparison instrument. Amylase measurements on both analysers correlate well but are not comparable without data correction due to different test methods. In addition, no drift effects were observed over a period of 9 hours. The ion-selective-electrode unit performed well in terms of throughput, precision and stability over time. The whole system showed good practicability with respect to patient sample and reagent handling, a short training period of technicians, ease of system software, maintenence, a robust barcode reader and a flexible host communication procedure.


Introduction
The Olympus AU 5200 series of selective multianalysers is well suited for medium and large laboratories processing several thousand samples per day. This paper reports on the performance characteristics of the AU 5200 analyser comparing it with the AU 5000 analyser. The evaluation procedure followed ECCLS guidelines [1]. Although only a multicentre evaluation allows a truly representative assessment for an analyser, the evaluation data presented here give a first impression of the quality and performance of the instrument.

The AU 5200
The AU 5200 is constructed following a modular concept ttiat originated from the AU 5000 instrument series. Any analyser consists of up to four basic identical units equipped with eight or 16 reagent lines. The stated sample throughput of 330 samples (300 samples with ISE unit), which is twice as much as that of an AU 5000 analyser, is achieved by two cuvette rings per unit with 240 cuvettes each. In an analyser equipped with eight reagent lines per module, 60 cuvettes are assigned to each chemistry channel (this compares with 24 cuvettes in a similar model of the series AU5000). A list of the main specifications of the AU 5200 analyser is given in table 1, along with a survey ofpossible configurations. Evaluation data on the comparison instrument AU 5000 has been published by Luley et al. [2].
The analyser configuration which was evaluated for this paper consisted of three units with eight reagent lines per unit for chemistry tests, and an additional ISE unit for electrolyte measurements.

Quality-control materials
The commercially available control materials used in the evaluation are specified in table 2.

Calibrators and standards
Multianalyte calibrators obtained from Olympus Optical Co. and Merck Company were used to calibrate the respective Substrate methods of both reagent suppliers.
Aqueous electrolyte standards obtained from Olympus Optical Co. and Boehringer Mannheim GmbH were used to calibrate flame photometry and ISE measurement on the Olympus AU 5000, AU 5200 and Hitachi 747 analysers respectively. To measure linearity of chemistry procedures analyte-spiked serum materials were obtained from Sigma Chemical Company (Multi-Analyte Lintrol, Multi-Enzyme Lintrol, CK-Lintrol, Lipid-Lintrol) or produced by adding pure analyte to a serum pool (iron and bilirubin standard solutions from Sigma Chemicals).

Chemistry reagents
The reagent kits were supplied by Olympus Optical Co.
in Hamburg and Merck Company in Darmstadt, Germany. A list of reagents, reagent manufacturers and methods of chemistry testing is given in table 3. Sigma Chemic GmbH, Germany BioChemica in Flacht, Germany) on the AU 5000 analyser and by indirect potentiometry on the AU 5200 and Hitachi 747 analyser.

Precision study
To determine the within-run imprecision, seven different control materials and two serum pools were analysed in a series of 20 assays. The series were repeated three times and the medians of the three CV values were taken as the final results.
The imprecision between-days was determined on 15 different, days using the seven control materials mentioned above.

Assessment of accuracy
Thirty-two commercial control materials (listed in table 2) were analysed in double assays on three days in both the AU 5200 and AU 5000. The medians of the measured values were compared with the assigned values of the different control materials (if available for the methodology used).
Linearity study The linearity of each chemistry method was assessed using serum material with high analyte concentrations and 25 equidistant levels of dilution with physiological saline solution or bovine serum albumin (for cholesterol, triglyceride and bilirubin dilutions). The dilutions were prepared using a robotic dilutor system (Tecan RSP 505, Zinsser Analytik). The measured concentrations were plotted against dilution levels and the resulting graphs were inspected visually for linearity.
Carry-over experiments Due to the construction principle that each cuvette and each mixer blade is used for one kind of test only and that the reagent lines are totally separated from one another, carry-over is restricted to sample carry-over caused by insufficient washing of subsequent cuvettes, the probe needle or the mixer blade.

Correlation study
Approximately 250 fresh human sera from daily routine samples were analysed simultaneously in the AU 5200 and in the AU 5000. In addition, sodium, potassium and chloride concentrations in the same samples were measured in a Hitachi 747 analyser equipped with a ISE unit. Care was taken to include pathological analyte levels.

Assessment of drift effects
Drift effects were studied using seven commercial control sera. The sera were dissolved before starting the drift experiment and kept at 4C until use. For determination of alkaline phosphatase, creatine phosphokinase and bilirubin the sera were freshly reconstituted every hour.
Five measurements of each control serum were carried out every 60 min for 9 h. Prior to starting a series of measurements the reagent lines were flushed with fresh reagent.
In order to detect short-term variations or sudden changes, a second experiment was conducted. Chemistry  As indicated in figure (a), 21 of 23 evaluated tests on the AU 5200 analyser yielded a median within-run imprecision well below 2"0%; urea showed a median imprecison of 2" 1%. Creatinine, measured with the kinetic Jaffe reaction method on the AU 5200, yielded rather high CV values from 1"5% to 4"7% (not shown in figures (a) and l(b).
The electrolyte measurements on the AU 5200 analyser showed a remarkably good precision. Slightly lower median CV values for the imprecision in series were found for 18 of 22 tests on the AU 5200, compared with the AU 5000 analyser.
The median between-day CV values were below 3"0% for 19 out of 23 tests on the AU 5200 analyser and were well below 5"0% for all tests, with the exception of the CPK activity measurement which showed CV values from 6"1% to 15-7%.
The majority of all tests gave lower imprecision from day to day on the AU 5200 analyser compared to those on the AU 5000.
The overall precision in series and from day to day on both analysers could be rated as very good, with the AU 5200 performing slightly better than the AU 5000 analyser.  Figure 3 (a). Drift during 9 h for three representative analytes measured in seven control sera. The medians, lOth, 25th, 75th and 90th percentiles of the recovery ratios in relation to the initial measurements are shown as box-plots.

