Sage Journals: Discover world-class research

Abstract

Background:

Functional arterial hemoglobin oxygen saturation is a critical clinical parameter and the reference for verifying accuracy of S_pO₂ values during performance testing of a pulse oximeter. Variable performance between blood gas analyzers that perform co-oximetry may complicate interpretation of research studies and regulatory guidance. This study quantifies systematic differences between multiple same model analyzers from two widely used brands of co-oximeters.

Methods:

Three sequential experiments were conducted using controlled hypoxemia in healthy adult volunteers according to regulatory clearance criteria for use of functional arterial oxygenation from co-oximetry analysis (sO₂), to simultaneous S_pO₂ values over the range of 70–100%. To account for potential methodological confounders in experiments 1 and 2, a third experiment analyzed 31 arterial samples across multiple same model analyzers for co-oximetry (two Werfen GEM Premier 5000, and three Radiometer ABL90 Flex) in a randomized order after simultaneous preparation. Statistical analysis included paired differences, mixed-effects modeling, and variance analysis to isolate analyzer-specific effects while controlling for confounding variables.

Results:

The GEM analyzers consistently measured sO₂ 2.3% higher on average than ABL analyzers (95% CI 2.01–2.63%, P < .001). This difference was more pronounced at lower saturations (3.2% in the sO₂ 70s range vs 1.4% in the 90s range). Despite strong correlation between measurements (r = 0.999), 96.8% of paired samples differed by >1% and 61.3% by >2%. Within-manufacturer variability was similar between brands (mean absolute difference: ABL 0.38%, GEM 0.47%, P = .55).

Conclusions:

Clinically important systematic differences in sO₂ measurement exist between GEM and ABL blood analyzers that perform co-oximetry, with potential implications for patient care and pulse oximeter verification studies. Larger studies with more co-oximeter brands, coupled with blood tonometry, are needed to better characterize these findings. In the meantime, regulatory standards and reports of pulse oximeter clinical trials should account for these findings and specify which manufacturer’s co-oximeters were used.

Keywords

blood gas analysis oximetry reproducibility of results technology assessment (biomedical)hypoxemia comparative study healthy volunteers monitoring (physiologic)

Introduction

Accurate assessment of blood oxygenation status is paramount in clinical decision-making for disease management, in bedside use of pulse oximetry, and in the development and verification of pulse oximeter performance. Co-oximeter (ie, a spectrophotometric analysis of arterial hemoglobin)-derived functional arterial oxygen saturation (sO₂) is required for regulatory clearances by the United States Food and Drug Administration (FDA) 510(k) guidance process and by the International Organization for Standardization (ISO) standard for manufacturers of pulse oximeters intended for medical purposes (ISO 80601-2-61).^1,2

The reliability of pulse oximetry, particularly its performance across diverse patient populations and skin pigments, is currently a significant area of investigation, underscoring the importance of a dependable reference standard like sO₂.³ However, the accuracy of this standard may be compromised by variations between different commercially available co-oximeters.^4–7 Inter-device variability in sO₂ poses several potential problems: (1) it can lead to potentially different clinical interpretations and management strategies for patients across settings using different co-oximeters; (2) it can undermine the comparability of pulse oximeter device verification studies or clinical trials across different study sites; (3) it can potentially impact regulatory clearance processes.

Despite the critical reliance on sO₂, standardized comparisons of different analyzer platforms using identical samples under rigorously controlled conditions are limited, especially for newer co-oximeters, leaving a gap in understanding the extent and clinical relevance of inter-analyzer differences. This study addresses this gap by conducting a detailed comparison of sO₂ measurements obtained from two commonly used blood gas analyzers that perform co-oximetry, the GEM Premier 5000 (Werfen North America, Bedford, MA, USA) and ABL90 Flex (Radiometer America, Brea, CA, USA).

