Accuracy of Oxygen Saturation Measurements in Patients with Obesity Undergoing Bariatric Surgery

Abstract

Background

We aimed to determine the magnitude, direction, and influencing factors of the concordance between arterial oxygen saturation (SaO₂) and peripheral capillary oxygen saturation (SpO₂) in patients with obesity undergoing bariatric surgery, supporting the measurement of SaO₂ and SpO₂ in key populations.

Methods

Patients with obesity undergoing bariatric surgery from 2017 to 2020 were included. Preoperative SpO₂ and SaO₂ were collected. Linear correlation and multiple linear regression analyses were performed to characterize the relationships between body mass index (BMI), age, and sex with pulse oximetry and arterial blood gas (ABG) parameters. Bland–Altman analysis was applied to determine the concordance between SpO₂ and SaO₂ and the limits of this concordance.

Results

A total of 134 patients with obesity undergoing bariatric surgery were enrolled. SaO₂ was negatively associated with BMI (p < 0.0001) and age (p = 0.006), and SpO₂ was negatively associated with BMI (p = 0.021) but not with age. SpO₂ overestimated SaO₂ in 91% of patients with a bias of 2.05%. This bias increased by 203% in hypoxemic patients compared with nonhypoxemic patients (p < 0.0001). The bias was 1.3-fold higher (p = 0.023) in patients with a high obesity surgery mortality risk score (OS-MRS) than in those with low or intermediate scores.

Conclusion

Compared with SpO₂, preoperative SaO₂ can more accurately reflect the real oxygen saturation in patients with obesity undergoing bariatric surgery, especially for those with BMI ≥ 40 kg/m², age ≥ 40 years, and high OS-MRS. ABG analysis can provide a more reliable basis for accurate and timely monitoring, ensuring the perioperative safety of susceptible patients.

Introduction

Obesity is associated with numerous chronic diseases [1–5]. It has reached pandemic proportions, and related metabolic disorders have become major health concerns in many developed and developing countries [3].

Bariatric surgery remains the most effective means of inducing long-term weight loss and improvements in or remission of obesity-associated comorbidities [6–8]. Patients with obesity require additional perioperative care, particularly with regard to anesthesia management. Alterations in metabolic demand and physical and respiratory conditions render these individuals highly susceptible to hypoxia [9–12]. Although there have been some strategies and guidelines [13, 14] aimed at improving oxygen saturation in patients with obesity undergoing surgery, which monitoring method is suitable for such patients has not yet been clearly stated.

Arterial oxygen saturation (SaO₂) measurements performed by arterial blood gas (ABG) testing and peripheral capillary oxygen saturation (SpO₂) measurements performed using pulse oximetry are vital clinical indicators of blood oxygenation. SaO₂ measurement is regarded as the gold standard, while invasive arterial blood sampling is required [15]. Pulse oximetry is commonly used because it is non-invasive and provides convenient and continuous measurement of blood oxygenation; however, the accuracy of SpO₂ is affected by various conditions, such as hypothermia, vasoconstriction, hypoxemia, and so on [16–18]. Previous researches confirmed that SpO₂ under- or overestimates SaO₂ in certain conditions, making it doubtful as a reliable surrogate for the measurement of absolute oxygen saturation [19–22]. In clinical practice, SpO₂ seems to be inconsistent with SaO₂ in patients with obesity, with an overestimate favoring SpO₂ compared with SaO₂. Timely and accurate monitoring of the actual oxygen saturation prompts an optimal preoperative intervention, thus ensuring the safety of these patient populations. Therefore, verifying the discrepancy between SpO₂ and SaO₂ in various patients with obesity undergoing bariatric surgery would be optimal essential.

However, controversy remains regarding the accuracy of pulse oximetry probes and the lack of comparison of these two methods in patients with obesity undergoing bariatric surgery. We hypothesize that SpO₂ inaccurately reflects the real oxygen saturation in these patient populations. Hereby, by conducting a retrospective study, we aimed to determine whether SpO₂ accurately reflects SaO₂ by determining the discrepancies in SaO₂ and SpO₂ in such patients and to explore the factors that affect the discrepancies.

