Pulse Oximetry Misclassifies Need for Long-Term Oxygen Therapy in Chronic Obstructive Pulmonary Disease

In the early 1980s, two landmark studies established the efficacy of long-term oxygen therapy (LTOT) for individuals with chronic obstructive pulmonary disease (COPD) using partial pressure of oxygen (Pa_O2) from arterial blood gases (ABGs) to define hypoxemia (1, 2). At the time, the first commercial pulse oximeters were coming to market based on the technologic discovery of Takuo Aoyagi less than a decade earlier; the use of pulse oximetry did not become ubiquitous in healthcare settings until the 1990s. Despite a lack of clinical evidence supporting its use in establishing a need for LTOT, approximate arterial oxygen saturation based on pulse oximetry (Sp_O2) was adopted into governmental and organizational guidelines (3, 4). The current insurance reimbursement criteria for LTOT in the United States use Sp_O2 cutoffs of <88–89% as an alternative to Pa_O2 cutoffs (⩽55 mm Hg or ⩽59 mm Hg in the presence of pulmonary hypertension, cor pulmonale, right heart failure, or polycythemia) despite previous studies raising concern that using saturation thresholds to establish hypoxemia for patients with COPD could lead to misclassification (5).

In this issue of AnnalsATS, Garnet and colleagues (pp. 1587–1594) use paired Sp_O2 and ABG measurements from 518 outpatients with COPD in stable condition at the Miami Veterans Affairs Medical Center to evaluate pulse oximeter performance in identifying severe resting hypoxemia and the need for LTOT (6). This analysis was facilitated by the Miami Veterans Affairs Medical Center’s adherence to the evidence-based practice of routinely evaluating LTOT candidates with ABGs, a practice that has largely been abandoned in favor of exclusively relying on Sp_O2 criteria, which are adequate for insurance reimbursement based on Centers for Medicare and Medicaid Services guidelines (4). The authors found that solely relying on Sp_O2 criteria would have misclassified 70% of patients qualifying for LTOT for severe resting hypoxemia based on Pa_O2 criteria, for an overall false negative rate of 10%, leading to inappropriate withholding of oxygen therapy. The proportion misclassified and false negative rate were higher among active smokers but not different by self-reported racial identification. Using receiver operating characteristic analysis, the authors determined that an overall threshold of resting Sp_O2 ⩽94% should be used to prompt reflex ABG testing, with Sp_O2 ⩽95% as the threshold for active smokers. These cutoffs are higher than that suggested by the Global Initiative for Chronic Obstructive Lung Disease report, which recommends ABG testing if Sp_O2 is <92% (7). Oximetry is similarly poor for establishing the need for LTOT in the inpatient setting. A recent letter in AnnalsATS reported a false negative rate of 28% for the detection of severe hypoxemia (Pa_O2 <56 mm Hg) using Sp_O2 ⩽88% in a cohort of 46 patients with 87 concurrent Sp_O2 and Pa_O2 measurements (8). Using receiver operating characteristic analysis, the authors of this letter suggested a screening cutoff of Sp_O2 ⩽92% to prompt reflex ABG testing. This is similar to the cutoff value established in a two-decade-old study of individuals seeking acute care for COPD exacerbation, which defined systemic hypoxia as a Pa_O2 <60 mm Hg and formed the basis of the Global Initiative for Chronic Obstructive Lung Disease recommendation for obtaining reflex ABG testing when oxygen saturation is <92% (7, 9). The study by Garnet and colleagues highlights that different clinical settings may affect pulse oximeter performance such that thresholds established in the inpatient setting cannot necessarily be applied to the outpatient setting, where exposures such as smoking and recent activity level may alter the prediction of oxygen saturation by noninvasive oximetry. Furthermore, whereas the Centers for Medicare and Medicaid Services regulate ABG testing in clinical settings through the Clinical Laboratory Improvement Amendments, ensuring a minimum quality threshold that produces comparable results across hospital laboratories, there are no standards for pulse oximeter calibration, recording, or reporting, making it difficult to compare results across studies or generalize them to the clinical setting.

The efficacy of supplemental oxygen among individuals with COPD without severe resting hypoxemia was recently reevaluated using pulse oximetry criteria. In a randomized controlled trial of individuals with COPD and moderate resting or ambulatory hypoxemia, LTOT had no effect on time to first hospitalization or death (10). Importantly, the determination of moderate hypoxemia relied solely on pulse oximetry (Sp_O2 of 89–93% at rest or ⩾80% for ⩾5 min and <90% for ⩾10 s during a six-minute-walk test). The study by Garnet and colleagues demonstrating that pulse oximeters misclassify a substantial proportion of individuals with COPD and severe resting hypoxemia based on Pa_O2 raises the question whether the trial inadvertently included individuals with severe resting hypoxemia. In the study, 39 of the 74 individuals with severe resting hypoxemia (53%) had an Sp_O2 of 89–92%. Inclusion of these individuals in a trial meant to evaluate moderate hypoxemia introduces uncertainty in interpreting the findings and suggests that the efficacy of LTOT among individuals with severe hypoxemia may need to be reevaluated in the modern era. Conversely, the trial may lead clinicians to conclude that supplemental oxygen therapy has no benefit among individuals with COPD and resting Sp_O2 between 89% and 93% without considering that individuals with severe resting hypoxemia may be “hiding” within that group as a result of pulse oximeter inaccuracy (i.e., overestimation of true oxygen saturation) and the misclassification of hypoxemia severity when using oxygen saturation.

The use of pulse oximetry to screen and diagnose hypoxemia in COPD is the latest example of rationalizing the use of a convenient medical device by applying data out of context without strong evidence to support clinical decisions. For instance, pulse oximetry has become ubiquitous in the acute care setting despite evidence at the turn of the 21st century of the poor performance of pulse oximeters in this context (11). Likewise, warnings that pulse oximeters were less accurate in individuals with darker pigmentation went unheeded for decades by the medical community and regulators until investigations spurred by the coronavirus disease (COVID-19) pandemic highlighted the tendency of pulse oximeters to overestimate true oxygen saturation among individuals of racial and ethnic minorities and the adverse clinical consequences of this health disparity (12–16). In addition to evaluating the misclassification of the need for supplemental oxygen, Garnet and colleagues evaluated the risk factors for inaccurate estimation of arterial oxygen saturation by pulse oximetry. They showed that Sp_O2 overestimates arterial oxygen saturation among smokers and individuals self-identifying as Black, although the latter did not reach statistical significance, possibly because of a lack of statistical power. Together, the use of oxygen saturation rather than Pa_O2 and the reliance on pulse oximetry to estimate oxygen saturation likely amplify the misclassification of severe hypoxemia (5).

Mounting evidence shows that pulse oximetry misclassifies individuals with COPD and severe hypoxemia, thereby leading to inappropriate withholding of supplemental oxygen therapy. Organizations that promote evidence-based guidelines should consider more nuanced recommendations for supplemental oxygen prescriptions that incorporate evidence of misclassification of severe hypoxemia using pulse oximetry.