FDA Pulse Oximeter Accuracy Across Skin Pigmentation: Guidance and 510(k) Clearance Expectations

FDA pulse oximeter accuracy across skin pigmentation: from the 2013 guidance to the 2025 draft

Key points

• Current final guidance (2013): FDA's final 510(k) guidance for pulse oximeters, issued March 2013, recommends a controlled desaturation study with at least 10 healthy subjects and at least 200 paired SpO₂/SaO₂ observations under ISO 80601-2-61:2011 ¹¹³. For skin pigmentation, it asks for a range of pigmentations that includes at least two darkly pigmented subjects or 15% of the subject pool, whichever is larger ¹⁶.
• Proposed approach (January 2025 draft): The draft guidance Pulse Oximeters for Medical Purposes recommends at least 150 participants ⁷⁸. Participants would be stratified by Monk Skin Tone (MST), with skin tone also measured objectively by Individual Typology Angle (ITA) ³²³⁴. Three co-primary success criteria would apply, two of which limit differences in SpO₂ bias across pigmentation ^106107108. The draft covers both prescription and over-the-counter (OTC) devices intended for medical purposes ¹³⁴.
• Status: The January 7, 2025 document is a draft for comment and not for implementation ¹³³. When finalized, it will supersede the 2013 guidance ¹⁴⁰. Among the FDA guidance records reviewed, it is still listed as a draft ²⁶. No final pulse oximeter guidance turned up in FDA press announcements after February 2025 ³⁴⁷.
• Clearance practice: 510(k) summaries for product code DQA cleared from 2024 through 2026 still largely follow the 2013 design: about 12 subjects, about 3 dark-skinned subjects (usually Fitzpatrick V–VI), and a pooled ARMS result ^183184177193. A few also report Monk scores ¹⁷¹¹⁷⁸. None of the summaries reviewed reports an ITA measurement.

The 2013 baseline: what FDA has required in 510(k)s

The final guidance Pulse Oximeters - Premarket Notification Submissions [510(k)s] is dated March 3, 2013 ¹²⁵. For clinical accuracy, it points sponsors to Clause 201.12.1.101.2 and Annex EE.2 of ISO 80601-2-61:2011. It recommends at least 10 healthy subjects of varied age and sex and at least 200 paired pulse oximeter/co-oximeter observations ¹³. Sponsors are asked to document the desaturation profile, the target saturation plateaus, and the criteria used to judge whether reference SaO₂ is stable at the sensor site ¹⁴. The ARMS specification should conform to Clause 201.12.1.101.1 of the same standard ¹⁵.

On skin pigmentation, the 2013 guidance asks for subjects with a range of pigmentations. This includes at least two darkly pigmented subjects or 15% of the subject pool, whichever is larger ¹⁶. FDA later told its advisory panel that this threshold came from U.S. racial census data, not from the population's distribution of skin tones ³¹⁴.

FDA recognizes ISO 80601-2-61, and pulse oximeter 510(k) submissions generally reference it ³⁴⁴.

Why FDA revisited the framework

November 1, 2022 advisory panel

FDA convened the Anesthesiology and Respiratory Therapy Devices Panel of the Medical Devices Advisory Committee on November 1, 2022 (docket FDA-2022-N-2110) ¹⁴². The question was whether pulse oximeters are less accurate in people with darker skin pigmentation ³⁰⁹. FDA's executive summary described studies suggesting missed hypoxemia ("occult hypoxemia"), delayed decisions on treatment eligibility, and worse outcomes in people with darker skin pigmentation ³¹¹. The summary made three cautions:

• Race or ethnicity is an inaccurate proxy for skin pigmentation ³¹².
• Real-world comparisons can be confounded by differences in patients, devices, and the timing of SpO₂ and SaO₂ measurements ³¹³.
• The Fitzpatrick scale was designed to assess susceptibility to sunburn. It does not directly measure skin color or melanin ³¹⁵.