Methods

This study was approved by the ethics committee of the First Affiliated Hospital of Jinan University (Ethics Approval No. KY-2019–023). Each participant or their legal guardian provided written informed consent. Patients with obesity undergoing bariatric surgery from 2017 to 2020 were included. The inclusion criteria were as follows: (1) the participants were 15–65 years of age, (2) had a BMI ≥ 30 kg/m², (3) had an American Society of Anesthesiologists (ASA) Physical Status Classification score of II or III, and (4) had no contraindications to general anesthesia. The exclusion criteria were as follows: (1) the participants had a severe genetic disease, (2) had heart failure or respiratory failure, (3) were positive in the modified Allen’s test, (4) had a local infection of the forearm, or (5) received oxygen therapy after admission.

A power calculation determined that a sample size of 36 was required to detect a 2% difference between SpO₂ and SaO₂ with a power of 0.8 and α = 0.05. In total, 138 patients were thoroughly screened. Excluded patients included 3 patients received oxygen therapy, and 1 patient diagnosed with Prader-Willi syndrome. Finally, 134 patients were recruited, which would be sufficient to identify a difference.

Diagnostic Criteria

Patients were diagnosed with hypoxemia when PaO₂ was below 80 mmHg as previously defined [23]. One hundred thirty-four patients were scored using the obesity surgery mortality risk score (OS-MRS) system [24, 25]. One point was gained for each of the following five factors: age > 45 years, hypertension, male sex, history of pulmonary embolism, and BMI > 50 kg/m². The OS-MRS system has 0 as the lowest risk score and 5 as the highest. On this basis, the participants were allocated to two groups: the first contained those with a low (L, score = 0) or intermediate (I, score = 1 or 2) risk score, and the second contained those with a high-risk score (H, score = 3, 4, or 5). Pulse oximetry and measurement of arterial oxygenation were detailed in the Supplemental material.

Statistical Analysis

Statistical analyses were performed using SPSS 24.0 (IBM, Inc., Armonk, NY, USA). Data are summarized as the means ± standard deviations (SDs). Linear correlation analysis was used to characterize the relationships of BMI, age, and sex with SpO₂ and ABG parameters. Multiple linear regression was used to determine multiple testing-corrected p values. Bland–Altman analysis was used to determine the concordance between SpO₂ and SaO₂ and the 95% confidence interval (95% CI). The Mann–Whitney rank-sum test was used to identify differences between normally distributed data with nonhomogeneous sample variances. The chi-square test or Fisher’s exact test was used to analyze categorical data. A p < 0.05 was considered statistically significant.

Results

Patient Characteristics

We recruited 134 patients with obesity (82 women and 52 men). The baseline characteristics of all participants are shown in Table 1. The mean ± standard deviation (SD) of age, BMI, SaO₂, and SpO₂ was 30.66 ± 8.72 years, 41.86 ± 9.68 kg/m², 96.48 ± 2.52%, and 98.53 ± 2.00%, respectively. The participants were recruited randomly, and the ages of the participants in the female and male groups were not matched. However, there were no differences between sex with regard to SpO₂ and the difference between SpO₂ and SaO₂, except for SaO₂, PaO₂, PaCO₂, and pH. The distributions of BMI and age in the female and male participants are shown in Supplemental Fig. S1.

Table 1.

Patient characteristics

	Total (n = 134)	Males (n = 52)	Females (n = 82)
	M ± SD	M ± SD	M ± SD
Age (years)	30.66 ± 8.72	28.58 ± 7.93	31.99 ± 8.99 *
Weight (kg)	117.11 ± 31.30	137.91 ± 32.97	103.93 ± 21.70****
BMI (kg/m2)	41.86 ± 9.68	45.58 ± 10.86	39.50 ± 8.07 ***
SaO2 (%)	96.48 ± 2.52	95.94 ± 2.86	96.83 ± 2.24**
SpO2 (%)	98.53 ± 2.00	98.15 ± 2.57	98.77 ± 1.49
SpO2–SaO2 (%)	2.05 ± 1.52	2.21 ± 1.31	1.94 ± 1.64
PaO2 (mmHg)	83.13 ± 8.97	80.56 ± 8.99	84.77 ± 8.61 **
PaCO2 (mmHg)	40.44 ± 5.60	42.30 ± 7.38	39.26 ± 3.70 **
pH	7.40 ± 0.030	7.39 ± 0.029	7.41 ± 0.029 *

^*p < 0.05,

^**p < 0.01,

^***p < 0.001,

^****p < 0.0001.

