Do Differences in Skin Pigmentation Affect Detection of Hypoxemia by Pulse Oximetry: A Systematic Review of the Literature

Abstract

Pulse oximetry is a widely used, noninvasive method for estimating arterial oxygen saturation (SaO2). However, emerging evidence suggests that skin pigmentation may affect its accuracy, potentially leading to occult hypoxemia in individuals with darker skin tones. This systematic review examines the impact of skin pigmentation on pulse oximeter accuracy by comparing pulse oximetry (SpO2) readings with arterial blood gas-measured SaO2 across diverse populations. A systematic search of PubMed and Embase was conducted following PRISMA 2020 guidelines. Eligible studies included those comparing SpO2 to SaO2 while stratifying results by skin pigmentation or race/ethnicity. Data extraction focused on bias in SpO2 readings, study design, and population characteristics. Risk of bias was assessed using the QUADAS-2 tool. Forty-two studies met the inclusion criteria. Consistent evidence indicated that pulse oximeters overestimate SaO2 in individuals with darker skin tones, particularly at lower oxygen saturations. This overestimation may delay recognition of hypoxemia and critical interventions. Methodological variability was noted, including inconsistent racial classifications and skin tone assessment methods. Pulse oximeters exhibit a systematic bias in individuals with darker skin tones. Standardized skin pigmentation assessment and improved device calibration are needed to enhance accuracy and ensure equitable patient care.

Introduction

Pulse oximetry (SpO2) is a quick, inexpensive, noninvasive assessment of blood oxygenation that estimates the percentage of oxygen bound to hemoglobin (SaO2). The estimate is derived from a ratio of red to infrared light at a given point in time, which is driven by the amount of arterial versus venous blood, which changes with the pulse. As one of the five core vital signs, oxygen saturation often informs patient care, and a SpO2 of >92% is normal for most adults. While many biological processes can interfere with pulse oximetry readings (e.g., methemoglobinemia, anemia, hyperbilirubinemia), little is known about the extent to which skin pigmentation can affect the accuracy of the device.

Previous research has shown that pulse oximeters are less accurate in people who identify as non-white, including black (Fawzy et al., 2024; Jubran & Tobin, 1990; Sjoding et al., 2020; Valbuena et al., 2022). Concerningly, these studies demonstrated that the SpO2 value was often falsely high in non-white individuals, leading to a phenomenon termed “occult hypoxemia,” wherein the actual arterial saturation was low enough to warrant clinical action while the noninvasive saturation was reassuringly “normal.” In the study by Sjoding et al., for example, occult hypoxemia occurred three times as often in black patients as it did in white patients in data from an intensive care unit (ICU) database (Sjoding et al., 2020). These results prompted a letter to the United States Food and Drug Administration (FDA) (2021) and a collaborative working group (FDA, 2022) to address this problem. Despite these advances, little has been done to characterize what, specifically, about non-white individuals leads to these reproducible discrepancies. Some have hypothesized that differential melanin, the pigment responsible for skin color, expression, affects the absorbance of light used in pulse oximeters. However, there are scant prospective data that address this issue. Rather, most studies have focused on categorization of skin pigmentation using patient self-report, visual scales that lack nuance (Verkruysse et al., 2024) or subjective scales like the Fitzpatrick skin type (FST) that are misapplied to the context (Fitzpatrick, 1975). We have undertaken this systematic review to discover and categorize studies that reported skin pigmentation in addition to the difference between the noninvasively measured saturation (SpO2) and the invasively measured saturation (SaO2) through arterial blood gas (ABG) analysis.

This review encompasses articles that compared the SaO2 and the SpO2 and reported the difference between the two values. We searched for articles that included classification of skin tone and compared the difference between the two saturation values based on race, ethnicity, or skin tone. The authors believe that while much research has been done on SpO2 and race, it is important to review key literature that addresses skin color or skin pigmentation and how it relates to pulse oximetry. This systematic review includes articles validating the accuracy of pulse oximeters and those showing the difference between the two saturation values and the precision of the device. Based on this conceptual framework, this systematic review aims to test the hypothesis that, based on the literature, a discrepancy exists between SpO2 and SaO2 as a function of skin pigmentation or color.