The panel's 24-hour summary reports that the panel judged the real-world evidence to show reduced accuracy in people with darker skin pigmentation ³¹⁹. It recommended:

• adequately sized studies across the full pigmentation spectrum ³¹⁹
• performance measures beyond pooled ARMS ³¹⁹
• prescription labeling that reports accuracy across pigmentation levels ³²⁰
• clear warnings that currently available OTC products are not for medical decision-making ³²⁰

FDA's background section in the 2025 draft also notes that its 2021 safety communication listed skin pigmentation among several factors that may affect readings ³⁶³.

November 2023 discussion paper and February 2, 2024 panel

In November 2023, FDA published a discussion paper for public comment on improving pulse oximeter performance studies. The paper took into account skin pigmentation and patient-reported race and ethnicity ²⁰²³. The panel reconvened on February 2, 2024 (docket FDA-2023-N-4916) ¹⁴¹.

FDA's proposal at that meeting included the following:

• Study size: A desaturation study with at least 24 participants and 480 paired datapoints, replacing the 2013 figures of 10 participants and 200 pairs ²⁶⁷³⁴².
• Scope: The approach would apply to both prescription and OTC devices intended for medical purposes ²⁶⁸²⁶⁹.
• Skin tone assessment: Visual forehead assessment on the 10-point MST scale. At least 25% of participants would come from each of three cohorts (MST 1–4, 5–7 and 8–10). Each MST value would need at least one participant or 15% of its cohort, whichever is greater ²⁷⁰.
• Objective measurement: ITA measured at the emitter site ²⁷¹.
• Non-disparate performance: The maximum absolute difference in mean SpO₂ bias across ITA and MST levels would have to be below 1.5 percentage points when SaO₂ is above 85%. Between 70% and 85%, it would have to be below 3.5 points ²⁷².
• Overall accuracy: ARMS below 3%, demonstrated with statistical significance ²⁷³.

The panel agreed that the 2013 guidance should be revised ²⁷⁴. It supported using MST, with participants spanning at least scores 2 through 9, and objective ITA measurement ²⁷⁴. It was generally in agreement with FDA's definition of non-disparate performance ²⁷⁴. Members questioned whether 24 participants would represent darker skin tones adequately. They also questioned whether measuring near, rather than in, the sensor's light path would capture the relevant pigmentation ²⁷⁵²⁷⁶.

FDA also distinguished two groups of OTC devices at this meeting ³⁴⁵:

• General wellness, sports or aviation oximeters: These generally do not undergo premarket review.
• OTC oximeters intended for medical purposes: These require premarket authorization.

The January 2025 draft guidance

FDA announced the draft on January 6, 2025. It proposes stronger collection and validation of clinical data across skin pigmentations for pulse oximeters intended for medical purposes ¹⁷²⁴. The full title is Pulse Oximeters for Medical Purposes - Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations, dated January 7, 2025 ²⁶¹³³.

Scope

The draft covers pulse oximeters intended for medical purposes, whether prescription or OTC, and whether standalone or part of a multi-parameter device ¹³⁴¹³⁵. It does not cover the sporting, aviation or wellness oximeter product codes it lists ¹³⁶¹³⁷.

Skin tone assessment and enrollment

• Forehead MST: Assess pigmentation visually on the 10-level MST scale. Use colorimetry to measure forehead L* and b* and calculate ITA ³². An appendix covers calibrated printing of MST charts and verifying them with a spectrophotometer ³³.
• Sensor site ITA: Measure ITA on the skin surface that directly contacts the sensor emitter ³⁵.
• Cohort allocation: At least 25% of participants in each of MST 1–4, 5–7 and 8–10. At least 50% of the MST 8–10 group should have a forehead ITA of −50° or lower. Each MST group should be at least 40% male and at least 40% female ³⁴.

Sample size

The draft recommends at least 150 participants, sized for at least 80% power on each success criterion ⁷⁸. Each participant should contribute an average of 20 to 24 paired SpO₂/SaO₂ points, with a minimum of 17 and a maximum of 30 ⁷⁸. At least 30% of pairs should fall in each of the three SaO₂ decades ⁷⁸.