QUICK LOOK

Current knowledge

Functional arterial hemoglobin oxygen saturation (sO₂) measured by co-oximetry serves as the reference standard for pulse oximeter accuracy verification in regulatory clearance processes and clinical research. Co-oximeters from different manufacturers are widely used interchangeably in clinical practice and research settings. Previous limited studies have suggested potential variability between co-oximeter brands, but comprehensive comparisons using identical samples under controlled conditions are lacking.

What this paper contributes to our knowledge

This study demonstrated a consistent 2.3% difference between GEM 5000 analyzers and ABL90 Flex analyzers across controlled hypoxemia experiments in healthy volunteers, with the GEM 5000 measuring higher. The systematic difference was more pronounced at lower oxygen saturation levels, with 96.8% of paired measurements differing by more than 1% and 61.3% differing by more than 2%. These systematic differences between co-oximeter manufacturers could impact patient management decisions and affect the interpretation of pulse oximeter performance studies, particularly those investigating accuracy across different patient populations.

Methods

Experimental design

The study was approved by the UCSF Institutional Review Board (IRB), and written informed consent was obtained from all subjects. We conducted three sequential experiments to assess systematic differences in sO₂ measurement between two manufacturers (Fig. 1).

Fig. 1.

Experimental design. Three protocols for processing blood samples with multiple co-oximeters are illustrated. Protocol 1 (left): A single sample syringe is analyzed on GEM 1, GEM 2, one ABL and a second ABL, as soon as machines are ready, with an optional measurement using a third ABL. Protocol 2 (center): The sample is aliquoted into four portions and analyzed immediately on four machines (GEM 1, GEM 2, ABL 1, ABL 2). Protocol 3 (right): The sample syringe is held until all machines are ready, then immediately processed by GEM 1, GEM 2, ABL 1, ABL 2, and ABL 3.

Co-oximeters

We used two GEM Premier 5000 [SN22082386 and SN24016212] and three ABL90 Flex Plus [SN092R0477N004, SN092R0392N006, and SN092R0392N007] analyzers. Before data collection, we performed internal calibration, tHb calibration, internal quality control, and external ampule quality control tests on the ABL90s as per the user manual.⁸ In addition, the hemolyzer frequency was checked daily and compared with acceptable ranges. The GEM automatically performed calibration after cartridge insertion (GEM 5K GB/Hct/Lytes/CO-Ox 300 test) and only allowed samples to be run once the process was successful. The sensors within the GEM were automatically monitored as per the protocol outlined in the GEM Premier 5000 user manual, utilizing pre-tonometered process control solutions (IL Performance Verification Product, product number 0024001515).⁹

Blood sampling

After informed consent, arterial blood samples were obtained from healthy adult volunteers undergoing controlled desaturation studies as previously described.¹⁰ These data were collected as part of the UCSF Hypoxia Lab Study: Accuracy of pulse oximeters with profound hypoxia (UCSF IRB# 21-35637). Briefly, the study investigator controlled the delivered oxygen concentration for each participant by adjusting nitrogen, oxygen, carbon dioxide, and room air flow meters via a custom breathing circuit setup to achieve different steady state levels of hypoxemia from 70– 100% sO₂. A stable plateau began once stability was reached (1-min slope of calculated S_aO₂ change <2%). Discard samples were drawn (3 mL, ie, 3 times the volume of the arterial line circuit at the start of a plateau, or 0.5–1 mL between samples within a plateau). Study samples were then drawn with 3 mL syringes (ICU Medical, San Clemente, CA, USA) that contained lyophilized balanced Li-heparin at a concentration of 23.5 IU/mL. Per our standard operating procedure, air bubbles were expelled through the syringe’s vented tip cap, and syringes were gently mixed by inverting and rolling prior to sampling unless otherwise stated below.

