Content area
Backgrounds
The cerebral oximetry index (CO
Our prospective observational cohort study enrolled patients scheduled for CEA. CO
One hundred and forty patients scheduled for CEA were enrolled. The incidence of delirium was 10.7 % (15/140) and the incidence of stroke was 3.6 % (5/140), including 1 patient who had both. The cumulative anesthesia time when CO
CO
Carotid endarterectomy (CEA) is a common treatment for stenosis or occlusion of carotid arteries. The incidence of postoperative cognitive decline and cerebral ischemia in these patients, including postoperative delirium and stroke, remains high. [ 1–3] Cerebral hypoperfusion and hypoxia are important risk factors [ 4–6] and are associated with impaired autoregulation of cerebral blood flow. As might thus be expected, patients with impaired autoregulation are most likely to develop postoperative delirium [ 7] and stroke [ 8].
The primary goal of hemodynamic management is to maintain adequate arterial pressure to ensure adequate cerebral perfusion, thereby reducing postoperative cerebral ischemia. Cerebral autoregulation refers to the ability of the cerebral vasculature to maintain constant cerebral blood flow despite changes in systemic blood pressure. [ 9] Patients presenting for CEA often have impaired autoregulation, but the shape and limits of individual autoregulation curves vary substantially and unpredictably. [ 10] Real time monitoring of perioperative cerebral autoregulation may therefore reduce ischemic events.
Cerebral oximetry index (CO x) is a promising tool for monitoring cerebral autoregulation. This real-time continuous metric is defined by the correlation coefficient between invasive mean arterial pressure (MAP) and regional cerebral oxygen saturation. When cerebral autoregulation is intact CO x fluctuates around zero, and becomes increasingly positive as autoregulation is progressively impaired. [ 11] Based on studies in patients having cardiac surgery, globally impaired autoregulation was likely when CO x increases to between 0.3 and 0.5. [ 8, 12, 13] Furthermore, stable or rising CO x values in shunted patients during awake CEA suggests clamp ischemia. [ 14] However, the extent to which impaired CO x predicts postoperative neurological complications in patients having CEA remains unclear.
We therefore conducted a prospective observational study to explore optimal CO x thresholds for predicting a composite of delirium or stroke in patients recovering from CEA conducted with general anesthesia.
2 Materials and methods2.1 Participants
The study population was derived from an ongoing prospective observational study, which was approved by the Chinese Registration Clinical Trial Ethics Committee (ChiECRCT20190235) and registered at www.clinicaltrials.gov ( NCT04362709). We enrolled consenting patients scheduled for elective CEA. Exclusion criteria were: 1) severe cognitive impairment (Mini-mental State Examination <18) [ 15], 2) current use of psychotropic medications; 3) history of intracranial surgery; and 4) visual or hearing impairment that precluded accurate assessment of delirium.
2.2 Anesthetic management and data collectionDuring surgery, routine monitoring was used, including electrocardiography, pulse oximetry, end-tidal carbon dioxide partial pressure, cerebral oxygen saturation, and invasive blood pressure. Anesthesia was induced with sufentanil (0.3–0.5 μg/kg), rocuronium (0.6 mg/kg) or cistracurium (0.15–0.2 mg/kg), and propofol (1.5–2 mg/kg) or etomidate (0.3–0.5 mg/kg). Tidal volume was set to 6–8 mL/kg with a positive end-expiratory pressure of 3–5 cmH 2O which has minimal effect on ICP. [ 16] Respiratory rate was set to keep arterial carbon dioxide partial pressure between 35 and 45 mmHg. 5–10 μg of sufentanil was given at incision and 0.05–0.1 μg/kg/min of remifentanil was given for anesthetic maintenance.
Mean arterial pressure was maintained within ±20 % of baseline during most of surgery by infusing norepinephrine, phenylephrine or nicardipine; however, we targeted a MAP 20 % above baseline during carotid clamping. [ 17] Anesthesia was adjusted to target Bispectral Index (processed electroencephalography) between 40 and 60 which is considered an ideal hypnotic depth. Anesthesia administration stopped at the end of surgery.
Patients were extubated when spontaneous respiration recovered with frequency of 12–20 breaths/min, tidal volume of 8–10 mL/kg body weight, and oxygen saturation > 90 % on room air. We also required normal PaCO 2 and stable cardiovascular variables. Patients were transferred to the post-anesthesia care unit after surgery, where monitoring included electrocardiography, pulse oximetry, and non-invasive blood pressure. Patients were initially given oxygen via a mask at a rate of 2–4 L/min. About an hour later, the patients were transferred back to the ward.