Success criteria

The draft proposes three co-primary criteria, each to be demonstrated with statistical significance (a one-sided test at 2.5%, or a two-sided 95% confidence interval) ¹⁰⁸¹⁰⁹:

1. Overall accuracy: ARMS below 3% across all participants ^103104105.
2. Bias across forehead MST groups: In pairwise comparisons of MST 1–4, 5–7 and 8–10, the largest difference in SpO₂ bias must be below 3.5% at SaO₂ 70% to 85%. It must be below 1.5% at SaO₂ above 85% up to 100% ¹⁰⁶.
3. Bias across sensor site ITA: For a 100-point change in emitter-site ITA, the difference in SpO₂ bias must stay below the same 3.5% and 1.5% limits ¹⁰⁷.

The subgroup criteria limit differences in bias. They are not per-subgroup ARMS limits ¹⁰⁶.

Labeling

• Prescription devices: Report ARMS overall and by SaO₂ decade (70–80%, 80–90%, 90–100%) ¹¹⁰¹¹¹. Report bias by forehead MST group and by sensor-site ITA ².
• OTC devices: Use plain-language directions that address normal ranges, when to seek medical attention, and how to interpret fluctuating readings ¹¹⁵. Summarize the desaturation study, including participant demographics, overall ARMS and accuracy by SaO₂ range ¹¹⁵. FDA also recommends usability testing, such as labeling comprehension testing ¹²⁷.
• Devices that demonstrate non-disparate performance: Include a prominent statement that the device was evaluated to perform comparably across a wide variety of skin tones, supported by clinical data ¹²⁸³⁶⁰.

Previously cleared devices and modifications

For pulse oximeters already cleared through a 510(k), the draft recommends that manufacturers gather clinical data to evaluate non-disparate performance, if they have not already done so ³⁵⁸³⁵⁹. These data may come from a controlled desaturation study or from real-world data. Manufacturers would submit the data in a new 510(k), with revised labeling and the 510(k) summary ^358359360.

The draft also gives examples of changes likely to require a new 510(k) ³⁵⁷:

• labeling based on non-disparate-performance data
• significant changes to the sensor
• significant changes to the SpO₂ algorithm or SpO₂ software inputs

How the expectations compare

What recent 510(k) clearances show

As long as the 2013 guidance remains the final guidance, recent DQA 510(k) summaries mostly reflect its structure, though some sponsors have started to add Monk data.

Older summaries show how pigmentation reporting has developed. In 2016 and 2018, Beijing Choice's MD300-series summaries counted subjects by pigmentation group, for example three subjects with "dark pigmentation" among 12 ⁶⁷⁶⁹. Other pre-2020 summaries described subjects only as "light to dark," with no count of darkly pigmented participants ⁷⁴⁷⁶⁷⁷.

Across the recent summaries, the subgroup sizes are small (three to six darkly pigmented subjects), and pooled ARMS remains the headline metric. That is the pattern the 2022 panel criticized when it called for measures beyond pooled ARMS ³¹⁹. Results are still reported mostly against the Fitzpatrick scale. Under the 2025 draft, the evidence would instead rest on MST- and ITA-based bias-difference testing in a much larger population ^78106107.

Practical implications for sponsors

• New prescription and OTC medical-purpose oximeters: The 2013 guidance remains the final reference ²⁵. The 2025 draft shows where FDA expects clinical evidence to go: about 150 or more participants, MST/ITA stratification and statistical tests of non-disparate bias ^7834106107. Sponsors planning new desaturation studies have a reason to design against the draft criteria and to discuss the protocol with FDA.
• Previously cleared devices: The draft anticipates new 510(k)s that carry non-disparate-performance data and the related labeling claims ³⁵⁸³⁶⁰. Changes to the sensor or the SpO₂ algorithm are flagged as likely to require a new 510(k) ³⁵⁷.
• Wellness, sports and aviation oximeters: These products remain outside the draft's scope ¹³⁶¹³⁷. FDA distinguishes them from medical-purpose OTC devices by intended use ³⁴⁵.
• Consumer messaging: FDA continues to publish consumer-facing material. In March 2025 it issued a Pulse Oximeter Basics Consumer Update on safe and accurate use, including at home ³⁴⁸.

Open questions for further research include how FDA will resolve public comments on the 150-participant design, how real-world data could support non-disparate-performance claims for legacy devices, and whether any 510(k) has yet been cleared with an MST/ITA-based non-disparate-performance labeling statement.