Methods

Published, peer-reviewed journal articles were included if the following criteria were met: pulse oximeter bias was reported as a result of SpO2 minus SaO2, the study was written or translated into English, the full text of the study was available, the population was adults >18 years old, race and/or ethnicity was reported, and the method by which race/ethnicity was determined was reported. In the initial search, we included any articles, case reports, or conference presentations that met these criteria. For conference presentations, we included those that met the inclusion criteria, which could be interpreted from written and poster presentations.

We conducted a systematic literature search in two databases, including PubMed and Embase, to identify relevant studies examining the relationship between oximetry and skin pigmentation (Supplement 1). No restrictions were placed on language or publication date.

We searched for English-language articles published through October 2023, with subsequent updates to PubMed and Embase searches through February 2025. The original and subsequent search strategies yielded 1,575 articles. After duplicates were removed, titles and abstracts of the remaining 1,125 studies were independently reviewed by two authors (SC and WCM). Following the database search, four other systematic reviews and their included articles and citations were checked for inclusion criteria. To be included, articles must have investigated the effects of skin pigmentation on oximetry measurements; used spectrophotometry, colorimetry, or a skin classification system to assess skin pigmentation; compared pulse oximetry to the gold standard ABG measurement by co-oximetry; and reported quantitative or qualitative data on the interaction between skin pigment and oxygen saturations. Articles were subsequently excluded if no English translation was available, if study groups were pediatric or neonatal, focused solely on non-human subjects, did not include primary data (e.g., reviews, editorials), or lacked sufficient methodological detail for assessment.

Screening was done using Covidence (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org, 2024). If there was disagreement between the two authors about study inclusion, a discussion was held, and if consensus could not be reached, a third author (AM) was consulted as the tiebreaker. After screening the 1,125 study titles and abstracts, 144 full-text articles were reviewed for inclusion. A similar approach was taken by the two authors (SC and WCM) with the 144 full-text articles, with a third author (AM) serving as a tiebreaker. The PRISMA diagram, as shown in Table 1 of the search results. Data were extracted by two reviewers (SC and WM) and managed using Covidence and Excel.

Table 1.

Pulse Oximetry and Skin Tone.

Results

Thirty-one studies were included from the search. Eleven additional studies were identified through citation investigation. Ultimately, 42 studies were included in the analysis (Table 1).

Studies from the literature search highlight the various pulse oximeter biases. The studies by Bickler et al. and Feiner et al. highlight this discrepancy (Bickler et al., 2005; Feiner et al., 2007), while Crooks et al. found that such inaccuracies led to delays in critical care interventions (Crooks et al., 2023). Seitz et al. and Gomaa et al. further confirmed the presence of racial bias in pulse oximetry, especially in critically ill patients, where unrecognized hypoxemia can have severe consequences (Gomaa et al., 2023; Kalra et al., 2023; Seitz et al., 2022). Clinical ramifications of these biases include delayed hypoxemia detection and treatment disparities, as demonstrated by Henry et al. and Sudat et al. (Henry et al., 2022; Sudat et al., 2023). Burnett et al. reported that occult hypoxemia was more prevalent among Black and Hispanic patients, raising concerns about potential inaccuracy (Burnett et al., 2022). Factors contributing to inaccuracy include probe technology and calibration errors, as examined by Mendelson et al. and Adler et al. (Adler et al., 1998; Mendelson et al., 1988), with Barker and Wilson showing device-specific variations in bias (Barker & Wilson, 2023). Blanchet et al. also identified inconsistencies among pulse oximeters, emphasizing the need for improved calibration (Blanchet et al., 2023). Errors persisted in clinical settings, as reported by Zeballos and Weisman during hypoxia and exercise (Zeballos & Weisman, 1991) and by Ebmeier et al. in ICU environments (Ebmeier et al., 2018). In addition, Ajizian et al. and Gudelunas et al. observed increased rates of missed hypoxemia under low perfusion conditions in darker-skinned individuals (Ajizian et al., 2023; Gudelunas et al., 2024). To address these issues, Khanna et al. and Bothma et al. advocate for recalibrating pulse oximeters using diverse skin pigmentation models (Bothma et al., 1996; Khanna et al., 2024). Wong et al. and Blanchet et al. emphasize the necessity of regulatory oversight to enhance accuracy across racial groups (Blanchet et al., 2023; Wong et al., 2023). Seitz et al. and Crooks et al. suggest incorporating ABG validation to counteract bias (Crooks et al., 2022, 2023; Seitz et al., 2022). Three studies relied on spectrophotometry to classify skin tone and concluded that, due to pulse oximeter biases at lower SaO2, the risk for occult hypoxemia could increase (Fawzy et al., 2024; Hao et al., 2024; Leeb et al., 2024). Ries and Lellouche grouped patients based on subjective “light” to “dark” and found bias was nearly equal across the studies (Blanchet et al., 2023; Ries et al., 1989). Both Balmaceda et al. and Wiles et al. utilized participant self-identified race and found that SpO2 and SaO2 were poorly correlated overall, with Balmaceda finding only a .814% SpO2 overestimation in Blacks (Balmaceda et al., 2022; Wiles et al., 2022). Using subjective skin scales, Marlar and Mastrototaro both found that while there was some bias in SpO2, the bias was within FDA guidelines (Marlar et al., 2025; Mastrototaro et al., 2024).