Clinical data were collected from electronic medical records at Beijing Tiantan Hospital, Capital Medical University, and included demographic characteristics, comorbid diseases, chronic medication, preoperative cognitive function screening, laboratory test results, and intraoperative information.
2.3 Cerebral oximetry index measurement and calculationAfter admission to the operating room, continuous bi-hemispheric regional cerebral oxygen saturation was assessed. Arterial blood pressure was sampled at 0.1 Hz captured by Anesthesia Information Management System (AIMS, version 5.0, Wangfeng Mingyue Ltd., China). When no blood pressure measurements were recorded or a value was marked as an artifact, pressures between measurements were linearly interpolated. [ 18] MAP was estimated as (SBP + 2 × DBP)/3.
Regional cerebral oxygen saturation was acquired through near-infrared spectroscopy (BRS-1, Casibrain Tech Ltd., Beijing, China) at a sampling rate of 1 Hz. Monitoring electrodes were placed symmetrically on both sides of the forehead below the hairline. Cerebral oxygen saturation was continuously monitored, but clinicians and investigators were blinded to avoid influencing clinical decisions.
Blood pressure and cerebral oxygenation data were exported, and CO x was estimated off-line with a program written in Python (version 3.9.13; Python Software Foundation, https://www.python.org). Data were filtered by resampling over a 10-s rolling average. Additionally, we interpolated MAP data to increase its frequency to 1 Hz to synchronize it with the cerebral saturation using spline interpolation method. [ 19] Next, as the MAP and cerebral oxygen saturation data were not normally distributed, a continuous, moving Spearman correlation coefficient between 300 consecutives, paired resampled MAP and cerebral oxygen saturation signals in ipsilateral carotid stenosis was calculated to generate the CO x. [ 20, 21] In a post hoc sensitivity analysis, we also estimated CO x using Pearson regression. [ 11, 22] Supplementary Fig. S1 presents two examples of measuring MAP and regional cerebral oxygen saturation and calculation of CO x.
Blood pressure and cerebral oxygenation data were exported, and CO x was estimated off-line with a program written in Python (version 3.9.13; Python Software Foundation, https://www.python.org). Data were filtered by resampling over a 10-s rolling average. Additionally, we interpolated MAP data to increase its frequency to 1 Hz to synchronize it with the cerebral saturation using spline interpolation method. [ 19] Next, as the MAP and cerebral oxygen saturation data were not normally distributed, a continuous, moving Spearman correlation coefficient between 300 consecutives, paired resampled MAP and cerebral oxygen saturation signals in ipsilateral carotid stenosis was calculated to generate the CO x. [ 20, 21] In a post hoc sensitivity analysis, we also estimated CO x using Pearson regression. [ 11, 22] Supplementary Fig. S1 presents two examples of measuring MAP and regional cerebral oxygen saturation and calculation of CO x.
2.4 Study outcomesOur primary objective was to investigate the ability of intraoperative CO x to predict a collapsed (either or both) composite of delirium and new-onset overt stroke during the initial 5 postoperative days, expressed as area under the receiver operating characteristic curve.
Delirium was assessed twice daily for 5 postoperative days using the Richmond Agitation Sedation Scale (RASS) and the 3-min Diagnostic Interview for CAM (3D-CAM). Delirium was not assessed when patients were unresponsive to verbal stimuli ( i.e., RASS score ≤ −4). Postoperative delirium diagnosis was made dichotomously and based on any positive within 5 postoperative days. Assessments were performed by independent research staff trained in delirium assessment.
Overt ischemic strokes were diagnosed by new-onset neurological symptoms such as limb weakness, paranesthesia, or language abnormalities, or visual field defects, or altered consciousness. We also assessed new ischemic foci with routine postoperative head computerized tomography. Assessments were performed independently by two investigators. Differences were resolved by consensus when possible, and if necessary adjudicated by the senior neurosurgeon.
Secondary outcomes include other postoperative complications and duration of hospitalization. Other postoperative complications included the incidence of reoperation, ICU admission, post wound infection or hematoma were monitored until discharge.
