Pulse Oximetry Performance in Darkly Pigmented Skin: Methodological Concerns

To the Editor

We read with great interest the article by Gudelunas et al,¹ which presented a laboratory pulse oximetry comparison study suggesting that low peripheral perfusion coupled with dark skin pigmentation results in greater pulse oximeter error during controlled hypoxemia. Our careful review identified numerous methodological limitations, which undermine the scientific rigor and render the study conclusions as invalid.

The first methodological error is the failure to standardize the sensor type and anatomic locations studied. The Masimo device was tested using both ear and finger measurements, whereas the Nellcor device was tested using only finger measurements. This introduces an “apples-to-oranges” comparison effect by using 2 different sensor sites for Masimo devices versus 1 sensor site for the Nellcor device. Also, it is well known that the measurement site affects sensor accuracy. For example, the Spo₂ accuracy specification of Masimo DCI reusable finger sensor is 2%, while the corresponding accuracy specification of Masimo’s ear clip sensor is 3.5% A_RMS in no-motion conditions. Importantly, the perfusion index (PI; measure of pulsatile signal) can vary greatly between finger and ear sites. The study did not disaggregate results or elucidate what proportion of measurements were ear versus finger on the Masimo devices.

The second important limitation is the failure to use current pulse oximeter models and sensors. The authors acknowledged that Masimo and Nellcor have newer models; however, the devices selected for evaluation were long out of date (no longer manufactured) pulse oximeter models. In addition, the Nellcor device studied was required to be removed from the market due to patent infringement of Masimo technology. Using outdated devices jeopardizes the validity and relevance to current practice.

The third confounder is the simultaneous data collection from more than 1 pulse oximeter sensor, without documented use of shielding to prevent sensor-to-sensor crosstalk. When multiple pulse oximeters are deployed simultaneously in close proximity, crosstalk can occur, which can cause data degradation.² The solution is to use an opaque optical shield (as per Masimo’s Directions for Use). Numerous publications demonstrating the importance of proper shielding of infrared light to minimize crosstalk in pulse oximetry are discussed in a recent paper by Panicos Kyriacou.²

The fourth concern with this study is use of nonstandardized PI values between the Nellcor and Masimo devices, which could systematically alter results. The Nellcor N-595 does not display a PI value on the device. We speculate that the Nellcor perfusion data was derived from trend data via the serial port. The study methodology divided the Nellcor PI value by 10 in an attempt to make it “numerically comparable” to the Masimo PI value. This is an invalid methodology, as our analysis shows that this practice significantly underestimates the perfusion values for the Nellcor device by a factor of 2 at the lower end of the perfusion spectrum.

The fifth methodological concern is the failure to standardize the hand-warming protocol. Warming of hands and/or forearms was mentioned in the study methodology, but the reader is unable to discern if the warming decisions were applied uniformly for all subjects via a defined protocol, or if warming was selectively applied, and whether this was applied equally to subjects with Masimo versus Nellcor pulse oximeter devices. Furthermore, Masimo sensors placed on the ears did not have the benefit of site warming.

Sixth, and lastly, the study did not consider outlier analysis, or describe how outlier data was treated. In statistical assessment of dynamically changing parameters, outlier treatment is necessary to avoid bias in conclusions. For example, the Masimo, Dark skin plot of pulse oximeter bias appears to include outlier data (potentially from 1 subject) that may be disproportionately skewing the overall results, as shown in the Figure. This outlier data may have been caused by sensor malposition or other data collection errors in the laboratory setting.

Figure.

Pulse oximeter bias (Spo₂-Sao₂) versus Sao₂ (measured arterial blood oxygen saturation) in light (Fitzpatrick I and II), medium (Fitzpatrick III and IV), and darkly (Fitzpatrick V and VI) pigmented human subjects. Regression lines at different level of PI are shown separately (blue: PI <1%; purple: PI ≥1% and <2%; pink: PI ≥2%). Values for the slopes with 95% confidence intervals are shown in the graphs. Data were analyzed by a mixed-effects multivariable model accounting for repeated measures. Example of outlier data encapsulated within oval demarcation. PI indicates perfusion index; Sao₂, arterial functional oxygen saturation; Spo₂, pulse oximeter measured oxygen saturation. Adapted from Gudelunas et al.¹

In summary, the authors’ conclusions are not supported due to the substantial methodological limitations as detailed in this letter. Moreover, Masimo recently published data collected over several years using contemporary RD SET sensors demonstrating clinically equivalent pulse oximetry accuracy in both light- and dark-skinned subjects, including no occult hypoxemia events during desaturation among darkly-skinned subjects.³ Furthermore, Masimo RD SET demonstrates high accuracy across all skin tones, even during low PI conditions.⁴

Vikrant Sharma, PhD
Masimo Corp
Irvine, California

Steven J. Barker, PhD, MD
Masimo Corp
Irvine, California

Jerry Novak, RT, MBA
Masimo Corp
Irvine, California

William C. Wilson, MD, MA
Masimo Corp
Irvine, California
william.wilson@masimo.com