BMI, body mass index. SaO₂, arterial oxygen saturation. SpO₂, peripheral capillary oxygen saturation. PaO₂, partial pressure of oxygen. PaCO₂, partial pressure of carbon dioxide. M, mean. SD, standard deviation.

Linear Correlation Analyses of the Relationships Between SaO₂ or SpO₂ and BMI, Age, PaO₂, and PaCO₂

BMI showed a negative correlation with SaO₂ (r = − 0.347, p < 0.0001) and PaO₂ (r = − 0.414, p < 0.0001) (Fig. 1A, Supplemental Fig. S2A). BMI had a negative correlation with SpO₂ (r = − 0.239, p = 0.006), the difference between SpO₂ and SaO₂ (r = 0.262, p = 0.002) (Fig. 1B–C), PaCO₂ (r = 0.299, p = 0.0004), and pH (r = − 0.208, p = 0.016) (Supplemental Fig. S2B–C).

Fig. 1.

The relationships between SaO₂, SpO₂, and SpO₂ − SaO₂ and BMI and age in patients with obesity. Linear correlation analyses of the relationships of SaO₂ (A, D), SpO₂ (B, E), and the difference between SpO₂ and SaO₂ (C, F) with BMI (A–C) and age (D–E). BMI, body mass index. SaO₂, arterial oxygen saturation. SpO₂, peripheral capillary oxygen saturation

Similarly, SaO₂ was negatively correlated with age (r = − 0.199, p = 0.021), whereas SpO₂ was not correlated with age (r = − 0.086, p = 0.325) (Fig. 1D–E). However, the difference between SpO₂ and SaO₂ (r = 0.217, p = 0.012) (Fig. 1F) and PaO₂ (r = − 0.220, p = 0.011) (Supplemental Fig. S2D) was correlated with age. There were no correlations between age and PaCO₂ or pH (Supplemental Fig. S2E–F).

Taking gender into account, linear correlation analyses of the relationships between SaO₂ or SpO₂ and sex, PaO₂, and PaCO₂ were displayed in Supplemental material.

The Difference Between SpO₂ and SaO₂

Bland–Altman analysis was performed to display the degree of bias and the limits of the concordance between SpO₂ and SaO₂. Across the full cohort, the analysis showed a bias of 2.05% (95% CI, − 0.943, 5.025) (Fig. 2A). In hypoxemic patients, the bias increased to 3.08% (95% CI, − 0.145, 6.309) (Fig. 2B). In contrast, the bias was only 1.52% (95% CI, − 0.722, 3.765) in nonhypoxemic patients (Fig. 2C).

Fig. 2.

Bland–Altman plots for the differences between SpO₂ and SaO₂. The differences between SpO₂ and SaO₂ in all the participants (A), hypoxemic participants (B), and nonhypoxemic participants (C) were plotted against the average SpO₂ and SaO₂. Red lines indicate the level of bias. SaO₂, arterial oxygen saturation. SpO₂, peripheral capillary oxygen saturation. CI, confidence interval

There was an insignificant increase (p = 0.067) in the difference between the bias in the obese (30 kg/m² ≤ BMI < 40 kg/m²) and severely obese (BMI ≥ 40 kg/m²) groups (Fig. 3A), although the bias correlated closely with BMI (Fig. 1C). However, the mean and limits of the bias were higher in the severely obese group (Fig. 3A). The biases did not differ between participants of the two sexes (Fig. 3B). The bias in participants ≥ 40 years of age was 1.38-fold higher than in those < 40 years of age (p = 0.054) (Fig. 3C), and the bias in hypoxemic patients was 2.03-fold higher (p < 0.0001) than that in nonhypoxemic patients (Fig. 3D). The biases and 95% CI for the subgroups are presented in Table 2.

Fig. 3.

Comparison of the differences between SpO₂ and SaO₂ in subgroups of participants categorized according to BMI, sex, age, and hypoxemic status. A–D Box plots of the differences between SpO₂ and SaO₂, according to BMI (A), sex (B), age (C), and hypoxemic status (D). E–H The percentages of participants with a bias ≤ 2% versus those with a bias > 2%, according to BMI (E), sex (F), age (G), and hypoxemic status (H). BMI, body mass index. SaO₂, arterial oxygen saturation. SpO₂, peripheral capillary oxygen saturation

Table 2.