Co-oximeter protocol 1

Twenty-one participants underwent controlled staged hypoxemia. Multiple arterial samples were drawn at least 20 seconds apart, at three stable plateaus (S_pO₂ in 90s, 80s, and 70s) (typically 4 samples over about 2 min). All samples were first processed on two ABL analyzers (in 15 cases there was a >0.5% difference in sO₂ between the first two machines, and the sample was run on a third ABL machine per routine lab protocol). When machines were available, after ABL analyses were completed, blood samples were then run on two GEM analyzers. Analyzer internal clocks were not synchronized over the days of the experiment and could not be corrected in analysis, and so timing variables were not included in the analysis.

Co-oximeter protocol 2

We designed a second experiment to quantify the potential impact of sample time-order on sO₂ values in protocol 1. A single study participant underwent controlled staged hypoxemia, and 30 arterial blood samples were drawn over the sO₂ range 70–100%, and each sample syringe was immediately split into four aliquots, each processed on different analyzers (two GEM and two ABL) as soon as each analyzer was available. To ensure coverage across all three ABL machines, we shifted the analyzer pair for each sample (eg, machines 1 and 2, then machines 2 and 3, then machines 3 and 1). Analyzer processing times and delays between draw and analysis were recorded for each sample. In interim analysis, we found that the within-manufacturer variability was higher than expected based on prior experience in our laboratory with the ABL analyzer (typically <0.5%), suggesting that the oxygen content of the syringes could have been impacted by the aliquoting process.

Co-oximeter protocol 3

To address potential impact of the fractionation procedure on sO₂ in protocol 2, we refined our approach. A single study participant underwent controlled staged hypoxemia. Each sample was maintained in a single syringe until all five analyzers (two GEM and three ABL) were simultaneously ready for analysis. When all analyzers were ready, the sample syringe was vortexed for 3 s, then analyzed by all five analyzers in a predetermined random order. Processing times were recorded as in protocol 2. This design minimized both the delay between analyses and potential bias from the analyzer selection order.

Statistical analysis

Descriptive statistics for experiments 1, 2, and 3

Descriptive statistics were calculated for each experiment, including mean bias and pairwise differences within analyzer manufacturers. We calculated the proportion of paired measurements falling within predefined clinical agreement thresholds (±0.5%, ±1%, ±2%). Pairwise comparisons were made by taking the difference between each possible pair of analyzers per manufacturer (eg, GEM1 vs GEM2, ABL1 vs ABL2, ABL1 vs ABL3).

Primary analysis: Paired differences

For each sample in experiment 3, we calculated the mean sO₂ value for each analyzer manufacturer (averaging the two ABL measurements and the two GEM measurements) to create a single paired comparison per sample. The difference was calculated as sO₂_GEM-sO₂_ABL. We used a paired t test to assess whether the mean difference was significantly different from zero. We report the mean difference, 95% CI, P value, and percentage of samples exceeding clinically meaningful thresholds (±1%, ±2%). We also report Bland–Altman plots across all experiments demonstrating the upper and lower limits of agreement.

Within-manufacturer variability in experiment 3 was analyzed by calculating empirical measurement differences between analyzers of the same manufacturer analyzing identical blood samples. We generated statistical summaries of these differences, including mean absolute differences, and determined the percentage of paired measurements exceeding clinically relevant thresholds (0.5%, 1%).

Secondary analysis: Mixed-effects model

A linear mixed-effects model was used to account for all individual measurements in experiment 3 and incorporate potential confounding factors. The model specified sO₂ as the response variable, treated sample_id as a random effect (accounting for correlation within samples), and included fixed effects for analyzer manufacturer, time to analysis, and sO₂ decile. A supplementary model including experiment as a random effect was also performed.

Results

Descriptive statistics for experiments 1, 2, and 3

The three experiments analyzed 288, 30, and 31 blood samples, resulting in 1,165, 120, and 155 total co-oximeter measurements respectively (Table 1). A consistent mean difference between manufacturers was observed across all experiments; GEM analyzers ranged from 2.1% (SD 1.2) to 2.3% (SD 0.8) higher than ABL analyzers. The mean within-manufacturer pairwise differences were also similar between experiments 1 and 3.

