Background
Near-infrared spectroscopy (NIRS) has been widely used in patients undergoing non-cardiac surgery because it is a non-invasive, continuous, and timely method for monitoring regional cerebral tissue oxygen saturation (SctO
2) [
1]. The light source of the NIRS device emits a constant intensity of radiation in the near–infrared spectrum which can penetrate tissues and be absorbed by oxyhaemoglobin and deoxyhaemoglobin [
2]. Sensors at a set distance receive a varied spectrum intensity to enable the calculation of the concentration of oxyhaemoglobin according to the Beer–Lambert law [
2]. Both ipsilateral and bilateral sensors on the forehead have been used for SctO
2 monitoring in different studies [
3,
4]. One major concern is the discrepancy between left and right SctO
2 readings. For example, a cohort study of healthy patients showed that the SctO
2 of the ipsilateral sensor showed better accuracy with reference to jugular bulb venous oxygen saturation (SjO
2) on the same side than on the contralateral side, and the comparison of readings between the two sides on a subject-by-subject basis showed a wide range of discrepancy [
5]. This difference in bilateral readings has also been reported in pre-term infants [
6].
Evidence in healthy volunteers showed that respiratory parameters significantly influence the accuracy of SctO
2 measurement; for example, hypoxia or hypocapnia may increase measurement bias by approximately 10% [
7,
8]. In clinical settings, respiratory parameters have also profound effects on cerebral saturation. First, the change of end-tidal carbon dioxide (EtCO
2) is highly associated with SctO
2. A cohort study of 20 surgical patients found that mild hypercapnia (approximately 45 mmHg) significantly increased SctO
2 (68% vs. 55% in the normocapnia group) [
9]. In contrast, hypocapnia (i.e., induced by hyperventilation) led to an approximately 10% decrease in SctO
2 in patients undergoing non-neurosurgical procedures [
10]. Second, lower arterial oxygen pressure is associated with a higher risk of cerebral desaturation [
11]. In contrast, a higher inspired oxygen fraction is useful for the elevation of SctO
2 in patients undergoing carotid endarterectomy [
12].
One-lung ventilation (OLV) is a necessary technique in thoracic surgery. The incidences of hypoxia and hypercapnia are as high as 50% during OLV [
4,
13]. In a cohort study of 15 patients undergoing OLV, hypercapnia (arterial partial pressure of carbon dioxide = 50 mmHg) increased SjO
2 from 54 to 69% [
14]. However, there is limited data to illustrate whether OLV affects the agreement between left and right SctO
2 and if the change in difference between left–right readings has clinical significance.
The present study was primarily designed to investigate whether OLV influences the agreement between the left and right SctO2 readings.
Methods
Ethic and registration
This study was a secondary analysis of a cohort study which was approved by the Clinical Research Review Board of Peking University First Hospital (#2017–1378) and registered in the Chinese Clinical Trial Registry (
http://www.chictr.org.cn, ChiCTR-ROC-17012627) [
4]. Written informed consent was obtained from all subjects and/or their legal guardians. All methods were performed in accordance with the relevant guidelines and regulations at Peking University First Hospital and Fourth Hospital of Hebei Medical University.
Patient inclusion and exclusion criteria
Adult patients (age ≥ 55-year-old) who received OLV under general anesthesia with or without epidural or paravertebral block were enrolled. Patients were excluded if they met any of the following criteria: 1) no record of SctO2 or 2) no record of EtCO2 or peripheral pulse oxygenation (SpO2).
Primary outcome
The primary outcome was agreement of left–right SctO
2 readings at 30 min after OLV. SctO
2 was monitored at the bilateral forehead using the FORE-SIGHT ELITE tissue oximeter (CASMED, Branford, CT, USA) from baseline to the end of surgery [
4]. The readings of SctO
2 were generated by oximetry every 2 s and extracted from the monitor at the end of surgery. Left–right SctO
2 readings were collected in pairs at 1 min intervals.
Baseline measurements were obtained before anesthesia induction, with the patients resting and breathing room air. The screen of the tissue oximeter was covered with an opaque bag to blind the anesthesia providers for monitoring. Dedicated research personnel checked the tissue oximeter every 10 min to ensure proper function.
Secondary outcomes
Secondary outcomes included the correlation of left–right SctO2 at eight time points before, during, and after OLV, the trend of left–right SctO2 over the first 30 min after OLV, the agreement of left–right SctO2 during OLV for each patient, the maximal difference between left–right SctO2, and its relationship with postoperative delirium.