While not all studies stated the pulse oximeter device tested or utilized in the study, several studies included in this review specified the brands or types of pulse oximeters used in their analyses. Barker and Wilson (2023) conducted their laboratory study using Masimo pulse oximeters, while Blanchet et al. (2023) compared the performance of multiple devices, including Philips, Nonin, Nellcor, and Masimo, reporting variability in bias by device type. Bothma et al. (1996) tested Simed, Nihon Koden, and Ohmeda probes (ear and finger), demonstrating differing levels of agreement with arterial oxygen saturation. Gudelunas et al. (2024) evaluated Masimo and Nellcor oximeters across FSTs, identifying increasing bias in darker skin tones. Lellouche (2023) similarly compared Nonin and Philips devices, finding that Nonin underestimated and Philips overestimated arterial saturation. Leeb et al. (2024) tested 11 different fingertip oximeters in a controlled laboratory environment and found that five devices failed to meet FDA accuracy standards (ARMS <3%) in individuals with darker skin. Earlier studies also reported on device-specific performance: Mendelson et al. (1988) evaluated the Datascope ACCUSAT, Ries et al. (1989) tested Ohmeda Biox and Hewlett-Packard (HP) models, and Zeballos and Weisman (1991) compared HP and Biox IIA oximeters, observing higher error rates in black subjects under hypoxic conditions. Collectively, these findings underscore that pulse oximeter accuracy can vary not only by skin pigmentation but also by device manufacturer and model.

The reviewed studies consistently demonstrate racial and ethnic discrepancies in pulse oximetry accuracy, particularly underestimating hypoxemia in Black, Asian, and other non-White patients. Bangash et al. found that pulse oximeters overestimated SaO2 more frequently in Black, Asian, and mixed ethnic groups compared to White patients, increasing the risk of occult hypoxemia (Bangash et al., 2022). Fawzy et al. reported similar findings, highlighting delayed COVID-19 treatment eligibility due to overestimated oxygen levels in Black and Hispanic patients (Fawzy et al., 2022). Pilcher et al. confirmed these discrepancies in a multicenter study using the FST scale, showing bias related to skin pigmentation (Pilcher et al., 2020). Schallom et al. found racial differences in accuracy between nasal and forehead sensors (Schallom et al., 2018). Sjoding et al. and Valbuena et al. demonstrated that Black patients were more likely to have occult hypoxemia compared to White patients, with implications for delayed clinical interventions (Sjoding et al., 2020; Valbuena et al., 2022). Wong et al. linked these discrepancies to increased organ dysfunction and mortality risks (Wong et al., 2021). By contrast, earlier studies, such as Jubran and Lee (Jubran & Tobin, 1990; Lee et al., 1993), did not explicitly address racial bias, though Lee noted variability in accuracy among ethnic groups in Singapore. Finally, McGovern et al. and Thrush and Hodges only studied white subjects in a laboratory setting and only noted minimal bias during hypoxic states (McGovern et al., 1996; Thrush & Hodges, 1994). Collectively, these findings underscore the urgent need for improved pulse oximeter technologies that account for racial and ethnic differences to ensure equitable healthcare outcomes.