Table 1.

Descriptive statistics for experiments 1–3

	Experiment 1	Experiment 2	Experiment 3
Statistic
No. blood samples	288	30	31
No. analyses	1,165	122	155
Mean sample diff. GEM-ABL (SD)	2.3 (0.9)	2.2 (1.1)	2.3 (0.8)
Mean diff. GEM-ABL >90%	1.6	1.5	1.4
Mean diff. GEM-ABL 80–90%	2.8	2.5	2.4
Mean diff. GEM-ABL 70–80%	3.0	2.7	3.2
ABL sO₂ mean (SD)	85.9 (9.5)	83.8 (10.0)	85.5 (9.9)
ABL sO₂ median [min–max]	85.7 [63.4–99.2]	85.0 [66.6–98.9]	87.2 [71.3–98.7]
ABL median analysis time [min–max] (minutes)	—	0 [0–5]	1 [0–2]
ABL mean absolute pairwise diff. (SD)	0.3 (0.3)	1.2 (1.1)	0.4 (0.3)
GEM sO₂ mean (SD)	88.0 (8.8)	86.3 (9.3)	87.8 (9.2)
GEM sO₂ median [min–max]	88.5 [67.6–100.0]	88.2 [69.4–99.1]	89.8 [74.4–99.4]
GEM median analysis time [min–max] (minutes)	—	0 [0–2]	0 [0–1]
GEM mean absolute pairwise diff (SD)	0.5 (0.5)	1.0 (0.8)	0.5 (0.4)

Time data unavailable for Experiment 1 as analyzer clocks were not synchronized.

The median time from first to last analysis was 1.0 min for ABL machines (range 0–2 min) and 0.4 min for GEM machines (range 0–1.2 min). The median time from collection to analysis in experiment 3 was 21.3 min (range 0.6–64 min) for ABL machines and 22.5 min (range 1.3–65 min) for GEM machines. Overall, regardless of machine type, the median time span from first to last analysis per sample was 1.5 min.

The difference in reported sO₂ between analyzer manufacturers varied across oxygen saturation levels (Fig. 2, Supplementary Table S1). In the lower saturation range (70s), the difference was 3.2%, while in the higher range (90s), the difference decreased to 1.4%.

Fig. 2.

sO₂ Measurements by analyzer manufacturer and saturation range (Experiment 3). Boxplots display sO₂ values measured by two blood gas analyzers across three pulse oximeter saturation (S_pO₂) ranges: 70s, 80s, and 90s. Each box represents the interquartile range (IQR), with the median marked as a horizontal line. Whiskers extend to 1.5 × IQR, and individual data points are overlaid. GEM measurements tend to show systematically higher sO₂ values than ABL in the 70s and 80s saturation ranges, with values converging in the 90s.

We observed variation in differences between individual analyzers (Table 2, Supplementary Figure S1). The analyzers with the largest deviation from the overall mean were GEM1/2, with a difference of 1.9% in the 70– 80% saturation range. This highlights that variation may exist not only between analyzer manufacturers but also between individual analyzers from the same manufacturer.

Table 2.

Analyzer-level performance metrics for sO₂ (experiment 3)

Machine ID	70–80%		80–90%		90–100%
Machine ID	Mean sO₂ (SD)	Diff from overall mean	Mean sO₂ (SD)	Diff from overall mean	Mean sO₂ (SD)	Diff from overall mean
ABL 1	72.8 (0.8)	−1.4	86.2 (1.9)	−0.8	96.6 (1.0)	−0.5
ABL 2	72.8 (0.9)	−1.3	85.6 (2.1)	−1.4	96.4 (1.1)	−0.7
ABL 3	73.0 (1.0)	−1.1	86.3 (1.9)	−0.7	96.7 (1.1)	−0.5
GEM 1	76.1 (1.1)	1.9	88.7 (2.0)	1.7	98.0 (0.8)	0.9
GEM 2	76.1 (1.5)	1.9	88.2 (2.1)	1.2	97.9 (0.9)	0.8

