RESUMEN
Although large differences in mean values and standard deviations of PO2 between models of blood gas instruments have been found in proficiency testing surveys employing commercial ampuled material, it is unknown whether or not equivalent differences would be found between instruments if a primary standard (tonometered fresh human blood) were used instead of ampules (secondary control specimens). Therefore, we compared the biases (differences between means) of six models of blood gas instruments using five varieties of ampuled quality control materials and a single source of tonometered fresh human blood at three levels of PO2 (67.5, 91.4, and 143.6 mm Hg). All materials correlated positively with blood, thus demonstrating consistency of machine biases. The coefficients of variation measuring blood and a perfluorocarbon mixture (abc) were not different, but were higher for aqueous and hemoglobin-containing mixtures. On average, at the PO2 levels tested, the differences among instruments were slightly smaller with abc than with blood, but larger with aqueous and hemoglobin-containing materials. Because abc "tracks" tonometered blood reasonably well at the PO2 levels tested, instrument differences in PO2 detected with abc are also likely to be seen with clinical blood samples. The other four quality control materials probably overestimate instrument differences in measuring PO2.