Continuous transcutaneous oxygen measurement: reproducibility and trace analysis in healthy adults

One hundred and fourteen transcutaneous oxygen traces were obtained from 46 adults using a mass spectrometer. Such traces usually rise to a plateau about which they fluctuate. To estimate the plateau level, here called the asymptote, two models are described--a time-series model and a growth model. If yt is the trace measurement at time t, for the time-series model. yt + 1 = m + alpha (yt-m) + epsilon t, and for the growth model dyt/dt = beta (m-yt), where parameters alpha and beta represent the rate of approach to the asymptote m. Least squares estimates of these parameters were determined for every trace. For every trace, the estimated asymptotes are similar for both models but the other parameters show less agreement. Considering the traces for each person in pairs, each trace pair is allocated to one of four groups according as the traces were recorded on the same or different sites and on the same or different days. The correlation coefficient (r) between the estimated asymptotes is greatest for traces recorded on the same site and day (r greater than 0.94, P less than 0.001) and least for traces from different sites and days (r greater than 0.71, P less than 0.001). The time series estimates of the asymptote are more credible than the growth model.