Paired differences analysis

Paired t test analysis demonstrated that the 2.3% difference in experiment 3 was statistically significant (95% CI: 2.0% to 2.6%, P < .01). A total of 90.9% of paired measurements differed by more than 1% in the 90–100% range, while none differed by more than 2%. In contrast, 100% of samples differed by 1% in the 70–80% range (Table 3). Despite the systematic difference, GEM and ABL measurements showed a strong correlation (r = 0.99) (Fig. 3), suggesting that while the absolute values differ, the relative changes in sO₂ are similarly captured by both analyzer manufacturers.

Fig. 3.

Correlation of sO₂ Measurements: GEM versus ABL (all experiments). Each point represents a paired sO₂ (%) measurement from the GEM and ABL analyzers, color-coded by experiment). The X axis shows ABL sO₂ (%) and the Y axis shows GEM sO₂ (%). The black dashed line represents the line of identity. The red solid line indicates the linear regression fit across all data points, with the reported Pearson correlation coefficient.

Table 3.

Percentage of paired samples exceeding 1–2% difference

sO₂ range	% Exceeding ± 1%	% Exceeding ± 2%
70–80%	100%	100%
80–90%	100%	90.0%
90–100%	90.9%	0%
Overall	96.8%	61.3%

The Bland–Altman plot (Fig. 4) across all experiments provides a visualization of agreement between the two measurement methods. This plot helps identify any systematic difference (mean difference not equal to zero) and whether the variability of differences is consistent across the measurement range. The Bland–Altman analysis confirms the systematic difference and establishes 95% limits of agreement ranging from 0.5% to 4.2%. This means that for 95% of measurements, we expect GEM readings to fall within this range relative to ABL readings.

Fig. 4.

Bland–Altman plot of sO₂ measurements. Each point represents the difference in sO₂ (%) between GEM and ABL analyzers plotted against the mean of the two measurements. Points are colored by experiment. The black horizontal line indicates zero difference. The red dashed line shows the mean difference. Red dotted lines represent the upper and lower limits of agreement (LoA).

When the same sample was analyzed on different ABL analyzers, the mean absolute difference was 0.38% (range: 0.0%–1.6%) (Fig. 5), while the mean absolute difference between different GEM units was 0.47% (range: 0.0%–2.0%). This difference was not statistically significant (Mann–Whitney U test P = .55). A total of 22.6% of ABL-ABL comparisons versus 32.3% of GEM-GEM comparisons differed by more than 0.5%.

Fig. 5.

Distribution of sO₂ measurement differences within analyzer manufacturers (Experiment 3). Each point represents a matched blood sample measured by two machines of the same type: ABL or GEM. The X axis shows the sO₂ value from one machine in the pair, and the Y axis shows the corresponding sO₂ from the second machine. The black dashed line indicates perfect agreement (line of identity), while the red dotted lines mark a ±0.5% difference threshold.

Mixed-effects model analysis

The mixed-effects model for experiment 3 confirms that the analyzer manufacturer significantly influences sO₂ readings even after accounting for saturation decile and time to analysis. Specifically, GEM read 2.3% higher than ABL (P < .01). In addition, we found that time delay to analysis had a small effect of −0.04% per minute (P = .009) (Supplementary Tables S1 and S2).

Discussion

Our comparison of sO₂ measurements between two common co-oximeters demonstrates a significant difference of 2.3%, a difference that remained consistent across multiple experimental approaches. Differences varied with saturation levels, with a larger difference in the lower range (sO₂ 70s) compared with the higher range (sO₂ 90s). The majority of paired measurements exceeded ±2% difference below 90%, indicating that the systematic difference between analyzer manufacturers could potentially impact clinical decision-making, especially near critical thresholds. The mixed-effects model also identified a small effect of time delay on sO₂ readings.