Comparisons of the key parameters in the two OS-MRS groups

	L + I (n = 102)	H (n = 32)	p
	M ± SD	M ± SD
Age (years)	29.66 ± 8.05	33.88 ± 10.06	0.029
Weight (kg)	107.7 ± 21.71	147.1 ± 38.10	< 0.0001
BMI (kg/m2)	39.31 ± 6.87	49.99 ± 12.61	< 0.0001
SaO2 (%)	96.88 ± 2.01	95.22 ± 3.46	< 0.0001
SpO2 (%)	98.79 ± 1.35	97.69 ± 3.20	0.081
SpO2–SaO2(%)	1.91 ± 1.53	2.47 ± 1.43	0.023
PaO2 (mmHg)	84.94 ± 8.21	77.37 ± 8.98	< 0.0001
PaCO2 (mmHg)	39.50 ± 3.65	43.41 ± 8.91	0.005
pH	7.40 ± 0.028	7.39 ± 0.032	0.032

L + I, low (L, score = 0) and intermediate (I, score = 1 or 2) OS-MRS group.

H, high (H, score = 3, 4, or 5) OS-MRS group.

OS-MRS, obesity surgery mortality risk score. BMI, body mass index. SaO₂, arterial oxygen saturation. SpO₂, peripheral capillary oxygen saturation. PaO₂, partial pressure of oxygen. PaCO₂, partial pressure of carbon dioxide. M, mean. SD, standard deviation.

SpO₂ overestimated SaO₂ in 91% of the participants (122/134). The percentage of participants with over- or underestimated oxygen saturation (when SpO₂ − SaO₂ > 2%) who were severely obese (55.7%) was higher than in those who were obese (41.1%) (Fig. 3E). Similarly, participants with an SpO₂ − SaO₂ bias > 2% were more common (62.5%) in the age ≥ 40-year group than in the age < 40-year group (45.8%) (Fig. 3G). Male patients tended to show biases of > 2% more frequently than female patients (Fig. 3F), but this difference was not significant. The percentage of participants with > 2% SpO₂ − SaO₂ bias who were hypoxemic (73.3%) was significantly higher than that in participants who were not hypoxemic (34.8%) (Fig. 3H).

Relationship Between the Difference in SaO₂ and SpO₂ and OS-MRS

We calculated the OS-MRS for those population (Table 2). Compared with the low or intermediate risk score (L + I) group, the difference between SpO₂ and SaO₂ (p = 0.023) increased 1.29-fold and was significantly higher in the high-risk score (H) group than in the L + I group. In addition, SaO₂ (p < 0.0001), SpO₂ (p = 0.081), PaO₂ (p < 0.0001), and pH (p = 0.032) were lower in the H group than in the L + I group, and PaCO₂ (p = 0.005) was significantly higher in the H group. Pearson correlation coefficient analyses showed that OS-MRS negatively correlated with SpO₂ (p < 0.0001) and SaO₂ (p < 0.0001).

Discussion

The present study provides an improved understanding of the magnitude and direction of the inaccuracy of pulse oximetry compared to ABG oxygen saturation measurements in patients with obesity. The data show that the discrepancy between SpO₂ and SaO₂ increases as BMI, age, the degree of hypoxemia, and OS-MRS increase, and compared with SpO₂, preoperative SaO₂ can more accurately reflect the real oxygen saturation in patients with obesity undergoing bariatric surgery, especially for those with a BMI ≥ 40 kg/m², age ≥ 40 years, and high OS-MRS. This means that simple pulse oximetry suffices for younger, lighter, and low or intermediate OS-MRS patients, while for older, heavier, or high OS-MRS individuals, more invasive blood gas monitoring is necessary. Therefore, this study supports the application of SaO₂ and SpO₂ in their suitable populations, stressing the use of ABG measurements in certain patients undergoing bariatric surgery to provide a reliable basis for accurate and timely monitoring and protection of patients’ perioperative safety.