Correlations of left–right ScO2 at eight time points were calculated at baseline, at 5 min interval during the first 30 min after the beginning of OLV, and at 5 min interval during the first 15 min after the end of OLV. The trends of left–right SctO2 over the first 30 min after OLV were analysed to investigate whether they had the same trends of change. Left–right SctO2 readings during OLV in each patient were collected in pairs, and their agreements were analysed individually. The maximal difference was defined as the maximal value between each pair of left–right readings, and its relationship with delirium was analysed accordingly.
Delirium was assessed twice daily (at 06:00–08:00 and 18:00–20:00) within postoperative 5 days using the Chinese version of the Confusion Assessment Method (CAM) in non-intubated patients and the CAM for intensive care unit (CAM-ICU) in intubated patients [
4,
15‐
17]. The researchers who were responsible for delirium assessment participated in a 4 h training session and were not allowed to access patient data during the research [
16,
17].
Anesthesia and perioperative care
All patients underwent the same monitoring on arrival at the operating room, including SpO
2, electrocardiography, and non-invasive blood pressure. Bispectral Index (BIS), EtCO
2, and nasopharyngeal temperature were monitored during general anesthesia which have been described in the original study as previously reported [
4]. Invasive arterial blood pressure and central venous pressure can be used when necessary.
Anesthesia was induced using propofol (2–4 mg/kg) and sufentanil (site-effect concentration, 0.2
–0.5 ng/ml). Anesthesia maintenance was administrated by propofol (4
–10 mg/kg/h) and sufentanil (site-effect concentration, 0.2
–0.5 ng/ml) via continuous infusion. Anesthesia depth was maintained a BIS value between 40 and 60 as previously reported [
4].
A double-lumen endotracheal tube was used for intubation. The tidal volume was set at 6–8 ml/kg. The aim of the minute ventilation volume was to maintain EtCO2 at 35–45 mmHg and SpO2 ≥ 92%. Fluctuation in blood pressure was maintained within 20% of the baseline value. Nasopharyngeal temperature was maintained at 36–37 °C.
Postoperative pain was assessed using numeric rating scale (an 11-point scale where 0 indicates no pain and 10 indicates the worst pain). Patient-controlled intravenous analgesia (PCIA) was used to maintain a pain score of 3 or lower. The program of PCIA was set to deliver a background infusion of sufentanil at 1.25 μg/h and a 2.5 μg bolus with a lock-out interval of 8 min for breakthrough pain [
4].
Statistical analysis
Power calculation
The Pearson’s correlation coefficient of left–right SctO2 at baseline was 0.690, and that at 30 min after OLV was 0.383. If the significance level was set at 0.05, the present sample size of 124 patients yielded a power of 0.99.
Outcome analysis
Continuous variables with normality are presented as mean (standard deviation, SD) and variables without normality are presented as median (interquartile range, IQR). The binary variables are presented as numbers (percentages).
For the primary outcome, the correlation between left–right SctO
2 at 30 min after OLV was first analysed using Pearson correlation. A linear regression model was then employed to investigate the correlation of left–right SctO
2 after adjusting for confounders. Selection of confounders was based on the results of previous trials, including comorbidities (age, diabetes, and hypertension) and parameters at 30 min after OLV (i.e., SpO
2, EtCO
2, mean arterial blood pressure) [
7,
8,
18‐
21].
For the secondary analysis, the correlation of left–right SctO2 at fixed time points was also analysed in line with the primary outcome. The Bonferroni method was used to control for type I errors. A linear mixed model was used to compare the trend of left–right SctO2 over the first 30 min after OLV, with adjustment for confounders. Pearson correlation was used to analyse the agreement between left and right SctO2 during OLV in each patient. The Pearson’s correlation coefficients were divided into three groups: strong (r ≥ 0.7), medium (0.5 ≤ r < 0.7), and weak (r < 0.5) correlations. The maximal differences are presented as medians (IQR). Its relationship with postoperative delirium was analysed using multivariate logistic analysis after adjusting for confounders. Confounders were first selected by univariate analysis, and variables with P < 0.05 were entered into multivariate analysis.
Statistical significance was set at P < 0.05. Analyses were performed using R 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria, 2019).