The research utilized various methods to determine skin pigmentation. In most studies, race and/or ethnicity was collected from the medical record or self-reported by the participant or the surrogate. A total of 17 studies used race or ethnicity as reported by the participant, surrogate, or recorded in the medical record (Balmaceda et al., 2022; Bangash et al., 2022; Burnett et al., 2022; Chesley et al., 2022; Crooks et al., 2022, 2023; Fawzy et al., 2022; Gomaa et al., 2023; Henry et al., 2022; Kalra et al., 2024; Schallom et al., 2018; Seitz et al., 2022; Sjoding et al., 2020; Sudat et al., 2023; Valbuena et al., 2022; Wiles et al., 2022; Wong et al., 2021). Some studies utilized subjective skin classification systems. Five studies utilized the FST (Blanchet et al., 2023; Ebmeier et al., 2018; Gudelunas et al., 2024; Leeb et al., 2024), three studies used the Massey-Martin scale (Barker & Wilson, 2023; Marlar et al., 2025; Wong et al., 2023), and two studies used both the FST and the Monk scale (Hao et al., 2024; Mastrototaro et al., 2024). Two studies used the Munsell color system and then classified participants into “light,” “intermediate,” or “dark” skin tone groups (Adler et al., 1998; Ries et al., 1989). Four studies utilized spectrophotometry to quantify light reflectance (Fawzy et al., 2024; Fong et al., 2024; Leeb et al., 2024). In six studies, the researchers classified the participants as “light” or “dark” pigmented based on observation (Ajizian et al., 2023; Bickler et al., 2005; Feiner et al., 2007; Jubran & Tobin, 1990; Khanna et al., 2024; Mendelson et al., 1988). Three studies enrolled only “lightly pigmented” or white men only (Bouvet et al., 2012; McGovern et al., 1996; Thrush & Hodges, 1994). One study enrolled only Black men (Zeballos & Weisman, 1991), and one study had patients self-identify ethnicity as Chinese, Malay, or Indian (Lee et al., 1993; Zeballos & Weisman, 1991). Sixteen studies enrolled only critically ill patients admitted to the ICU (Balmaceda et al., 2022; Blanchet et al., 2023; Bothma et al., 1996; Bouvet et al., 2012; Chesley et al., 2022; Crooks et al., 2023; Ebmeier et al., 2018; Fawzy et al., 2024; Hao et al., 2024; Henry et al., 2022; Jubran & Tobin, 1990; Lee et al., 1993; Marlar et al., 2025; Schallom et al., 2018; Seitz et al., 2022). Six studies included patients roomed in the hospital but outside the ICU (Adler et al., 1998; Bangash et al., 2022; Burnett et al., 2022; Crooks et al., 2022; Ebmeier et al., 2018; Fawzy et al., 2022). Eleven studies were completed in a pulmonary laboratory or clinic (Ajizian et al., 2023; Barker & Wilson, 2023; Bickler et al., 2005; Feiner et al., 2007; Gudelunas et al., 2024; Mastrototaro et al., 2024; McGovern et al., 1996; Mendelson et al., 1988; Ries et al., 1989; Thrush & Hodges, 1994; Zeballos & Weisman, 1991). Five studies utilized large databases (Kalra et al., 2023; Sjoding et al., 2020; Sudat et al., 2023; Valbuena et al., 2022; Wong et al., 2021). In three studies, the setting in which data were obtained was not reported.

Overestimation of SaO2 saturation by SpO2 in darker-skinned individuals was a consistent finding across numerous studies. Bias varied between devices, with levels ranging from minor (≤1%) to significant (+5%) overestimations depending on the population and device. A few studies identified that occult hypoxemia was more prevalent in Black patients when compared to White patients.

