Effects of Skin Pigmentation on Pulse Oximeter Accuracy at Low Saturation

This study was approved by the University of California at San Francisco Committee on Human Research, and informed consent was obtained from all subjects. Twenty-one healthy, nonsmoking subjects were studied. None of the subjects had lung disease, obesity, or cardiovascular problems. They ranged in age between 24 and 47 yr. Eleven subject were very darkly pigmented individuals of African-American descent, and the remaining 10 were light-skinned individuals of northern European descent. No Asians or Hispanics were included in the study. The average age of the light-skinned subjects was 28 yr, and that of the dark-skinned subjects was 30 yr.

Subjects were studied semisupine (30° head up) with a nose clip while deliberately hyperventilating air-nitrogen-carbon dioxide mixtures via  a mouthpiece from a partial rebreathing circuit with 10- to 20-l/min fresh gas inflow. An indwelling 22-gauge radial artery catheter was placed to facilitate arterial blood sampling for measurements of Sao². Five oximeters were mounted on each subject: one Nellcor N-595 with the OxiMax A finger probe (Nellcor Inc., Pleasanton, CA), two model 513s from Novametrix Inc. (Wallingford, CT), and two Onyx models from Nonin Inc. (Plymouth, MN). The Novametrix and Nonin instrument readings were recorded manually. Oxygen saturation measured by pulse oximetry (Spo²) from the Nellcor oximeters, the end-tidal gases, and estimated Sao²were recorded by a computer running LabVIEW 6.0 (National Instruments, Austin, TX). Data from the two Novametrix and two Onyx instruments were averaged separately. Probes were not mounted on the thumbs or little fingers. The Nellcor oximeter was tested in all subjects, and the Novametrix and Nonin instruments were tested in nine light- and seven dark-skinned subjects each.

A series of 10–12 stable target Sao²plateaus between 60 and 100% were sought by an operator who adjusted the inspired air-nitrogen-carbon dioxide mixture breath by breath in response to an analog meter displaying the estimated Sao²derived from mass spectrometer end-tidal gas analysis.17Input parameters for the computer oxygen dissociation curve included arterial pH estimated from end-tidal partial pressure of carbon dioxide, base excess adjusted if needed after each arterial sample was analyzed, and alveolar-arterial oxygen difference as needed especially at low Sao²to attempt to match the predicted with the measured Sao². At each level, arterial blood was sampled after a plateau of 30–60 s had been achieved, followed by a second sample at the same plateau 30 s later. To insure that each subject had good circulation to the fingers, each hand was wrapped in a warming pad. “Functional” arterial Sao²(HbO²/(hemoglobin + HbO²)) was determined by multiwavelength oximetry (Radiometer OSM-3, Copenhagen, Denmark). Quality control standards were run each day.

Bias is computed as Spo²minus Sao²from the reading of each oximeter minus the corresponding blood sample value. Bias is reported as mean ± SD. The relation of bias to Sao²was analyzed by linear regression. Bias was also analyzed for decadal subgroups of Sao²(60–70, 70–80, 80–90, and 90–100%) using analysis of variance. Differences between oximeters were analyzed by analysis of variance and the Tukey-Kramer multiple comparison technique. Multivariate models were used to analyze the interrelations of bias with skin pigment, oximeter, and Sao². P < 0.05 was considered statistically significant. Statistical analysis was performed with JMP 4.0 (SAS Institute, Cary, NC).

Table 1presents all of the data. The mean of all 1,067 data points at all Sao²levels and all three types of oximeters indicated that Spo²reads approximately 1% higher in dark- than in light-skinned subjects (P < 0.0001). These overall mean bias errors due to pigment were +0.4% with Nonin, +0.6% with Novametrix, and +1.6% with Nellcor.

At lower oxyhemoglobin saturation, greater differences in bias between light- and dark-skinned subjects were apparent. With all three instruments, the effect of skin pigment on bias increased approximately linearly as Sao²decreased (fig. 1). In the range of 60–70% Sao², the mean difference in bias between light- and dark-skinned subjects was +1.4% (Nonin), +4.4% (Novametrix), and +4.3% (Nellcor) and 3.2% for all three instruments combined. Dark skin pigment-related bias was statistically significant with Nellcor in all Sao²decades and with Novametrix between 60 and 80%, but only in the 70–80% range with Nonin. Multivariate models analyzing bias with respect to skin pigment, Sao²(either as a continuous variable or by decadal range), and oximeters showed statistically significant relations for all of the variables.

