How accurate are pulse oximeters in patients with acute exacerbations of chronic obstructive airways disease?

The aim of this study was to determine the extent of correlation and agreement between arterial oxygen saturation and oxygen saturation as recorded by transcutaneous pulse oximetry, with a view to identifying whether pulse oximetry can be used as an alternative to arterial values in the clinical management of patients with acute exacerbations of chronic obstructive airways disease (COAD) in the emergency department. It also aims to determine whether there is a cut-off level of oxygen saturation by pulse oximetry that can screen for significant systemic hypoxia in this group. This prospective study of patients with acute exacerbations of COAD who were deemed by their treating doctor to require an arterial blood gas analysis to determine their ventilatory status, compared arterial oxygen saturation with simultaneously recorded oxygen saturation measured by transcutaneous pulse oximetry. Data were analysed using Pearson correlation, bias plot (Bland-Altman) methods for agreement and the receiver operator characteristic (ROC) curve method for determination of a screening cut-off. Sixty-four sample-pairs were analysed for this study. Nine (14%) had significant hypoxia (arterial PO2 less than 60 mmHg). The correlation coefficient was 0.91. The bias (Bland-Altman) plot shows a constant bias of -0.758% and only fair agreement, with 95% limits for agreement of -8.2 to + 6.7%. With respect to the ROC curve analysis, the 'best' cut-off for detection of hypoxia was at oxygen saturation by pulse oximetry of 92% (sensitivity 100%, specificity 86%). In conclusion, there is not sufficient agreement for oxygen saturation measured by pulse oximetry to replace analysis of an arterial blood gas sample in the clinical evaluation of oxygenation in emergency patients with COAD. However, oxygen saturation by pulse oximetry may be an effective screening test for systemic hypoxia, with the screening cut-off of 92% having sensitivity for the detection of systemic hypoxia of 100% with specificity of 86%.