PERFORMANCE-CHARACTERISTICS AND INTERANALYZER VARIABILITY OF PO-2 MEASUREMENTS USING TONOMETERED HUMAN BLOOD

We performed a side-by-side comparison of the ability of four blood ga s analyzers (IL-1312, Corning-178, AVL-995, and ABL-330) to measure PO 2 across a wide range under controlled laboratory conditions. Samples of fresh whole human blood, tonometered with analytic quality gas, wer e prepared with partial pressures of oxygen from 0 to 283 mm Hg. Fifte en determinations were made at 16 levels of tonometric PO2 (tPO2) on e ach of the four blood gas analyzers. The bias, precision, and root mea n squared error (RMSE) of the PO2 measurement relative to tPO2 were de termined for each analyzer at each tPO2 level. Mean bias and precision across the range tested were 2.78 +/- 1.29 mm Hg (IL), -0.35 +/- 1.91 (Corning), 2.14 +/- 1.43 (AVL), and 3.00 +/- 1.47 (ABL). RMSE was 3.2 8, 2.61, 3.57, and 2.41 for IL, AVL, ABL, and Corning, respectively. P ercent RMSE (RMSE/ tPO2 X 100%), ranged from 0.9% (AVL at 75 mm Hg PO2 and IL at 283 mm Hg tPO2) to 9.1% (IL at 29 mm Hg tPO2). Three analyz ers (AVL, ABL, and Corning) showed a statistically significant (p < 0. 0001) correlation between RMSE and tPO2, and no correlation between pe rcent RMSE and tPO2. This demonstrates that, for these instruments, ac curacy is a function of the magnitude of the tPO2 value. IL did not sh ow a significant correlation between RMSE and tPO2 but did demonstrate a significant negative correlation (r = -0.78, p > 0.001) between per cent RMSE and tPO2, indicating that, for this analyzer, accuracy is no t a function of tPO2. The differences in PO2 measurements between pair s of analyzers were also examined. The mean +/- 95% confidence limits for the overall differences ranged from a low of -0.87 +/- 3.90 (AVL - ABL pair) to a high of 3.30 +/- 7.49 (ABL - Corning pair), owing to t he analyzer-specific differences in systematic tendency to over- or un derestimate the target PO2 (ie, positive or negative bias). There are statistically significant differences in both means and variances betw een pairs of blood gas analyzers across most of the ranges of tPO2 tes ted. Under controlled laboratory conditions, the bias ranges from -3.0 6% to 8.62% and precision from 0.82% to 3.76% of the tPO2 on any one a nalyzer; however, interanalyzer differences in reported PO2 Of 10% or more (95% confidence limit) could occur if the same sample were run co ncurrently on two different brands of blood gas analyzers.