QP-11: Blood Gas Applications
|
|
Blood gas measurements are often performed in Point-Of-Care (POC) settings as well as larger clinical or pulmonary laboratories. Patients may be monitored during surgery and in the recovery room via the POC tests. Later, when the patients are moved back to their hospital rooms, they may be monitored by measurements performed in a central clinical laboratory or pulmonary function laboratory. The comparability and consistency of test results will be important.
When a new POC analyzer is obtained, the performance of the analyzer should be evaluated by comparison to the measurements being performed in an established laboratory, such as the central clinical chemistry laboratory or the pulmonary function laboratory. Method validation studies will provide estimates of imprecision and bias that can then be used for planning QC procedures.
The US CLIA criteria for acceptable performance in proficiency testing are as follows:
Blood gas pH Target Value +/- 0.04 pH unit Blood gas pCO2 Target Value +/- 5 mm Hg or +/- 8% (greater) Blood gas pO2 Target Value +/- 3 SD To quantify the requirement for pO2, results from an AACC/CAP proficiency testing survey during 1999 showed the following SDs and CVs for pO2 for "all" analyzers, which represent approximately 3500 laboratories that participated in the survey:
- At 145 mm Hg, SD=6.7 mm Hg, CV=4.6%
- At 115 mm Hg, SD=7.1 mm Hg, CV=6.1%
- At 118 mm Hg, SD=7.5 mm Hg, CV=6.4%
- At 59 mm Hg, SD=14.7 mm Hg, CV=25.1%
- At 144 mm Hg, SD=7.2 mm Hg, CV=5.0%
The SD and CV for the 4th specimen (59 mm Hg) appear to be inconsistent with the rest of the figures and further inspection showed that different manufacturers' systems provided large differences in their mean values, demonstrating that the "all" or overall SDs and CVs may reflect a large contribution from systematic errors between different types of analyzers. Averaging the results from the other four specimens gives an SD of 7.1 mm Hg and a CV of 5.5%. The CLIA requirement for pO2 can therefore be estimated to be about +/- 16.5%.
To estimate the imprecision of the analyzer, initial method validation studies should include a replication study performed over a period of 20 days. Typically, three different control materials would be analyzed and the SDs and CVs calculated for each level. To estimate inaccuracy or bias, a comparison of methods experiment should be performed on a minimum of 20 and preferably 40 fresh patient samples. The paired data could be analyzed by regression and t-test statistics.
For the examples here, the estimates of method CVs are similar to the performance documented in a recent published evaluation study of a POC analyzer [1]. Three different levels of control materials are generally provided for blood gas measurements. Typical levels would be 7.20, 7.40, and 7.60 for pH; 60, 40, and 20 mm Hg for pCO2, and 60, 100, and 140 mm for pO2.
NOTE:
- Normalized OPSpecs operating points can be calculated on this website using the Normalized OPSpecs Calculator. Click here to see the calculator.
- The Normalized OPSpecs charts can be download from this website in PDF format. Click here to download the charts.
The average method SD for the three control materials is 0.005 pH unit and the method bias is 0.01 pH unit. The CLIA requirement is 0.04 pH units.
- The normalized operating point has an x-coordinate of 12.5% [(0.005/0.04)100] and a y-coordinate of 25% [(0.01/0.04)100].
- When plotted on the normalized OPSpecs chart for N=3 with 90% AQA, the operating limits of all the QC procedures are above the operating points (see figure above). Any control rule with N=3 can be used, but it would be best to select a 13s or 13.5s rule to keep false rejections very low.
- The Total QC strategy should depend on statistical QC to detect problems and include the minimum preventive procedures recommended by the manufacturer and by good laboratory practice.
The average CV for the three control materials 2.5%. Method bias is observed to be 0.2 mm Hg at the mean of the comparison data (at approximately 40 mm Hg). The quality requirement at 40 mm Hg would be 5 mm Hg or 12.5%.