These findings are generally consistent with prior, more limited, studies that have demonstrated variable performance across co-oximeter manufacturers.^4,5 In 2007, Gehring et al found differences in sO₂ values amongst sets of identical devices amongst five different manufacturers’ devices, as well as an increase in these differences as saturation levels decreased. Gehring found up to a 2.6% difference for sO₂ readings at the 97% level, and up to a ±4% difference at the 75% level, although the mean differences appeared much smaller. In 2012, another report continued to find systematic differences in hemoximeter measurements of sO₂.⁶ Gehring determined that one can “assume that a measurement error also exists even for the gold standard of hemoximetry, and this will influence pulse oximeter calibration.”⁵ Our findings support this assessment. In addition, our findings are in good agreement with previous studies that report a 2.2% difference in sO₂ between the GEM Premier 5000 and Radiometer ABL 800 Flex.⁴

There are several important implications that arise from our findings. First, for analyzer manufacturers and regulatory bodies, these results highlight critical issues with current calibration and verification practices. Co-oximeter manufacturers commonly calibrate analyzers against predicate reference instruments, meaning the selection of the reference analyzer could introduce systematic bias.^8,9 The outcome of regulatory submissions for pulse oximetry instruments may change depending on the reference co-oximeter used, especially if testing facilities use different instruments or if the same analyzers are not used consistently. Blood tonometry has historically been the reference standard, as outlined in Burnett et al’s recommendations to the International Federation of Clinical Chemistry (IFCC).^2,11,12 Tonometry reliably creates fully oxygenated or deoxygenated samples, and has been used to produce samples with intermediate oxygen saturations.¹³ However, tonometry is difficult, time-consuming, and not practical for routine use. Radiometer states in their reference manual that tonometry was the primary reference method for co-oximeter sO₂ readings,⁸ and GEM utilizes pre-tonometered solutions for pO₂ and pCO₂ device verification.⁹ To our best knowledge, there is currently no certified reference material available for hemoglobin fractions as measured by co-oximetry, hindering harmonization of sO₂ values between manufacturers. Therefore, given the current state of potential for differences in sO₂ values between co-oximeter manufacturers, regulatory submissions for new pulse oximeters should require disclosure of which co-oximeter manufacturer and model were used as the sO₂ reference.

Second, a 2% difference can significantly impact clinical care, especially when decisions are threshold-based, such as prescribing or titrating supplemental oxygen, managing mechanical ventilation, or prescribing medications, among other decisions.^14–17

Finally, these findings have important implications for pulse oximetry research, particularly studies investigating the potential impact of factors like skin color on pulse oximeter performance. Most research concerning skin color effects on pulse oximetry has identified small differential effects with dark skin, typically in the range of 1–2% bias. A systematic review of 32 studies by Shi et al found that “pulse oximetry overestimates oxygen saturation in people with a high level of skin pigmentation” with a pooled mean bias of 1.11%, and a CI of 0.29% to 1.93%.³ This wide interval may be confounded by the use of different analyzers within the studies reviewed. It is conceivable that depending on the analyzer used, not only could the magnitude of bias be impacted but also its direction—positive or negative. Specifically, the use of the GEM may minimize the apparent impact of dark skin color on bias and reduce the observed rate of occult hypoxemia. Previous evidence may need reevaluation considering which analyzers were used as reference analyzers, as the choice of reference analyzer could either mask or amplify true differences attributable to patient factors. Our findings regarding co-oximeter measurement differences do not invalidate previous studies demonstrating clinically important systematic differences between oximeters themselves in patient populations.^18,19