In figures 2A–C, the data obtained with each instrument are plotted against Sao², using filled circles for dark-skinned subjects. The slopes of the regression lines differ most between dark and light skin with Nellcor and least with Nonin. Compared with light-skinned subjects, the mean effect of desaturation on bias with dark skin is ΔSpo²=Ω(100 − Sao²), where Ω is 0.064 with Nonin, 0.123 with Novametrix, and 0.125 with Nellcor. Ω could be considered as a correction factor for clinical use to correct (reduce) low Spo²values in dark-skinned patients. However, the linear equation would overcorrect data at intermediate Sao²with Nonin.

The scatter of data increases at low Sao²as shown in table 1. The SD averages approximately 1.2 in all instruments at high Sao²but increases in the 60–70% range to 2.5 (Nonin), 3.0 (Novametrix), and 3.7 (Nellcor). However, the SD is not greater with dark than light skin at low Sao².

We found small but statistically significant differences between oximeters. When data from light- and dark-skinned subjects were combined, the mean bias of Novametrix (1.40 ± 1.94%) was statistically higher than that of Nellcor (0.49 ± 2.18) and Nonin (0.08 ± 1.45%). Small differences were also found within decadal Sao²intervals.

Pulse oximetry theory predicts that the ratio of the pulsatile to the total transmitted red light divided by the same ratio for infrared light should be dependent only on arterial saturation. In practice, there are many minor deviations from this ideal. Major known variability is caused by anemia, light scattering, venous and tissue pulsation by mechanical force from nearby arteries, pulsatile variations in tissue thickness in the light path other than in the arteries, nail polish, and skin pigment.3Therefore, pulse oximeter design is often based on empirically determined correction factors obtained by in vivo  comparison of oximeter readings with arterial oxyhemoglobin saturation (Sao²) of volunteer subjects. In our 18 yr of testing pulse oximeter accuracy, and probably in other testing laboratories, the majority of subjects have been light skinned. Most pulse oximeters have probably been calibrated using light-skinned individuals, with the assumption that skin pigment does not matter. The current data show that skin pigment introduces a consistent positive bias at low Sao²in the Nonin, Novametrix, and Nellcor instruments. We infer that most pulse oximeters show similar pigment-related bias at low saturation. A previous study of pulse oximeter accuracy in dark-skinned individuals only measured bias at room air saturation levels and therefore concluded that oximeter accuracy is independent of skin color.2 

We chose to study extremes of skin pigment, dark and light, to most easily detect an effect with a small number of subjects. As shown in figures 2A–C, we found a bias ranging from −4 to +8% in dark-skinned subjects and a slightly smaller range in light-skinned subjects. However, as noted above, the SD was not found to be greater in dark-skinned subjects at any Sao²range. This implies that the bias error does not arise simply from inadequate signal intensity and confirms the problem as a real pigment-related difference in pulse oximeter optical factors, which deserves attention and possible provision of correction factors, tables, or even built-in user-optional adjustments. Furthermore, the larger bias in oximeter readings at low saturation did not arise because of data from a few subjects. Scatter plots of data from separate subjects indicate some intersubject variability but no obvious outliers.

Because we tested only three types of oximeters, our results may not apply to pulse oximeters made by other manufacturers. However, all three instruments tested showed increasing positive bias in oximeter readings at low Sao²in subjects with dark skin pigmentation. This may be a property common to a wide variety of modern pulse oximeters, given that the basic technology is similar. In addition, the mean differences in readings between light- and dark-skinned individuals were larger than the differences between the different oximeters. The results of this study call for further investigation of the effect of skin color on oximeter accuracy at low Sao²in other oximeter types. It is also possible that fingers have a different bias than ears or other monitoring sites.

The magnitude of the oximeter error in all three oximeters due to dark skin pigmentation is relatively small at saturations greater than 80% and probably of no clinical significance. However, in individuals with darkly pigmented skin, bias of up to 8% was observed at lower saturations, which may be quite significant under some circumstances. For example, in congenital heart disease, many patients have stable low saturation values, and accuracy in an outpatient setting or during surgery may be desired. In high-altitude medicine, oximeters are frequently relied on for accurate readings in both research and clinical settings.

In conclusion, dark skin pigmentation results in overestimation of arterial oxygen saturation, especially at low saturation in the three tested pulse oximeters. A notice warning of this effect may currently be the best available action.

The authors thank Renchell J. Andres, B.S. (Medical Student, University of California at San Francisco, San Francisco, California), for help with recruiting subjects and data collection.