Pulse Oximetry in Acute Respiratory Failure: What Should Be Expected?-Reply.

To the Editor: In the April issue of RESPIRATORY CARE, Amalakanti and Pentakota1 present a comparison of arterial oxygen saturation values between a 2-wavelength pulse oximeter and blood gas multi-wavelength hemoximetry in COPD subjects with acute respiratory failure. The authors conclude that the 2-wavelength pulse oximeter significantly overestimates arterial oxygen saturation and that COPD and its phenotypes have a causal role. However, this study has a number of shortcomings, which may limit the applicability of the authors’ conclusions. By comparing functional and fractional oxygen saturations, the authors are really comparing green apples with red apples. Two-wavelength pulse oximeters report functional oxygen saturation. Functional oxygen saturation (SpO2 via pulse oximetry) only accounts for oxyhemoglobin and reduced hemoglobin: oxyhemoglobin/(oxyhemoglobin reduced hemoglobin) 100. Functional oxygen saturation can also be calculated by blood gases (SaO2). Multiwavelength oximeters measure fractional oxygen saturation, which includes carboxyhemoglobin and methemoglobin: oxyhemoglobin/(oxyhemoglobin reduced hemoglobin carboxyhemoglobin methemoglobin) 100. It is well documented that 2-wavelength pulse oximeters are not reliable in the presence of increased carboxyhemoglobin and methemoglobin.2,3 The authors concede that not accounting for the effect of carboxyhemoglobin on their results is a shortcoming; however, the methods section states that carboxyhemoglobin was measured, but it was not reported in the study. Had the authors compared functional saturation by the 2 methods, they would have probably seen closer agreement, as shown in Figure 1.4 To compare oxyhemoglobin between blood hemoximetry and pulse oximetry, a multi-wavelength pulse oximeter is required.5 It is notable that the mean PO2 the authors reported is what would be expected from the mean SpO2 via pulse oximetry (a slight rightward shift of the oxyhemoglobin dissociation curve due to hypercapnia and presumed acidosis). For the purposes of oxygen titration and the avoidance of hyperoxemia, the data seem to indicate that pulse oximetry performed quite well. The authors correctly used Bland-Altman analysis to assess agreement between the 2 methods and conclude that “the pulse oximeter consistently overestimates oxygen saturation.” However, the agreement plots show that in many instances, pulse oximetry markedly underestimated oxyhemoglobin (SaO2 SpO2 0). The agreement plots also indicate that in some comparisons, the disparity between the 2 methods is quite large ( 30% and 20%). These findings are unusual, yet there is no recognition or explanation for these observations in the paper. The authors propose a causal relationship between COPD (and 2 of its phenotypes) and pulse oximetry inaccuracy. A control group is necessary to make the case for a causal relationship. In this study, a control group (ie, subjects with acute respiratory failure without COPD) is missing, precluded by design a priori. There is a subgroup called “hypoxia” in a table but there is no description of these subjects in the paper. In addition, the use of an FEV1/FVC 0.7 to define COPD instead of a statistically valid lower limit of normal (eg, Z score 1.64) can be expected to include subjects without COPD.6 Using a binary approach to COPD phenotyping (ie, chronic bronchitis or emphysema) is inappropriate because COPD has several phenotypes.7 In our opinion, this study fails to test what can be expected from a 2-wavelength pulse oximeter and by design cannot validate a causal relationship between COPD phenotypes and pulse oximetry inaccuracy.