Reliability of six pulse oximeters in chronic obstructive pulmonary disease

Six pulse oximeters with finger probes were studied in three groups of 17 hypoxemic patients with COPD aged 50 to 75 years. Transcutaneous arterial oxygen saturation (SpO2) was measured with the Nellcor N101 (oximeter 1a), the Ohmeda Biox III (oximeter lb), the Nellcor N200 (oximeter 2a), the Critikon Oxyshuttle (oximeter 2b), the Radiometer Oxi100 (oximeter 3a), and the Ohmeda Biox 3700 (oximeter 3b). The SpO2 was compared with SaO2 measured in simultaneously withdrawn samples of arterial blood (Radiometer OSM2) at three 20-minute steady-state levels of FIO2 ranging from 0.21 to 0.40 (SaO2, 62 to 100 percent). The bias (mean SpO2-SaO2 difference) and the error in precision (SD of the differences) were both below 4 percent for instruments 1a and 1b and remained below 1.2 and 3 percent, respectively, for the others. A good agreement between SpO2 and SaO2, as reflected by the Bartko intraclass correlation coefficient, was observed in instruments 2a, 3a, and 3b. The individual relationships between SpO2-SaO2 differences and SaO2 appeared to be linear and parallel. With four instruments (1a, 1b, 2a, and 2b), the mean slope of this relationship was negative, showing a systematic instrumental error: the lower the SaO2, the larger the overestimation of SaO2. The scattering of the data (precision) principally reflects a subject source of error. In most instruments a technical adjustment could greatly improve instrumental errors and accuracy. The correction of the errors due to between-subject variation would require a system of calibration adjustable by the users to each individual.