Assessment of recently developed blood gas analysers: a multicentre evaluation

Providing guidelines for testing expected inaccuracy and imprecision is still a matter under debate. The Expert Panel of the French Society of Clinical Chemistry has developed a protocol, which was based on a comparative multi-centre evaluation of four instruments: the Ciba-Corning 278, the Instrumentation Laboratory 1306, the Nova SP 5 and the ABL 330. The purpose was to evaluate the analytical performance and efficiency of the analysers. Another aim was to design a valid approach for evaluating any new system. As buffered aqueous solutions and fluorocarbon emulsions give only partial information, tonometered blood was used at different levels of gas mixture, even though it is both difficult and time-consuming. Comparisons have been established on patients' blood samples with the analysers currently used in the evaluation sites. The tests showed that the four analysers have the same degree of precision, and interinstrument comparisons demonstrated a very high degree of reliability. This analysis emphasizes that the evaluation of instruments for pH and blood gas analysis is neither easy nor is it often done, mainly due to the choice of a quality-control material and the lability of the measured parameters.


Introduction
Modern pH and blood gas analysers measure pH, pCO2 and pO2 by electrochemical methods. Built-in microprocessors have not only improved automation and signal processing but also the calculation of parameters for oxygen status and acid base balance [1]. A multi-centre evaluation was organized by the French Society of Clinical Chemistry in order to compare the analytical performance and the practicability of four recent models so that guidelines could be given on the choice of equipment. The study also looked at quality control in bood gas analysis. Site  Tonometr,y Tonometry was performed according to the recommendations of the proposed IFCC reference method for tonometry of blood [3] on. the Laue bulb tonometer (L. Eschweiller & Co., Kiel, FR Germany) and the IL 237 tonometer (Instrumentation Laboratory, Inc. Lexington, Massachusetts, USA) with a gas mixture of known composition (pO and pCO). Blood  Measurements should be ideally performed on whole blood but stabilization ofthis medium is critical [4]. Thus the two phosphate buffers were run in triplicate over five days on both comparison and tested instruments.
Precision study Within-run precision was estimated with aqueous solutions or fluorocarbon emulsions and with two levels of tonometry El:CO2 5%, O2 12%; L2:CO2 12%, O2 5"5%). For each sequence, six measurements were performed daily on the three first and last days of the evaluation.
Drift was checked using tonometered blood tested immediately after a calibration and before the subsequent one.
Linearity was assessed using three successive measurements of tonometered whole blood containing O2 from 0 to 95% and CO2 from 2 to 15%. The sequence was repeated three times during the evaluation period.
Neither drift nor carry-over were observed in any of the instruments. Practicability: The systems are easy to operate. They are not bulky, they are quick, they are available 24 h a day and require only small volumes of whole blood. Membrane replacement procedures have been facilitated by the new electrode design.

Inter-instrument comparisons
As an example, the Ciba Corning instrument is equipped with maintenance-free electrodes and so remembraning is unnecessary. When an electrode needs to be replaced, the procedure is quick and easy: simply pull out the old electrode and slide in the new one. The software flexibility were two parameters with nearly uniform measurements. When discrepancies occurred, preanalytical errors were probably responsible. Imprecision and inaccuracy were more difficult to appreciate for pO2 than for pCO, mainly due to the high  [8]. This parallels to the larger question of whether the instrument differences found using quality-control ampoules are likely to predict similar differences if clinical samples of whole blood were to be measured.
Tonometered blood is ideal for examining interinstrument bias in pO, but when it is necessary to use more than one tonometer and more than a single source of tonometry gas, the variability of the blood gas measurement of pO increases. Thus the analysis and interpretation of systematic bias becomes more difficult.
Whole-blood tonometry allows the accuracy and the precision of the instruments to be evaluated [9]. However, this method requires knowledge of the composition of the gas mixture, control of the tonometer temperature and of the time for complete equilibration.
No mishandling should occur during the transfer of the equilibrated sample from the tonometer to the instrument [2]. Not only is the technique time-consuming but also the equipment is not widely available in France.
pO inter-instrument comparison with blood specimens identified the difficulty in maintaining the validity ofpO measurement. There are instrument differences in calibration methods for pO2 electrodes and measuring chamber size, measuring chamber content prior to sample introduction, sample introduction technique, sample size, sample warming, analysis time, and electrode signal processing, all of which can contribute to model-specific differences. Whether the inter-instrument differences in pO are real is still much discussed 10-12].
The linearity study showed that the pO responses of the sensors were in line with the manufacturers' specifications. This indicates that model specific algorithms to correct design and imperfections are valid. In fact, the discrepancies observed for the high values of pO are not really clinically relevant.
In this comparative study, the basic models were tested for each brand (see table 5). Thus three out of four analysers measured only pH and pO2, pCO. NOVA SP 5, chosen in agreement with the manufacturer, is a good example of a multichan.nel analyser combining determinations of other analytes. The appearance of combined electrochemical sensors for electrolytes, analytes, pH and blood gases have raised new issues in sensor calibration, sample collection and handling. Phosphate buffers, developed for pH calibration, interfere with the activities of Na+, K+, Ca 2+ in calibrating solutions. On the other hand, organic buffers, which do not affect these ions activities, lead to a pH bias on sample measurement 13].
Besides the quantification of the analytical performance of the analysers, this protocol draws users' attention to appropriate indicators of functioning for each system, and the difficulty in obtaining reliable data with respect to the various quality-control materials. The qualitycontrol materials which are available today are not ideal.
Aqueous materials have such advantages as a long shelflife and being prepared in ampoules which are ready to use; their disadvantage is poor oxygen buffering [14].
However, storage temperature of the samples will affect measurement of glucose or potassium. Specimens are stable for blood gas analysis for two hours when maintained at about C, reducing the glycolytic effect, but potassium increased significantly under the same conditions [19]. A compromise can be suggested: if the sample is measured within 10 min it can be maintained at  Finally, the choice of anticoagulant is critical when a Ca 2+ result is needed. Anticoagulant containing calcium or citrate, as well as heparin, will afect ionized calcium values. Calcium titrated heparinate is the best means to minimize calcium chelation, either as a solution in glass ampoule or as a dry preparation in syringe or capillary tube [20][21][22].
These multichannel analysers offer many advantages, but the question is still open as to the most efficient association of the parameters [23]. The answer depends on the total workload, type and proportion of tests usually ordered, and staffing.

Conclusion
The modern automated pH and blood gas instruments, with built-in micro-processors, have become a standard equipment in most clinical laboratories or intensive care units and remain the reference technique. The supplementary help of computer programs means that optimal information can be extracted from various and complex data. Although these types of sensors and analysers have proved to be efficient, another field of analytical research is the development of new sensors; the ultimate goal is continuous and non-invasive monitoring of pH and other blood gas parameters.