Discussion
The present study found that OLV increased the discrepancy in left–right SctO2 readings in patients undergoing thoracic surgery. The maximal difference in the left–right SctO2 was not associated with delirium.
Although there are data to support the poor correlation of bilateral SctO
2 readings in healthy volunteers, the present study is the first to investigate this phenomenon in patients undergoing OLV [
5]. It is obvious that the correlation between left and right SctO
2 became poor at 30 min after OLV in comparison with baseline. We selected the correlation of left–right SctO
2 at 30 min after OLV as the primary outcome because patients experience the most severe changes in respiratory parameters during this period. For example, the overall incidence of hypoxia during OLV was 58.1% (72/124), while 52.8% (38/72) occurred in the first 30 min after OLV.
One strength of our study is that we employed a linear regression model to adjust for confounders that might affect SctO
2 measurement. Most studies only used simple correlation analysis, such as Pearson correlation, to compare bilateral SctO
2 [
5,
7]. This method is practical for healthy volunteers but is not suitable for patients in clinical settings. Age, diabetes, hypertension, SpO
2, EtCO
2, and mean arterial blood pressure may significantly affect the accuracy of SctO
2 readings [
7,
8,
18‐
21]. Our results showed that the correlation between left and right SctO
2 at 30 min after OLV decreased from 0.383 to 0.323 after adjustment. This indicated that the interpretation of SctO
2 during OLV should consider preoperative comorbidities and respiratory parameters.
Because the correlation at a single time point could not reflect the trend of change, we provided three methods to illustrate the question. First, we used a linear mixed model to compare the trend of SctO
2 change during the first 30 min after OLV. Unlike classic correlation that measures the agreement between two variables, a linear mixed model can provide an analysis of the trend of repeated measurements [
22]. Our results showed that the change trend between left–right SctO
2 presented statistically significant over time. Second, we provided Pearson’s correlation coefficients at eight fixed time points as representative of changes before, during, and after OLV. Our results showed that the phenomenon of poor correlation between left and right SctO
2 still existed at 15 min after the end of OLV. Third, we compared the correlation of left–right SctO
2 during OLV for each patient with the subject-to-subject purpose. Our results showed that a strong correlation existed only in 62.9% of patients. An inverse correlation was observed in some patients.
Another interesting finding was that the maximal difference between the left and right SctO
2 readings was approximately 13.5 which might have statistical significance for diagnosing cerebral desaturation. Our previous analysis showed that a relative decrease of 10% from the baseline SctO
2 value was highly associated with delirium [
4]. Taking this criterion as a reference, the incidence of cerebral desaturation was 61.5% on the left side and 57.5% on the right side. However, multivariate analysis showed that the maximal difference in left–right SctO
2 was not related to postoperative delirium. The clinical significance of this difference warrants further investigation.
The discrepancy in the left–right SctO
2 readings may be mainly attributed to two aspects. First, blood supply to the left and right hemispheres is generated from different arteries. Cerebrovascular disease, such as severe stenosis of the internal carotid artery and circle of Willis may impair oxygen supply and consumption on the ipsilateral side [
23‐
25]. One limitation of the present study was that we did not perform preoperative ultrasound screening of the carotid artery; however, all patients underwent thorough physical examination, including auscultation of the carotid bruit and no positive events were reported in the medical records. Second, measurement bias may be attributed to the calculation algorithm. The readings of SctO
2 are based on a calculation algorithm with a fixed ratio of cerebral mixed venous and arterial blood (i.e., 30:70 claimed by CASMED) [
7]. Many factors induce variations in the in vivo ratio. For example, two studies reported that artificial hypoxia or hypocapnia may increase measurement bias in approximately 10% of healthy volunteers [
7,
8]. In our study, we could not evaluate the accuracy of the left and right readings, because SjvO
2 was not measured in the present study which was considered as a “gold” reference to calibrate SctO
2.
Our results show that bilateral SctO2 at baseline is well correlated, but the trend of bilateral SctO2 is different and is significantly affected by OLV. This result suggests the application of bilateral SctO2 in clinical practice and studies. Further studies are needed to illustrate how to clinically act if a patient suffers from one or two-sided cerebral desaturation.
One limitation of the present study is its secondary analysis design. However, all these data were prospectively collected in a previous study which ensured the quality of the data. Second, the sample size was limited to 124 patients. Based on the results, the sample size yielded a statistical power of 0.99. Third, as discussed above, SjvO2 and carotid artery ultrasound were not performed.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.