Comparative analysis of signal accuracy of three SpO(2) monitors during motion and low perfusion conditions

To compare pulse oximetry performance during simulated conditions of motion and low perfusion in three commercially available devices: GE HealthCare CARESCAPE ONE TruSignal SpO(2) Parameter, Masimo RADICAL-7 and Medtronic Nellcor PM1000N. After IRB approval, 28 healthy adult volunteers were randomly assigned to the motion group (N = 14) or low perfusion (N = 14) group. Pulse oximeters were placed on the test and control hands using random assignment of digits 2–5. Each subject served as their own control through the series of repeated pair-wise measurements. Reference co-oximetry oxyhemoglobin (SaO(2)) measurements from the radial artery were also obtained in the motion group. SpO(2) readings were compared between the test and control hands in both groups and to SaO(2) measurements in the motion group. Accuracy was assessed through testing of accuracy root-mean squared (ARMS) and mean bias. In the simulated motion test group the overall Accuracy Root Mean Square (ARMS) versus SaO(2) was 1.88 (GE), 1.79 (Masimo) and 2.40 (Nellcor), with overall mean bias of − 0.21 (Masimo), 0.45 (GE), and 0.78 (Nellcor). In the motion hand, ARMS versus SaO(2) was 2.45 (GE), 3.19 (Masimo) and 4.15 (Nellcor), with overall mean bias of − 0.75 (Masimo), − 0.01 (GE), and 0.04 (Nellcor). In the low perfusion test group, ARMS versus the control hand SpO(2) for low PI was 3.24 (GE), 3.48 (Nellcor) and 4.76 (Masimo), with overall bias measurements of − 0.53 (Nellcor), 0.96 (GE) and 1.76 (Masimo). Experimental results for all tested devices met pulse oximetry regulatory and testing standards requirements. Overall, SpO(2) device performance across the three devices in this study was similar under both motion and low perfusion conditions. SpO(2) measurement accuracy degraded for all three devices during motion as compared to non-motion. Accuracy also degraded during normal to low, very low, or ultra low perfusion and was more pronounced compared to the changes observed during simulated motion. While some statistically significant differences in individual measurements were found, the clinical relevance of these differences requires further study.