Assessment of accuracy
The recovered values of all analytes in 32 commercial control sera were expressed as percentages of the assigned values. The medians, the 10th, 25th, 75th and 90th percentiles of the respective values are shown in table 5 and in figure 2. Fourteen of 23 tests yielded median recoveries within the 5% range; five tests were in the range between 5% and 10% (lactate dehydrogenase, bilirubin, cholesterol, triglyzerides and creatinine measured enzymatically); and three tests were in the range between 10% and 15% (alanine aminotransferase, creatine phosphokinase).
The low recoveries of some enzyme activity tests may be due to the conversion of the values measured to corresponding values at 25C. These conversion factors are based upon human serum, whereas most commercial control materials are essentially non-human or spiked with non-human analytes. The low recoveries of albumin (83%), cholesterol and triglyzerides were caused by calibration errors. As a consequence of this evaluation the calibrator values for these tests were reassigned by the manufacturer.

Linearity
Both the linearities found, and the linearity ranges given by the manufacturers, are shown in table 6. The Carry-over A potential carry-over ofsample residue resulting from an insufficient washing of the cuvettes was investigated by measuring creatine phospokinase activity (27 U/1 in a serum pool) in cuvettes which had been used for the determination of the same analyte with a very high activity (approximately 1500 U/l). The results were compared with measurements of low activity in cuvettes which had also been used for measurements of low activity. Figure 4(a) shows that there was no detectable difference. Sample carry-over due to incomplete washing of the probe needle and the mixer blades was investigated by alternating two serum pools with high and low creatine phosphokinase activity; the samples which immediately followed the groups with high activity did not show elevated activities ( figure 4(b)). In conclusion, no specimen-related carry-over could be detected.
Method comparison with patient samples The Olympus AU 5200 series analyser was compared with an Olympus AU 5000 analyser and a Hitachi 747 analyser (ISE measurements only). The statistical evaluation was performed according to the method of Passing and Bablock [3]. The numerical results are given in

Assessment of practicability
An instruction period of four days was necessary to acquire the knowledge to operate the Olympus AU 5200 analyser, users of the older AU 5000 analyser can operate the new system instantly. Despite the size and capacity of the analyser, the Olympus AU 5200 is easy to run and can easily be integrated into the organization of a large laboratory. The sample feeding mechanism operates in a linear fashion, allowing continuous supply of subsequent samples. One sample which is nowadays not an aliquot, but delivered in a primary tube, passes through the instrument in 5 min and even tubes on a fully loaded sample tray are accessible after 70 min. The sample belt is in front of the instrument, which is a major advantage over the older AU 5000 series analyser where it is in the rear. The analyser can be reached from all sites and all mechanical movements can be watched.
The system-processing software for routine operation is clearly arranged; the main jobs, for example activating the analyser out of a standby status, starting measurements, washing the cuvettes and flushing the reagent lines, are presented in one menue and are initiated by one keystroke. The software contains an extensive diagnostic program, which allows a specific detection of mechanical or electronic problems.
The preanalytical preparation of the analyser consists of four steps requiring the time periods given in parenthesis: warming up to 37C (60 min), washing of cuvettes and filling reagent lines (23 min), reagent blanking, calibration and running controls (20 min). Using the automatic power-on facility reduces the preanalytic phase by the time period for warming up the instrument. Thus, after 45 min, including the time needed for checking the results of calibration and control materials, the analyser is ready for routine operation.
During the evaluation period and the subsequent routine work no total shut down of the analyser occurred. The most frequent problems were rack jams on the transport belt and needle crashes. In such situations the analyser does not stop completely, but stops only the feeding mechanism. Results of tests pipetted so far can still be obtained. The most severe faults were worn out gears of the cuvette wheel motor causing a small displacement of the cuvette wheel hindering the action of the washer unit; and, a leakage in the vacuum system of the washer unit causing an incomplete draining of the cuvettes. Both problems could be attributed to some kind of material fatigue. The manufacturer has now replaced the materials for these machine parts. Another point of criticism is the badly arranged report format in which the results are printed out-there is no way ofachieving patient-oriented reports.

Conclusion
With respect to imprecision, accuracy, linearity and carryover data, the Olympus AU 5200 analyser compared favourably with and showed slightly better results than the AU 5000 analyser. The Olympus AU 5200 analyser reaches a sustained peak performance of 300 samples/h including ISE measurements, operates reliably from day to day and is thus well suited for medium and large sized laboratories.