- The normalized operating point has an x-coordinate of 20% [(2.5%/12.5%)100] and a y-coordinate of 5.0% [(0.2mm/5mm)100].
- When plotted on the normalized OPSpecs chart for N=3 with 90% AQA, there are four possible QC procedures (see figure above). A multirule procedure with 13s/2of32s/R4s rules or a single rule procedure using a 12.5s rule could be selected. The multirule procedure is preferred because it will give fewer false rejections (1.0% vs 3.0%).
- The Total QC strategy can again depend on statistical QC to detect important problems whenever they occur.
The average CV is observed to be 4.1% and the bias between methods is nearly zero. As shown earlier, the CLIA quality requirement is 16.5%, as calculated from results of a proficiency testing survey where the group CV is estimated as 5.5% on the basis of 4 different PT specimens.
- The normalized operating point has an x-coordinate of 25% [(4.1%/16.5%)100]. The y-coordinate is zero.
- When plotted on a normalized OPSpecs chart for N=3 and 90%AQA, the only solution is the 12s rule, which has an unacceptably high level of false rejections (approximately 14%).
- When plotted on a normalized OPSpecs chart for N=6 and 90% AQA, there are four solutions available. However, it's unlikely that POC testing sites will be willing or able to run 6 controls to monitor pO2. Therefore, an N=3 solution will be needed.
- When plotted on a normalized OPSpecs chart for N=3 and 50%AQA, there are two multirule procedures that will provide at least 50% error detection (see figure above). The fact that their operating lines are considerably above the operating point indicates that the error detection should be considerably higher than 50%. The best selection to give maximum error detection would be the multirule procedure that includes a 31s control rule.
- The Total QC strategy for pO2 should include an aggressive preventive maintenance program to minimize the problems that occur. Additional QC checks such as the use of tonometered blood would be a valuable addition to monitor performance. Correlation with oxygen saturation may be a useful addition to check the reliability of some specimens.
It is not unusual to see a range of performance for different blood gas parameters - from excellent for pH, good for pCO2, and marginal for pO2. The method CVs used in the examples here are representative of the performance that can be expected from current portable blood gas analyzers [1], thus there is a need for well designed QC procedures to assure comparability of results between point-of-care and central laboratory applications. The problem of how to QC pO2 measurements is further complicated by matrix effects of the control materials [2].
- Use subsets of a multirule procedure to provide different QC designs. The differences in analytical performance and quality requirements will typically lead to different QC designs for each of these tests. One way to relate these different QC procedures is to select rules that are part of a particular multirule combination. In this case, select the 13s/2of32s/R4s/31s combination and utilize the whole set of rules for pO2, drop the 31s rule for pCO2, and drop all but the 13s rule for pH. This provides a set of QC designs that are all related to a single multirule procedure, which will be the focus of in-service training.
- Implement multistage designs for startup and monitoring. In situations where it is not possilbe to achieve the ideal performance of 90% error detection with less than 5% false rejections, implement two different QC designs - one having 90% error detection (with some increase in false rejections if necessary) and the other having low false rejections (with moderate error detection). Use the high error detection design to check carefully at "startup" and the low false rejection design to "monitor" performance after startup. For blood gas measurements, it may be useful to have an N=3 startup design that is performed once a day and an N=1 monitoring design that is performed by new operators and at eight-hour shift changes.
- Accept lower error detection when there is frequent calibration. Many blood gas analyzers have automatic calibration functions that are scheduled at short intervals, often hourly. With this frequent calibration, most instrument sources of shifts or drifts are automatically corrected and therefore do not require detection by the QC procedure. Optimal QC performance then depends on maintaining a low level of false rejections. Moderate error detection - 50% or more - should be tolerable.
- Utilize instrument function checks as part of TQC strategy. Again, because of the automatic checks programmed into many blood gas analyzers, the TQC strategy can support QC designs with moderate error detection. Instrument function checks, such as electronic QC, can be used to complement the detection capabilities of statistical QC procedures.
