Accuracy of Pulse Oximeters to Get a Close Look From FDA Advisors

An FDA advisory committee on Tuesday will consider whether pulse oximeters give accurate results for all patients, including those with darker skin.

"As part of its premarket evaluation of pulse oximeters, FDA has long required premarket data assessing pulse oximeter performance in subjects with different skin pigmentation," FDA staff members noted in a briefing document posted prior to the meeting of the Anesthesiology and Respiratory Therapy Devices Panel of the Medical Devices Advisory Committee. "However, a December 2020 published report suggests that pulse oximeters may be less accurate in patients with darker skin pigmentation."

"Since the December 2020 publication, additional real-world studies have been published suggesting increased risk for missed diagnosis of hypoxemiavii (i.e., 'occult hypoxemia'), delays in treatment eligibility decisions and worse patient outcomes among subjects with darker skin pigmentation," the report continued.

One recent study of more than 3,000 patients in the intensive care unit found that average oxygen delivery rates were lower for Asian (coefficient -0.291, P=0.03), Black (coefficient -0.294, P=0.001), and Hispanic (coefficient -0.242, P=0.03) patients compared with white patients, in large part due to poorer performance of pulse oximeters. Another study showed that pulse oximetry overestimated oxygen saturation by more than 1 percentage point in minority groups with COVID-19, leading to delayed recognition of eligibility for treatment.

FDA staff noted that several factors can impact the accuracy of pulse oximeters, including:

• Skin pigmentation
• Dyshemoglobinemias
• Severe anemia
• Low perfusion
• Dyes
• Nail polish
• Ambient light
• Motion artifacts

In the briefing document, they noted that in its guidance for pulse oximeter manufacturers who are seeking approval for their devices, "the FDA ... recommends that desaturation studies include 10 or more healthy subjects that vary in age and gender, include 200 or more data points (i.e., paired observations of SpO2-SaO2), and for the study subjects to have a range of skin pigmentation, including at least 2 darkly pigmented subjects or 15% of the study group, whichever is larger. The recommendation for distribution of darkly pigmented subjects was based on census data for distribution of race in the U.S. and not based on the distribution of skin tone of the general U.S. population."

Because of the concerns about pulse oximeter accuracy in darker-skinned populations, FDA staff "performed an initial examination of prior cleared 510(k) submissions ... to better understand skin pigmentation assessments and reporting in regulatory submissions. The evaluation examined premarket clinical study characteristics such as reporting of skin pigmentation classification and reporting of factors potentially impacting device accuracy (e.g., skin pigmentation) in device labeling."

Staff members chose 44 submitted 510(k) applications from a pool of 420 submissions that occurred from Jan. 1, 2000 through Dec. 31, 2020. "Submissions were excluded if clinical data were not required, the device was not a pulse oximeter, the device was for pediatric use only, or if the data was not relevant," the report noted.

The staff found that "submissions used inconsistent methods for categorization, with many of these classifications lacking clear and standardized definitions. Specifically, among the clinical studies, 2 (4.5%) indicated subjects with darker skin pigmentation but did not report skin pigmentation data. For others, 8 (18.2%) used 2 categories, 13 (29.5%) used 3 categories, 6 (13.6%) used 4 categories, 1 (2.3%) used 5 categories, 1 (2.3%) used the Von Luschan scale, 1 (2.3%) used the Fitzpatrick scale. Of note, 12 (27.3%) studies did not specify skin pigmentation categories." In all, 32 of the 44 submissions, or 72.7%, reported race and ethnicity data for participants.

"While any differences in optical based medical device performance correlated with race/ethnicity could potentially be caused by a variety of factors, skin pigmentation is typically considered the most likely cause," they wrote. "This is due to the strong, spectrally varying optical absorption of epidermal melanin, and evidence that variations in skin pigmentation have a significant impact on visible to near-infrared reflectance signals, and thus impact the performance of optical devices (e.g., regional oximeters, bilirubinometers)."

"Further complicating the ability to differentiate patient populations based on skin pigmentation is the variance in skin pigmentation at different anatomical sites," they continued. "The most common pulse oximetry site is the finger -- specifically the fingernail and distal palmar finger, which contain minimal melanin, especially when compared to anatomical sites with high pigmentation levels such as the arm or torso. Furthermore, the fingernail bed has a melanocyte content that is approximately 5% that of normal skin and, unlike melanocytes in the skin, these cells do not produce melanin."

FDA staff also pointed out that "there appears to be mounting real-world evidence that pulse oximeter accuracy varies by self-reported race, and skin pigmentation. However, there is a need for prospective studies that utilize standardized measurement of skin pigmentation, capture simultaneous measurement of SaO₂ and SpO₂ paired data and systematically collect data on important confounders, to have more robust evidence about the impact of skin pigmentation on real-world pulse oximetry."

At the meeting, committee members will be asked to discuss the clinical evidence from the scientific literature about the accuracy of pulse oximetry among patients with darker skin pigmentation and address whether the currently available clinical evidence demonstrates disparate performance in patients with darker skin pigmentation, as well as whether they believe such disparate performance may lead to increased risks. They will also be asked to discuss how accurate pulse oximeters need to be, and to give their recommendations for studies evaluating the devices.

Comment