Evaluation of the Ohmeda 3700 pulse oximeter: steady-state and transient response characteristics

The authors determined the accuracy of the Ohmeda 3700 (version J) pulse oximeter in healthy volunteers rendered hypoxic (SaO2 from 60-98%) by breathing mixtures of O2 in N2. When equipped with an ear probe, the pulse oximeter reading (y) reliably predicted arterial saturation (x) under steady-state conditions (y = 1.05x - 4.66, r = 0.98) as well as when oxygen saturation was rapidly decreasing (y = 1.05x - 6.38, r = 0.96). Conversely, when equipped with a finger probe, the oximeter tended to significantly underestimate steady-state arterial saturation (y = 1.21x - 19.1, r = 0.98, P less than 0.001). In response to this information, the manufacturer modified the oximeter's software (version XJ1), resulting in improved agreement between oximeter readings and arterial values (y = 0.96x + 4.59, r = 0.99). Despite the close correlation between steady-state oximeter readings and arterial saturation, the 99% prediction limits for both the ear and finger probes (version XJ1) were +/- 8%. Finger probe readings did not reliably reflect radial arterial oxygenation during rapid desaturation (y = 0.55x + 45.2, r = 0.78). This may be related to the time required to "arterialize" the blood in the finger; during acute resaturation, we found that the ear- to finger-probe delay was 24.0 +/- 2.3 s (means +/- SE, P less than 0.001).