2.5 Sample size estimation and statistical analysisSample size was based on detecting an AUC of 0.75 with an alpha significance level of 0.05. Previous studies report that about an eighth of similar patients experience delirium and/or strokes. [ 2, 3] A population with 14 event among 126 patients provides 80 % power to detect COx differences in patients who did and did not experience neurologic complications. Sample size was estimated using the PASS 15.0 software (NCSS LLC, Kaysville, Utah). We included 140 patients to account for an anticipated 10 % loss to follow-up.
Shapiro-Wilk test was used to verify normality of continuous data. Normally distributed measurement data were expressed as means ± standard deviations, and non-normally distributed data were expressed as medians (interquartile ranges, IQR). Categorical variables were expressed as frequencies (percentages). Independent sample t-tests were used to analyze normally distributed continuous data and independent sample Mann-Whitney U tests were used for continuous data with non-parametric distributions. Differences in categorical variables were analyzed using X 2 or Fisher exact tests. Baseline characteristics and intraoperative variables that differed between the groups at P < 0.1 were included in the multivariable logistic regression model to identify independently risk factors associated with ischemic events.
Based on previous studies, we initially set CO x > 0.3 as a predictor of impaired autoregulation. [ 12] A sliding threshold approach was then used to identify CO x values ranging from 0.4 to 0.8, incremented by 0.1. The cumulative time above each value was computed and compared among patients with postoperative neurological complications and without using t or Mann-Whitney U tests. Receiver operating characteristics curves with Youden's index were analyzed using k-fold cross-validation (10 folds) method [ 23] to determine which CO x threshold most accurately distinguished prognostic composite outcomes. The optimal value was identified as the CO x at maximum AUC and maximum Youden's index. The association between CO x and outcomes was assessed by k-fold (k = 10) logistic regression analysis, to estimate odds ratio and 95 % CI of duration at the optimal CO x threshold. We defined a range of durations from 5 to 30 min, with a step size of 5 min. CO x was calculated from insertion of the arterial catheter until the end of anesthesia. In a post hoc subgroup analysis, CO x was divided into three phases: before, during, and after clamping.
SPSS 23.0 software (IBM, Armonk, New York) and STATA17.0 software (Stata Corporation, College Station, Texas) were used for statistical analysis. P < 0.05 were considered to indicate statistical significance.
3 ResultsAmong 175 patients who had CEA between 17th March 2021 and 5th June 2023, we enrolled 140 who met our inclusion and exclusion criteria. The incidence of postoperative neurological complications was 13.6 % (19/140), including 14 patients who presenting with delirium, 4 patients who had new-onset strokes, and 1 patient who had both ( Fig. 1 ).
Demographic and baseline characteristics were similar in patients who did and did not have delirium or stroke ( Table 1 ). Nearly all intraoperative factors were comparable between the groups including surgical duration, duration of carotid artery occlusion, recovery time, intraoperative anesthetic doses, amount of vasoactive medication used, and the arterial partial pressures of carbon dioxide and oxygen. Patients without neurological complications recovered consciousness after surgery slightly more quickly than those who did experience delirium or stroke, but not by a clinically meaningful amount: 15 min [IQR 10–20 min] vs 18 min [IQR 14–26 min], P = 0.031 ( Table 2 ). There were no significant differences between the two groups in MoCA-B score on the fifth postoperative day, duration of hospitalization, or other complications ( Table 3 ).
Because perioperative variables were comparable in patients with and without neurological complications, we analyzed the relationship between cerebral autoregulation and postoperative neurological complications without adjustment. Arterial pressures or cerebral oxygen saturation was missing in 6 patients. Consequently, CO x analyses were based on 134 patients of whom 17 had complications and 117 did not.
MAPs and cerebral oxygen saturations were comparable, except that MAP (90 vs. 82 mmHg, P = 0.032) was slightly higher in patients with complications than those without during the post-clamping phase (Supplementary Table S1). Over the entire anesthesia duration, when CO x was equal to and more than 0.3, the cumulative time when CO x exceeded each threshold was longer in patients with delirium or stroke than those without ( Fig. 2A ). When the CO x threshold was 0.6, the corresponding predictive ability was AUC of 0.69 and Youden index of 0.61 ( P = 0.0003, Fig. 2B) with a positive predictive value of 100 %. CO x exceeding 0.6 for a minute increased the odds of delirium or stroke by 9 % (Odds ratio = 1.09, 95 % CI 1.02 to 1.17) and for 30 min with the odds ratio 4.97 (95 % CI 1.39 to 10.97) ( Fig. 2C).