We have shown that SaO₂ is inversely correlated with BMI, consistent with the findings of several previous studies [26–28]. Peppard et al. reported an inverse correlation between BMI and SaO₂ [27], and Littleton et al. reported that as BMI increased in patients with obesity, their PaO₂ decreased [26]. Similarly, our results show that the SpO₂ values demonstrate a similar association with BMI. Few previous studies have evaluated the relationship between SaO₂ and age. However, Vold et al. and Kapur et al. reported that SpO₂ is negatively associated with age in older patients with obesity (mean age was 65.8 years or age > 64 years), which suggests an inverse correlation between the level of oxygen saturation and age [28, 29]. We also found a significant correlation between age and SaO₂, but not SpO₂, suggesting that SpO₂ is not associated with age in patients with obesity. However, a larger study should be conducted to confirm this association in patients with obesity.

Differing from previous studies that only compared the relationship between BMI and oxygen levels [26, 27], our present study emphasizes determining the discrepancy between SpO₂ and SaO₂ in various patient populations undergoing bariatric surgery and defining factors affecting the discrepancy, including BMI, age, sex, the degree of hypoxemia, and OS-MRS. Our data show that SpO₂ overestimated SaO₂ in 91% of the participants. SpO₂ overestimated SaO₂ by a mean of 2.05% in the participants with obesity, and the level of bias increased to 3.08% in people with obesity suffering from hypoxemia. The difference between SpO₂ and SaO₂ was positively correlated with BMI, and the limits of the concordance between the measurements were higher in participants with severe obesity than in participants with obesity. A similar positive association was also found in the difference between SpO₂ and SaO₂ and age. The bias and the limits of the concordance between the measurements were higher in participants ≥ 40 years of age than in those < 40 years of age. Taken together, our data suggest that as BMI, age, and the degree of hypoxemia increase, the inaccuracy of SpO₂ for the estimation of SaO₂ also increases. Hypoxemia leads to some fatal diseases [30–32]; thus, timely identification and intervention would be highly essential.

There was also a strong influence of sex on the associations of SaO₂ with BMI and age. However, the findings of previous studies suggest that respiratory function and airway behavior differ between the sexes, and sex hormones [33, 34] and waist-to-hip ratio [35] may play a role. In addition, smoking status may contribute to hypoxemia. Carbon monoxide produced by tobacco combustion will combine with hemoglobin to form carboxyhemoglobin, thus affecting oxygen transport [36]. The smoking rates varied by sex in this study, with 48% of male patients smoking and only 17% of female patients smoking.

OS-MRS is a useful means of assessing the risk of postoperative complications in patients with obesity undergoing surgery [25]. Therefore, in the present study, we determined whether the OS-MRS correlated with the difference between SpO₂ and SaO₂ in patients with obesity, with the objective of guiding the use of pulse oximetry or ABG testing in patients undergoing bariatric surgery. We found that the difference between SpO₂ and SaO₂ was significantly higher in patients with high-risk scores than in those with low or intermediate risk scores. Based on these findings, we conclude that while pulse oximetry provides valuable, timely, and continuous measurements of intraoperative oxygen saturation in comparison to baseline values, patients who have a high OS-MRS would benefit from the use of intraoperative ABG testing and absolute oxygen saturation measurements.

Nevertheless, our present study also encounters some limitations. This single-center study recruited 134 patients with obesity, and no comparison was made between these patients and healthy lean controls. To reduce measurement bias, at least two measurements per time point per person is ideal; however, due to the nature of this retrospective study, one SaO₂ measurement per patient was taken. For better perioperative monitoring in patients with obesity undergoing bariatric surgery, further studies to compensate for these limitations are necessary.

Conclusion

Compared with SpO₂, SaO₂ can more accurately reflect the real oxygen saturation in patients with obesity undergoing bariatric surgery, especially for those with a BMI ≥ 40 kg/m², age ≥ 40 years, and high OS-MRS. Simple pulse oximetry suffices for younger, lighter, and low or intermediate OS-MRS patients, while for older, heavier, and high OS-MRS patients, SaO₂ gas monitoring is necessary. Our results provide a better understanding of the magnitude, direction, and factors affecting the concordance between SpO₂ and SaO₂ and support the use of ABG during the care of certain populations with obesity undergoing bariatric surgery.

Supplementary Material

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s11695-022-06221-7.

Key Points.

• SpO₂ overestimates SaO₂ in patients with obesity.

• SaO₂ better reflects the lower oxygen saturation in older male patients.

• The inaccuracy of SpO₂ increases with BMI, age, and OS-MRS.