Abstract
The aim of this study was to compare quantitatively the changes in tissue oxygen saturation (TOS), determined by two algorithms (TOSc and TOSa) based on near-infrared spectrophotometry, to the changes in arterial oxygen saturation (SaO2) measured by pulse oximetry. TOSc is an algorithm derived by the manufacturer (Critikon) based on a modified Beer-Lambert law; TOSa, our own algorithm, uses the diffusion approximation of light transport for the semi-infinite boundary condition.
Slow changes of more than 3% in SaO2 were carried out in 20 mechanically ventilated neonates by altering the inspired oxygen fraction. For each change the regression lines of TOSc versus SaO2, TOSa versus SaO2 and TOSc versus TOSa were calculated. For each infant the mean slope, intercept and r2 of these lines were determined.
In 18 preterm infants we obtained median 9.5 (range one to 13) measurements corresponding to a total of 166 measurements.
The mean SaO2 was 91.6 (SD 2.3)%, TOSc was 64.7 (SD 7.2)% and TOSa was 71.4 (SD 11.0)%. Changes in TOSc and TOSa were strongly correlated to changes in SaO2 (r2 = 0.86 and r2 = 0.87). TOSc considerably but systematically underestimated the size of the change: ΔTOSc = 0.49 ΔSaO2. TOSa quantified changes reasonably correctly: ΔTOSa = 0.90 ΔSaO2. Changes in TOSc and TOSa were highly correlated (r2 = 0.98). These results are promising, but the large inter-individual variation requires further work.