In the post hoc subgroup analyses before clamping, the cumulative time when CO x exceeded 0.7 was longer in patients with delirium or stroke ( Fig. 2D) and was most predictive ( Fig. 2E), especially when the 0.7 threshold was exceeded by at least 20 min (Odds ratio = 3.10, 95 % CI 2.20, 3.78) ( Fig. 2F). In contrast, CO x was not predictive during clamping ( Fig. 2G-I). After clamping, the optimal CO x threshold was 0.4 (AUC = 0.85, Youden index = 0.82, P < 0.0001, Fig. 2K) and the positive predictive value was 100 %.
In the sensitivity analysis, we applied Pearson's coefficients to calculate CO x, resulting in identical outcomes to those previously observed ( Supplementary Fig. S2).
In the sensitivity analysis, we applied Pearson's coefficients to calculate CO x, resulting in identical outcomes to those previously observed (Supplementary Fig. S2).
4 DiscussionPerioperative neurological complications are common during and after CEA. For example, the incidence of overt postoperative strokes was 3.6 % in our patients which is consistent with previous reports. [ 3] The incidence of delirium, at 10.7 % in our patients, was slightly higher than previous reports of 3 % [ 1] to 8 % [ 2], possibly reflecting use of different cognitive assessment tools and our 5-day twice-daily assessments.
Near-infrared spectroscopy is designed to assess oxygen saturation of superficial frontal lobe tissue [ 24]. Some studies suggested that use of near-infrared spectroscopy reduces unnecessary shunting during carotid artery surgery [ 25, 26], whereas the others suggest no benefit [ 27, 28]. Rather than simply using frontal lobe saturation, we combined changes in saturation and changes in MAP into an index of cerebral autoregulation. Specifically, CO x is based on a moving linear correlation coefficient between MAP and reginal saturation of oxygenation. [ 10, 29] CO x is based on the theory that there is little correlation between cerebral saturation and MAP when autoregulation is intact, and that correlation becomes increasingly apparent as autoregulation fails.
Our primary goal was to determine the CO x correlation coefficient most associated with neurological complications (delirium and stroke) in patients having CEA. Patients with impaired cerebral autoregulation are at elevated risk of developing postoperative delirium [ 12] or stroke [ 8] after cardiac surgery. The underlying mechanism may be attributed to cerebral hypoxia. Although delirium is multifactorial, inadequate perfusion consequent to loss of cerebral autoregulation likely contributes. [ 5] As might therefore be expected, globally impaired autoregulation and cumulative time of MAP below the autoregulation limit is associated with delirium after cardiac surgery. [ 12].
A CO x of 0.7 in the pre-clamping phase predicted delirium or stroke after CEA which is considerable higher than the 0.3–0.5 range that is taken to indicate impaired autoregulation in patients having cardiac surgery. [ 8, 12, 22, 30] The high threshold was necessary to provide a good balance between sensitivity and specific for an event as rare as stroke. An additional factor is that carotid lesions impair cerebral autoregulation [ 31] which is further aggravated by hypertension, hyperlipidemia, or chronic cerebral infarction. Half of our patients had a history of stroke and two-thirds had a history of hypertension. CEA patients usually have bilateral carotid artery stenosis and cerebral autoregulation remains persistently impaired bilaterally after single-side surgery. [ 32, 33] In contrast to CO x in the pre-clamp period, CO x during clamping did not predict delirium or stroke, possibly because the exposure period was too short.
One important limitation of our analysis was that we considered only overt strokes, although covert strokes (detected by MRI screening) are nearly 6 times as common [ 34]. Presumably we would have had more power had we included covert strokes, and possibly our estimate for the optimal CO x threshold would differ. A second limitation is that delirium or strokes have multifactorial etiologies and surely only a fraction are due to cerebral hypoperfusion. A third concern is that we evaluated CO x during surgery which precluded consideration of postoperative hemodynamic events which may have contributed to strokes and/or delirium. A fourth concern is that the small number of outcome events limits the precision of our optimal CO x estimate. A larger population with more outcome events would provide a more robust estimate. Another consideration is that vasopressors increase reginal cerebral oxygenation as well as MAP [ 35–37], but the effect of vasopressors on CO x remains unknown. We cannot address the issue because all our patients were given vasopressors. And finally, baseline CO x was not obtained because arterial catheters were inserted after anesthetic induction.
5 ConclusionIn conclusion, the optimal CO x threshold for predicting postoperative delirium or stroke in patients having CEA was 0.7 in the pre-clamping phase and 0.4 in the post-clamping phase. Whether the observed association between loss of cerebral autoregulation and postoperative delirium or stroke is causal and thus amenable to intervention remains unknown.