This study has several important limitations. Our experimental design evolved across three sequential protocols to address methodological challenges identified during data collection: ABL machines always run before GEM in Experiment 1, and alterations in oxygenation from the aliquoting process in Experiment 2. While all protocols yielded consistent findings regarding inter-manufacturer differences, this adaptive approach limits direct comparability across experiments. In addition, the time from collection to analysis in Experiment 3 was longer than typical; this was due to the requirement that all machines be ready to analyze before starting any analysis. Because of automated intermittent self-QC cycles, all sampling waited until all machines were ready. However, evidence suggests that co-oximetry results may remain stable up to 1–3 hours.²⁰ Moreover, this would not affect cross-device results, and in fact, once analyses were begun, the time from first to last sample was short. Finally, while this study demonstrates differences between these two analyzers, the results cannot necessarily be generalized to other models by the same manufacturers nor across manufacturers. These results do not establish which analyzer provides more accurate measurements, only that there was a difference. Caution should be used in trying to account for this difference across Werfen and Radiometer devices in the clinical setting.

Conclusions

In this study of two co-oximeter models from different manufacturers, we demonstrate a significant difference in sO₂ results. Such differences (ie, choice of blood analyzer that performs co-oximetry) could impact clinical care, as well as conclusions from pulse oximeter performance studies in the literature. We have two major recommendations: 1) Disclosure of the co-oximeter manufacturer and model used as a reference for pulse-oximeter studies should be required for regulatory submission as well as for reports in the medical literature. 2) Manufacturers of co-oximeters should make efforts to harmonize their sO₂ results so sO₂ values between manufacturers are interchangeable.^21,22

Footnotes

Author Disclosure Statement

X.M.R.v.W. received honoraria from Becton, Dickinson and Company (BD) and Abbott Diagnostics, and is a former employee of Beckman Coulter (part of Danaher, owner of Radiometer).

Funding Information

The study was performed at the University of California, San Francisco. There was no specific funding for this study; costs were covered by the Hypoxia Research Laboratory at UCSF.

Supplemental Material

References

Center for Devices, Radiological Health. Pulse Oximeters - Premarket Notification Submissions [510(k)s]: Guidance for Industry and Food and Drug Administration Staff. FDA; 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pulse-oximeters-premarket-notification-submissions-510ks-guidance-industry-and-food-and-drug [Last accessed: Feb 12, 2025].

International Organization for Standardization. Particular Requirements for Basic Safety and Essential Performance of Pulse Oximeter Equipment: ISO; 2017.

Shi

, Goodall

, Dumville

, et al. The accuracy of pulse oximetry in measuring oxygen saturation by levels of skin pigmentation: a systematic review and meta-analysis. BMC Med, 2022; 20(1):267; doi: 10.1186/s12916-022-02452-8

Katar

, Demi̇r

. Gem Premier 5000 Cihazı Ile Radiometer ABL 800 Flex Kan Gazı Analizörlerinin Karşılaştırılması. Gnj, 2023; 8(Özel Sayı):639–648; doi: 10.5281/ZENODO.8402027

Gehring

, Duembgen

, Peterlein

, et al. Hemoximetry as the “gold standard”? Error assessment based on differences among identical blood gas analyzer devices of five manufacturers. Anesth Analg, 2007; 105(6 Suppl):S24–S30; doi: 10.1213/01.ane.0000268713.58174.cc

Brandt

, Opp

, Gehring

. Next Generation Hemoximeter Accuracy during Rapid Desaturation Procedures. In: American Society of Anesthesiology Conference; 2012; doi: 10.13140/2.1.3184.6401

Aoyagi

. Is the Co-oximeter a gold standard?. Anesth Analg, 2002; 94(1S):S101, A1.

Radiometer Medical ApS. ABL90 FLEX Instructions for Use. Brønshøj, Denmark: Radiometer Medical ApS; 2016.

Instrumentation Laboratory. GEM 5000 Premier Operator’s Manual. Bedford, Massachusetts, USA: Instrumentation Laboratory; 2020.

10.

UCSF Hypoxia Lab. Protocol for Basic Pulse Oximeter Hypoxia Performance Study. 2022. https://openoximetry.org/study-protocols/ [Last accessed: Jul 30, 2025].

11.

