Pulse Oximetry Error in Patients With Single-Ventricle Palliation

Introduction

Pulse oximetry is a mainstay technology used to manage children with cyanotic congenital heart disease (CCHD). Using light absorption, the percentage of oxyhemoglobin (SpO₂) is calculated as a noninvasive estimate of arterial oxygen saturation (SaO₂). While managing children with CCHD, providers must ensure that the patient is within the acceptable range of hypoxemia because unexpectedly low SaO₂ indicates the need for intervention. In children with structurally normal hearts, the SpO₂ accuracy declines at lower SaO₂. The fact that pulse oximetry overestimates SaO₂ during hypoxia in patients with darker skin pigmentation has also been documented.¹ However, the degree of errors due to race and/or single-ventricle physiology in patients with CCHD has not been fully elucidated.2, 3, 4, 5 In this research letter, we used time-matched SpO₂-SaO₂ pairs collected during cardiac catheterization to investigate SpO₂ error as a function of SaO₂, self-reported race, and single-ventricle palliation.

Methods

We conducted a single-center, retrospective review of children with single-ventricle palliation undergoing cardiac catheterization. SpO₂ was measured using the IntelliVue X3 system (Philips), and SaO₂ was measured from arterial samples using the GEM Premier 4000 system (Werfen). Time-matched data pairs (recorded at a maximum of 60-second intervals) were extracted from the procedure log and race identifiers from the Society of Thoracic Surgeons and IMproving Pediatric and Adult Congenital Treatments databases. Self-identified African Americans (AAs) were compared with those who did not identify as AA. Patients were dichotomized to those with (Blalock-Thomas-Taussig shunt], patent ductus arteriosus stent, central shunt, and hybrid) and without diastolic runoff lesions.

Linear mixed-effect modeling was used to describe the relationship among the AA race, SaO₂, and SpO₂. The model included a random effect for the subjects and an interaction term to determine whether the AA race affects the relationship between SaO₂ and SpO₂. In the models, the SaO₂ was converted to a percentage <100% such that the intercept represented the expected SpO₂ when the SaO₂ was 100%. This study was approved by the Office of Research Integrity and Compliance.

Results

Four hundred thirty-nine SpO₂-SaO₂ pairs were extracted from 172 catheterizations performed on 130 patients (Figure 1D). Thirty patients (23%) representing 112 SpO₂-SaO₂ pairs (26%) were AAs. The median difference between SpO₂ and SaO₂ was +1% (IQR, −1% to +4%). The difference between SpO₂ and SaO₂ increased at lower SaO₂ (B ​= ​−0.38; P ​< ​.001). ​A significant interaction existed between the AA race and SaO₂ in the model of SpO₂ (P ​< ​.001), indicating that the relationship between SaO₂ and SpO₂ varied with the presence of the AA race. A significant interaction with SaO₂ also existed for age, sex, weight, body surface area, cardiac index, arterial blood pressure, pulse pressure, P_CO2, bicarbonate, the stage of palliation, and the presence of diastolic runoff, meaning that each of these factors affects the accuracy of SpO₂. The interaction between the AA race and SaO₂ in determining SpO₂ persisted even after controlling for these variables (P ​< ​.001). In the final model, which compared AA with non-AA subjects, the intercept was lower (91% vs 96%, respectively; P ​< ​.001) and the slope was less negative (−0.46 vs −0.75, respectively; P ​< ​.001), indicating that in AA patients, SpO₂ increasingly overestimated SaO₂ at lower SaO₂ (Figure 1A). For example, at an SaO₂ of 80%, the expected SpO₂ was 82% for AA patients vs 81% for non-AA patients, whereas at an SaO₂ of 60%, the expected SpO₂ was 73% for AA patients vs 66% for non-AA patients. ​Diastolic runoff was also correlated with SpO₂ values, which overestimated SaO₂ (Figure 1B), with the AA race and diastolic runoff representing additive risk factors (Figure 1C).

Figure. 1.

(A) Percentage of oxyhemoglobin (SpO₂) as a function of arterial oxygen saturation (SaO₂) and race. (B) SpO₂ as a function of SaO₂ and diastolic runoff lesions. (C) SpO₂ as a function of SaO₂, race, and diastolic runoff lesions. (D) Patient demographics. AA, African American; BSA, body surface area; BTTS, Blalock-Thomas-Taussig shunt; cath, catheterization; Hgb, hemoglobin; PA, pulmonary artery; PDA, patent ductus arteriosus; SaO₂, arterial oxygen saturation; SpO₂, percentage of oxyhemoglobin.

Discussion

Pulse oximetry significantly underestimates critical hypoxemia in self-identified AA patients and in the presence of runoff lesions, with these errors being additive. Falsely elevated SpO₂ readings may cause clinicians to underdiagnose critical hypoxemia in AA patients with single-ventricle palliation and diastolic runoff, a population known to be at a high risk of mortality in the interstage period.

Without access to the manufacturer’s proprietary algorithm for SpO₂, we can only speculate the reason for increased errors in the setting of runoff lesions. In principle, all pulse oximeters measure the ratio of pulsatile-to-nonpulsatile absorbance. These data were collected from healthy volunteers, and an empiric algorithm was generated. We posit that in the setting of runoff lesions, the pulsatile component is significantly increased, thereby rendering the empiric formula less accurate. A next-generation device capable of correcting for the presence of widened pulse pressures could be invaluable.

The primary limitation of this study is that the correlation between skin tone and SpO₂ accuracy was not directly investigated. Rather, self-identified race was used as a proxy for skin tone. A single brand of pulse oximeter at a single institution was studied, and additional data are required to validate these results more broadly. Finally, potential factors, such as methemoglobinemia or carboxyhemoglobin, were not directly measured in this study but are known to affect the accuracy of pulse oximetry.

Conclusion

SpO₂ accuracy is negatively impacted by worsening hypoxemia, the AA race, and the presence of diastolic runoff, a novel finding of this study. The wide pulse pressure associated with runoff lesions was outside of the standard physiology based on which pulse oximetry was designed and validated and, thus, warrants further study. Ultimately, these data demonstrate that pulse oximetry can be falsely reassuring in our highest-risk patients with CCHD and present concerns that warrant further attention from device manufacturers and clinicians alike.