Accuracy and reliability of pulse oximetry at different arterial carbon dioxide pressure levels

The present study aimed to assess whether arterial carbon dioxide pressure (Pa,CO(2)) has an impact on agreement between oxygen saturation measured with pulse oximetry (Sp,O(2)) or arterial blood gas co-oximetry (Sa,O(2)). Sa,O(2) and Sp,O(2) determinations were obtained simultaneously from 846 patients under assessment for long-term home oxygen therapy in a specialised outpatient clinic. Both measurements were taken with patients seated and breathing room air. Agreement between Sa,O(2) and Sp,O(2) results was analysed by the Bland-Altman method and the Lin concordance coefficient. In addition, potential interactions of arterial oxygen tension (Pa,O(2)) or Pa,CO(2) on agreement were analysed by adjusted multivariate analysis. Upon comparison of Sa,O(2) and Sp,O(2) results, the Bland-Altman technique yielded a bias (95% confidence interval (CI)) of -1.24 (-6.86-4.38) and -1.32 (-7.78-5.15) when Pa,CO(2) >48 mmHg (6.39 kPa) or Pa,O(2) <54 mmHg (7.20 kPa), respectively. Estimate by Lin's coefficient (95% CI) in these cases was 0.88 (0.85-0.90) and 0.81 (0.77-0.85). Adjusted multivariate analysis, performed to assess the impact of pH, Pa,O(2), Pa,CO(2) and bicarbonate on bias, showed that Pa,O(2), Pa,CO(2) and their interaction terms were the most important predictors of the bias (standardised estimates of -0.54, -0.94, and 0.85, respectively). The effect of pH, although statistically significant, was small, and bicarbonate had no significant effect. Arterial carbon dioxide pressure status can contribute to impaired agreement between arterial oxygen saturation and arterial oxygen saturation measured with pulse oximetry, particularly in patients with hypercapnia.