Differences in Arterial Oxygen Saturation Measurements Between Two Co-Oximeters From Different Manufacturers

Background: Functional arterial hemoglobin oxygen saturation is a critical clinical parameter and the reference for verifying accuracy of S_pO2 values during performance testing of a pulse oximeter. Variable performance between blood gas analyzers that perform co-oximetry may complicate interpretation of research studies and regulatory guidance. This study quantifies systematic differences between multiple same model analyzers from two widely used brands of co-oximeters. Methods: Three sequential experiments were conducted using controlled hypoxemia in healthy adult volunteers according to regulatory clearance criteria for use of functional arterial oxygenation from co-oximetry analysis (sO₂), to simultaneous S_pO2 values over the range of 70-100%. To account for potential methodological confounders in experiments 1 and 2, a third experiment analyzed 31 arterial samples across multiple same model analyzers for co-oximetry (two Werfen GEM Premier 5000, and three Radiometer ABL90 Flex) in a randomized order after simultaneous preparation. Statistical analysis included paired differences, mixed-effects modeling, and variance analysis to isolate analyzer-specific effects while controlling for confounding variables. Results: The GEM analyzers consistently measured sO₂ 2.3% higher on average than ABL analyzers (95% CI 2.01-2.63%, P < .001). This difference was more pronounced at lower saturations (3.2% in the sO₂ 70s range vs 1.4% in the 90s range). Despite strong correlation between measurements (r = 0.999), 96.8% of paired samples differed by >1% and 61.3% by >2%. Within-manufacturer variability was similar between brands (mean absolute difference: ABL 0.38%, GEM 0.47%, P = .55). Conclusions: Clinically important systematic differences in sO₂ measurement exist between GEM and ABL blood analyzers that perform co-oximetry, with potential implications for patient care and pulse oximeter verification studies. Larger studies with more co-oximeter brands, coupled with blood tonometry, are needed to better characterize these findings. In the meantime, regulatory standards and reports of pulse oximeter clinical trials should account for these findings and specify which manufacturer's co-oximeters were used.