Studies performed in laboratory settings were able to titrate oxygen and other gases to achieve hypoxia and test oximeter accuracy in induced hypoxic states. These studies were performed on healthy volunteers. Bias reported from pulmonary laboratory studies ranged from 0% to 2%, staying within the FDA requirement for use in a healthcare setting. Studies performed in hospital settings or those that used large databases of hospitalized patients varied.

Discussion

This systematic review encompasses all studies through February 2025 that study the bias of pulse oximeters by race, ethnicity, or skin tone. A total of 42 studies were reviewed (Table 1). In November 2023, the FDA stated, “about 66% of SpO2 values were within 2 or 3% of blood gas values and about 95% of SpO2 values were within 4 to 6% of blood gas values, respectively.” However, the FDA goes on to comment that “current scientific evidence from lab desaturation studies suggests that there are some accuracy differences in pulse oximeter performance, especially in lower arterial saturations between lightly and darkly pigmented participants.”

During the height of the COVID-19 pandemic, there were reports that the pulse oximeter failed to accurately assess the saturation in patients who identified as Black or African American (Crooks et al., 2023; Fawzy et al., 2022; Sjoding et al., 2020; Sudat et al., 2023; Wiles et al., 2022). The reports highlighted that this group was at higher risk for unrecognized occult hypoxia, defined as a SpO2 in the normal range with a SaO2 of <88%. This failure of the device has been investigated retrospectively and prospectively in the studies found during this review. In the studies reported here, there is significant reliance on subjective, investigator-reported methods of race or self-reported races, which can introduce variability across the different studies. Objective tools like spectrophotometry were rarely used, but may be a potential area for standardization in future research on skin pigments and pulse oximetry. The studies that reported results by race/ethnicity found that SpO2 to SaO2 bias did vary by race, highlighting the fact that there may be less reporting and/or recognition of hypoxia in Black patients. In general, in patients with SpO2 >90%, the difference between SpO2 and SaO2 is generally less than 2 points (Jubran, 1999).

These reports of bias within the FDA guidelines are reassuring, but there is no reasoning in these studies as to why the bias is greater in some race categories when compared to White. In 2024, the FDA standard required devices to be tested with a minimum of 200 data points evenly distributed over an SaO2 range of 70% to 100% and that devices be tested on 10 or more healthy subjects who vary in age and gender and that the study has subjects with “a range of skin pigmentations, including at least 2 darkly pigmented subjects or 15% of the subject pool” (Goodman, 1988). However, many pulse oximeters were historically calibrated using data from light-skinned individuals. This can lead to less accurate measurements in individuals with darker skin tones because the algorithms may not adequately account for variations in light absorption due to pigmentation.

The authors acknowledge that there have been many published reviews focused on skin tone and pulse oximeter bias. This review is different because it only includes articles that compare the SpO2 to the gold standard SaO2 obtained by ABG sampling. In addition, this review encompasses both pulmonary laboratory and hospital setting studies and shows the differences between induced hypoxic state pulse oximetry in a pulmonary laboratory-induced and clinically relevant hypoxic states of patients admitted to hospitals. This is important because oftentimes perfusion during induced hypoxia remains normal in a healthy adult, while perfusion in a hospitalized patient may be decreased due to various factors such as sepsis, vasopressor administration, or other comorbid states like peripheral vascular disease.

The findings of our study are important for several reasons. First, we have observed that in aggregate, the discrepancy between SpO2 and SaO2 tends to be small, providing some reassurance that pulse oximeters are providing useful information for most patients. Second, there are situations where outliers are observed, such that in some patients a major discrepancy exists between the SpO2 and the SaO2. Third, the existing literature is somewhat flawed for a few reasons, including retrospective designs, non-concurrent measures of SaO2 and SpO2, and inadequate assessment of objective skin color. Thus, further work is clearly needed in this area.

The utility of pulse oximetry has been questioned in part because of an FDA warning regarding the accuracy of the recordings in people with dark skin pigment. Some providers rely on ABG measurement in the ICU because many of these patients have indwelling arterial lines for hemodynamic monitoring; thus, the risk of obtaining the gold standard measurements is minimal. On the other hand, some have questioned the need for indwelling arterial lines because of the availability of non-invasive blood pressure cuffs, which provide a reasonable estimate of arterial blood pressure (Li-wei et al., 2013). In addition, the definition of acute respiratory distress syndrome traditionally required the measurement of PaO2 to calculate the P/F ratio. However, increasingly the use of S/F ratio is being substituted as it provides a reasonable surrogate for the ratio of partial pressure of oxygen to the fraction of oxygen (P/F ratio; Cotton et al., 2022). Thus, the accuracy of SpO2 remains critically important, given the potential for misdiagnosis in this subset of patients.

