HOMERuN Collaborative: Summary of FDA Panel on Racial Bias in Pulse Oximetry |
|
The Hospital Medicine Reengineering Network (HOMERuN) is a rapidly growing collaborative made up of more than 50 Hospital Medicine groups from academic and non-academic hospitals across the United States. |
|
Panel Summary by: David Sterken
|
|
Racial Bias in Pulse Oximetry |
|
The FDA had public meeting about racial bias in pulse oximetry on November 1. A think tank forum on the issue the following day had great speaker presentations. The public meeting lasted 10 hours, and we condense some of the takeaways below, including new and as well as unpublished studies which address some of the issues.
|
|
Some helpful links to resources are below:
|
|
Key Takeaways from the Panel Discussion
1. The panel unanimously agreed that there is racial bias in pulse oximetry and that it is a significant clinical problem. At least one panelist said he was initially not convinced that this problem was real, particularly since manufacturers' presentations spent much of their time criticizing the published literature as being methodologically flawed, but the presentations from experts ultimately did convince him. They also agreed that skin pigmentation is probably the main cause, but that other factors like poor perfusion in sick patients likely contribute and need to be better studied.
2. The panel agreed that objective skin pigmentation measurements need to be used, but they didn't agree on what scale to use. Experts disagree on this, too. There are issues of how big of "buckets" to create, since too few buckets would lump too wide of pigmentation ranges together, but too granular buckets would make them cumbersome and lessen interobserver agreement. The optimal way to measure skin pigmentation was also unclear (e.g., spectrophotometry or something else), since some are very expensive and infeasible and also some would be affected by temporary confounders like erythema of the skin.
3. The panel agreed that the overall expectations of accuracy for pulse oximeters should be tightened, but didn't seem to set a specific threshold. Masimo also recommended that the FDA intensify its requirements for root mean square accuracy from 3% to 2%. For context on how to interpret root mean square accuracy (or "Arms"), "If Arms = 3%, then the probability that an SpO2 value is within 3% of the SaO2 value is roughly 68%." This is significantly less accurate than we assume pulse oximeters to be.
There were some interesting comments on other accuracy issues. For example, many experts mentioned that pulse oximeters might not accurately be measuring the wavelengths of light they say they are which could be a huge source of error. Others mentioned that some companies had gotten their initial approval on a specific probe and monitor, but they then used that approval to market a different probe just because it used the same monitor (i.e., the basic idea being that they swap out FDA-approved components with other components that didn't undergo such testing and yet they do so under the same FDA approval).
Side note, some expert speakers mentioned that the entire 510(k) approval pathway at the FDA is inadequate and there have been known problems for years, so they urged the FDA to overhaul the whole thing.
One comment from the panel chair was that "a rising tide raises all ships," which was meant to imply that intensifying the overall accuracy requirements would also automatically improve the accuracy for patients with dark skin pigmentation even without the FDA specifically targeting accuracy in that population. That statement is probably very slightly true, but is a problematic way to look at it because for many reasons (discussed in the HOMERuN letter), manufacturers could easily meet intensified overall requirements by cherrypicking their study subjects and data without actually improving accuracy for patients for certain racial groups. This seems like a "colorblind" approach to the issue that ignores the need to directly address racism.
4. The panel generally agreed that labeling requirements by the FDA were inaccurate. For example, the FDA leaves it up to manufacturers to decide what factors to mention that can affect pulse oximeter accuracy, even though those factors are the same for all pulse oximeters. As such, only 18.2% of manufacturers listed skin pigmentation as affecting accuracy in their labeling, at least in the recent premarket submissions that the FDA reviewed. However, just listing that skin pigmentation affects accuracy is insufficient, since it's obviously not equitable to just tell Black people, "This device won't work for you and there are no other alternatives that would work for you," and leave it at that. The panel also agreed that the labeling for consumers was inaccurate, which will be addressed in more detail in Point 4 below.
|
|
Key Takeaways Related to HOMERuN Recommendations:
It's not known whether the panelists read the HOMERuN letter, but they did touch on just about everything and for the most part seemed to come to good conclusions.
|
|
1. The panel only briefly touched on specifically performing subgroup analyses, but they did say that large enough numbers of patients with dark skin pigmentation should be included in premarket trials in order for it to be possible to do subgroup analyses on these subgroups that reach the same level of statistic and clinical significance in these subgroups as in the whole study group, which is basically exactly what we recommended. Masimo specifically recommended that each of these subgroups should contain equal numbers of participants in their submitted comments.
They did discuss the distinction between skin pigmentation and race, but many panelists seemed to suggest that by including requirements regarding skin pigmentation that they would also inherently be addressing the issue of race as well. The one Black person on the panel pushed back on this several times mentioning that every time race is ignored in discussions and regulations, that Black people end up suffering as a result, but unfortunately it seemed like her voice was mostly drowned out on this issue. We need to be looking both at objective skin pigmentation and the social construct of race because there are definitely ways that racism itself could be contributing to this bias.
One study illustrates an analogous concept, which showed not just that the prevalence of hypertension is associated with race, but the prevalence of hypertension is associated with experiencing racial discrimination, which suggests that racism itself is causing hypertension. It's certainly possible that racism might be increasing the prevalence of disease in certain racial groups and that these diseases might affect pulse oximetry's accuracy (e.g., peripheral vascular disease, like mentioned in the letter), and furthermore that these diseases are even more prevalent in hospitalized populations and affect perfusion even more in these patients, thus helping to explain why more racial bias is seen in hospitalized patients than in healthy patients. Therefore, the social construct of race may be relevant to the accuracy of pulse oximeters beyond just differences in skin pigmentation.