Burnett

, Covington

, Fogh-Andersen

, et al. Approved IFCC recommendation on definitions of quantities and conventions related to blood gases and pH. Clin Chem Lab Med, 1995; 33(6):399–404; doi: 10.18452/12993

12.

Burnett

, Covington

, Maas

, et al. IFCC document stage 3, draft 1, dated 1989 02 01. An approved IFCC recommendation. IFCC method (1988) for tonometry of blood: reference materials for pCO2 and pO2. International Federation of Clinical Chemistry Scientific Division. Committee on pH, Blood Gases and Electrolytes. Clin Chim Acta, 1989; 185(1):S17–S24; doi: 10.1016/0009-8981(89)90141-1

13.

Balcerek

, Steinach

, Lichti

, et al. A broad diversity in oxygen affinity to haemoglobin. Sci Rep, 2020; 10(1):16920; doi: 10.1038/s41598-020-73560-9

14.

Moore

Jr , Gudelunas

, Lipnick

, et al. Pulse oximeter bias and inequities in retrospective studies–now what? Respir Care, 2022; 67(12):1633–1636; doi: 10.4187/respcare.10654

15.

Brower

, Matthay

, Morris

, et al.; Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med, 2000; 342(18):1301–1308; doi: 10.1056/NEJM200005043421801

16.

Anonymous. IDSA Guidelines on the Treatment and Management of Patients with COVID-19. n.d. https://www.idsociety.org/practice-guideline/covid-19-guideline-treatment-and-management/ [Last accessed: Jun 27, 2025].

17.

Humbert

, Kovacs

, Hoeper

, et al.; ESC/ERS Scientific Document Group. 2022 ESC/ERS Guidelines for the diagnosis and treatment of pulmonary hypertension. Eur Heart J, 2022; 43(38):3618–3731; doi: 10.1093/eurheartj/ehac237

18.

Blanchet

M-A

, Mercier

, Delobel

, et al. Accuracy of multiple pulse oximeters in stable critically ill patients. Respir Care, 2023; 68(5):565–574; doi: 10.4187/respcare.10582

19.

Lellouche

, Bouchard

P-A

, Branson

. Impact of S_pO₂ targets and pulse oximeter brand on oxygen flow requirements and oxygenation. Respir Care, 2023; 69(1):110–113; doi: 10.4187/respcare.11358

20.

Zavorsky

, van Wijk

XMR

. The stability of blood gases and CO-oximetry under slushed ice and room temperature conditions. Clin Chem Lab Med, 2023; 61(10):1750–1759; doi: 10.1515/cclm-2022-1085

21.

Leeb

, Auchus

, Law

, et al. The performance of 11 fingertip pulse oximeters during hypoxemia in healthy human participants with varied, quantified skin pigment. EBioMedicine, 2024; 102:105051; doi: 10.1016/j.ebiom.2024.105051

22.

Bourassa

, Bouchard

P-A

, Dauphin

, et al. Oxygen conservation methods with automated titration. Respir Care, 2020; 65(10):1433–1442; doi: 10.4187/respcare.07240

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

3.56 MB

Differences in Arterial Oxygen Saturation Measurements Between Two Co-Oximeters From Different Manufacturers

Abstract

Background:

Methods:

Results:

Conclusions:

Keywords

Introduction

QUICK LOOK

Current knowledge

What this paper contributes to our knowledge

Methods

Experimental design

Co-oximeters

Blood sampling

Co-oximeter protocol 1

Co-oximeter protocol 2

Co-oximeter protocol 3

Statistical analysis

Descriptive statistics for experiments 1, 2, and 3

Primary analysis: Paired differences

Secondary analysis: Mixed-effects model

Results

Descriptive statistics for experiments 1, 2, and 3

Paired differences analysis

Mixed-effects model analysis

Discussion

Conclusions

Footnotes

Author Disclosure Statement

Funding Information

Supplemental Material

References

Supplementary Material