The COVID-19 pandemic was particularly problematic since many patients were described as having “happy hypoxia” or “silent hypoxemia” (Simonson et al., 2021). The accuracy of SpO2 was critical during this period because some patients with occult hypoxemia may have been misclassified due to inaccurate readings. The COVID-19 pandemic affected communities of color preferentially, at least in some regions of the world, making the issues around skin color particularly important.

Despite our study’s strengths, we acknowledge several limitations. First, our analyses relied on published literature and thus we are limited by the design of the parent studies. Thus, many of the prospective studies were done in healthy participants in pulmonary laboratories, in whom estimation of SaO2 is not typically considered challenging. In addition, retrospectively designed studies did not always have simultaneous measurements of SpO2 and SaO2, which may be problematic with variables that are dynamically changing in critically ill patients. Second, the measurement of skin pigment is best done using spectrophotometry since self-report is subjective and may lead to reporting bias. Semi-quantitative scales, such as the FST, are considered by many to be culturally insensitive and thus better replaced by objective measurements. Third, prospective studies focused on clinically meaningful outcomes may be helpful to draw firmer conclusions (i.e., does aggressive treatment of occult hypoxemia lead to improved outcomes or are such abnormalities clinically inconsequential). Unfortunately, our findings suggest that there were few prospective studies among our 42 included references. Large-scale trials will be required to draw rigorous conclusions. Despite these limitations, we believe that our findings are important and hope that they drive further research in this area. For now, the findings are reassuring and may be clinically directive until more definitive data are available.

Conclusions and Nursing Implications

The variation in pulse oximetry accuracy based on skin pigmentation has significant implications for nursing practice. Nurses frequently rely on SpO2 values to assess oxygenation, guide clinical decision-making, and escalate care. However, overestimation of oxygen saturation in patients with darker skin tones may lead to missed or delayed recognition of hypoxemia, potentially resulting in undertreatment or poor outcomes. Nurses must remain vigilant, especially when caring for African American patients, by integrating additional assessment methods such as clinical signs of hypoxia, ABGs, and patient-reported symptoms to ensure equitable and accurate care. Awareness of device limitations and advocating for more inclusive technology are also essential components of nursing advocacy and patient safety.

Pulse oximetry should be interpreted in the context of patient presentation and symptoms. SpO2 alone may overestimate SaO2 in some patients, and this can be exacerbated by the patient’s race/ethnicity and perfusion. Care should be taken in hospitalized patients when deciding on and executing interventions based on SpO2 in isolation. Finally, it should be noted that while only a few studies in our review looked at occult hypoxia, overall, pulse oximeters become less accurate as the patient’s true SaO2 decreases. This finding underscores the need for inclusive research and development of practices to ensure equitable healthcare outcomes.

Supplementary Material

Supplemental material for this article is available online.

Biographies

Dr. Shannon A. Cotton is a critical care nurse and researcher in the intensive care unit at UC San Diego.

Dr. Jung-ah Lee is a nurse scientist and faculty at UC Irvine with expertise in gerontology and healthcare system. Her research focuses primarily on translational approach to improve care delivery and quality of life for patients and caregivers. Within this focus, a recurring theme of her work is to explore racial and ethnic disparities in quality of care.

Dr Atul Malhotra is a board-certified pulmonologist, intensivist and research chief of Pulmonary, Critical Care and Sleep Medicine at UC San Diego. He is active clinically in pulmonary, critical care and sleep medicine.

Dr. W. Cameron Mcguire completed his pulmonary and critical care medicine fellowship at UC San Diego School of Medicine. Dr. McGuire earned his medical degree from Tulane University School of Medicine. His research interests include pulmonary physiology and pulmonary vascular medicine, and their overlap in the intensive care unit.