2. The panel did briefly discuss "race correction factors," but sometimes in the opposite way than hoped. They discussed the potential for a race button on pulse oximeters to tell the device what race the patient is so that the device can adjust accordingly, which is not the way to address this issue and would make the problem worse. Fortunately, others noted that this isn't the way to address the issue, but this point of caution doesn't seem likely to be included in their recommendations to the FDA.
In the think tank forum the following day, this was discussed much more directly. One speaker presented a preprint study that shows that "race correction" would not work for pulse oximetry, which is really important evidence to have. That paper was in part a response to this paper which suggested that maybe we should try race correction factors.
3. Fortunately, the panel definitely seemed to recognize that studies in hospitalized patients need to be required. One person made the analogy that the studies in healthy patients that the FDA currently requires are akin to Phase 1 clinical trials of drugs, but that we really need the equivalent of Phase 3 trials. Exactly how to operationalize this, however, was less clear. Some people pointed out that it could be difficult to obtain IRB approval for studies in hospitalized patients because performing additional ABGs on subjects wouldn't necessarily have a potential benefit to those subjects themselves. As a medical ethics side note, this gets to the principle of beneficence in the Belmont Report.
4. The panel mostly seemed to think that consumer-grade pulse oximeters need to undergo some form of FDA review because consumers are obviously using them for medical purposes. One panelist initially said something along the lines that labeling them as not for medical use was sufficient, but another panelist strongly pushed back, saying that doing so was a copout and not fulfilling the FDA's duty to protect consumers. After that, everyone seemed to agree that consumer-grade devices should be regulated. Some panelists said that consumer devices should be held to the same standards as medical devices, but the majority of panelists thought that these approval requirements should be kept separate because otherwise consumer devices would become so expensive that it would create an access issue wherein some patients couldn't afford them, which is in line with our suggestions in the HOMERuN letter.
5. The panel only briefly discussed the idea of having recalls of noncompliant devices, but some of the public speakers did promote this idea. It is extremely unlikely that the FDA will enforce any sort of recalls regarding this. To be fair, one could guess that essentially every device on the market exhibits some degree of racial bias and obviously we cannot recall all of them, but maybe in a perfect world the worst offenders could be recalled.
6. The panel also only very briefly discussed other devices. Given the FDA's history, it is unlikely that FDA will start proactively looking for racial bias in other optical devices. Therefore, it will likely fall to academia to again reveal such biases and force the FDA to act.
7. Same as 6 above.
8. Fortunately, the FDA has already taken steps to fund independent prospective research in hospitalized subjects. Through its Centers of Excellence in Regulatory Science and Innovation (CERSIs), the FDA is conducting two studies, one in children at Stanford and one in adults at UCSF. Interestingly, Epic is looking into doing retrospective research on racial bias in pulse oximetry through its Cosmos database.
|
|
Further reading and resources:
This is a new website from UCSF's Hypoxia Lab which did some of the pioneering studies into racial bias in pulse oximetry about 15 years ago, and which is now conducting the FDA study above. They will share all their work open source. They also have a database of performance statistics of different pulse oximeters where they independently test their accuracy so you can see how accurate they actually are (spoiler alert: some FDA-approved devices do not actually meet FDA requirements when tested independently). They even have some consumer-grade devices, so you can see how accurate a device is before you buy it for personal use. Unfortunately, they don't have data on Masimo, the manufacturer of our devices at Wisconsin and many other institutions.
This is an article on addressing medical bias from an engineering perspective.
This is a fascinating article describing how the earliest oximeters designed by HP and NASA actually did a good job at accounting for differences in skin pigmentation all the way back in the 1970s. After that, essentially industry shed those features over time, presumably for cost-cutting reasons and because no one in power really cared about racial equity in these devices at the time.
This is the article that arguably started this whole discussion about racial bias in pulse oximetry because it was the impetus for the December 2020 study. The author presented at the think tank forum, and an interesting story that she shared is that this article was going to be published in the New York Times but was then pulled the day before publication because the FDA emailed the New York Times' fact checker and downplayed the possibility of racial bias. Thankfully she was able to get it published elsewhere, otherwise we probably wouldn’t be having these discussions at all.
This is a great article about government policies surrounding medical innovation with specific recommendations for the Biden administration.
|
|
HOMERuN-Affiliated Study Seeking Partners |
|
Roflumilast or Azithromycin to Prevent COPD Exacerbations (RELIANCE) Study |
|
Purpose of RELIANCE:
Both roflumilast and azithromycin have been shown to reduce the risk of COPD exacerbations compared to placebo. However, there has not been a head-to-head comparison of these two FDA-approved medications. RELIANCE is intended to support hospital efforts to reduce the risk of all-cause hospitalization and premature deaths in individuals with COPD.
RELIANCE is Seeking Community-Based Hospitalists:
Hospitalists are critical in the development of post-discharge care plans and medications used by people with COPD. We found from preliminary work that identification of people with COPD while they are hospitalized is an efficient recruitment method for RELIANCE.
Benefits and Compensation:
- $500/year honorarium for being a community partner (paid after registration) plus $200 per patient enrolled.
- Option to participate in clinical roundtables with COPD thought leaders.
- Contribute topic ideas for future grant proposals or publications related to hospitalist care.
- Community Partners will not be investigators / authors, but will be acknowledged in the RELIANCE publication.
Please reach out to [email protected] if you are interested in participating and would like more information about RELIANCE.
|
|
Our next meeting will be on December 9, 2022. |
|
Image Attributions: Vector images from vecteezy.com; Doll4179, CC BY-SA 4.0 via Wikimedia Commons; Adoscam, CC BY-SA 4.0, via Wikimedia Commons |
|
If you would like to join the HOMERuN Collaborative calls, please reach out to [email protected].
|
|
|
|
|
|
|