The following are the supplementary data related to this article. Supplementary Fig. S1 CO x monitoring and calculation in two example patients. MAP data (left), rScO2 data (middle), and the CO x curve (right). MAP, mean arterial pressure; rScO2, regional cerebral oxygen saturation; CO x, cerebral oximetry index; the dashed red line indicates clamping and the dashed green line indicates declamping; solid black lines were horizontal lines of CO x > 0.3 or > 0.7; the X-axis indicates elapsed operating room time. Supplementary Fig. S1 Supplementary Fig. S2 Predicting postoperative neurological complications at each threshold of CO x in the pre-clamping phase by using Pearson regression. Analyzed the data using Pearson regression in the pre-clamping phase. A. Cumulative time of CO x over each threshold in patients with complications and without. B. ROC curves for the cumulative time of CO x over each threshold to predict postoperative neurological complications. C. Logistic regression model was used to evaluate associations between CO x and outcomes. The odds ratios for various durations are reported along with 95 % CI. Supplementary Fig. S2 Supplementary material Supplementary material
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jclinane.2024.111733.
FundingCapital Development Research Fund Project ( 2024-2-2047).
CRediT authorship contribution statementMuhan Li: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Tingting Ma: Writing – original draft, Data curation. Xueke Yin: Writing – original draft, Resources, Data curation. Xin Zhang: Software, Methodology, Formal analysis. Tenghai Long: Software, Methodology, Formal analysis. Min Zeng: Investigation, Data curation. Juan Wang: Investigation, Data curation. Qianyu Cui: Data curation. Shu Li: Investigation, Data curation. Daniel I. Sessler: Writing – review & editing. Rong Wang: Supervision, Data curation. Yuming Peng: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.
Declaration of competing interestThe authors declare that they have no conflicts of interest.
AcknowledgementThe authors thank Mengchen Lu (Casibrain Technology Limited Company, Beijing, China) for her assistance with data analysis.
| Characteristics | Without neurological complications | With neurological complications | |
| (
| (
| ||
| Age, years, median (IQR) | 66 (61, 71) | 67 (64, 74) | |
| Male, n (%) | 104 (86.0) | 17 (89.5) | |
| BMI, kg/m 2 a, median (IQR) | 24.8 (22.9, 27.4) | 25.1 (22.8, 27.7) | |
| Educational background, n (%) | | ||
| | 5 (4.1) | 0 (0) | |
| | 35 (28.9) | 7 (36.8) | |
| | 60 (49.6) | 9 (47.4) | |
| | 21 (17.4) | 3 (15.8) | |
| Symptomatic stenosis, n (%) | 105 (86.8) | 15 (79.0) | |
| | 55 (45.5) | 8 (42.1) | |
| | 39 (32.2) | 7 (36.8) | |
| Medical history, n (%) | |||
| | 80 (66.1) | 15 (79.0) | |
| | 40 (33.1) | 7 (36.8) | |
| | 23 (19.0) | 4 (21.1) | |
| | 54 (44.6) | 11 (57.9) | |
| Medication history, n (%) | |||
| | 71 (58.7) | 14 (73.7) | |
| | 9 (7.4) | 2 (10.5) | |
| | 100 (82.6) | 14 (73.7) | |
| | 86 (71.1) | 10 (52.6) | |
| Preoperative functional score | |||
| | | ||
| | 56 (46.3) | 7 (36.8) | |
| | 65 (53.7) | 12 (63.2) | |
| | 26 (24, 28) | 25 (22, 27) | |
| Preoperative laboratory inspection, n (%) | |||
| | 9 (7.4) | 1 (5.3) | |
| | 21 (17.4) | 4 (21.1) | |
| Without neurological complications | With neurological complications | | |
| (N = 121) | (N = 19) | ||
| Type of anesthesia, n (%) | | ||
| | 102 (84.3) | 18 (94.7) | |
| | 19 (15.7) | 1 (5.3) | |
| Occlusion duration, min, median (IQR) | 25 (18, 38) | 22 (17, 51) | |
| Surgical duration, min, median (IQR) | 160 (128, 194) | 153 (135, 190) | |
| Anesthesia duration, min, median (IQR) | 211 (181, 240) | 234 (184, 290) | |
| Awakening duration, min, median (IQR) | 15 (10,20) | 18 (14, 26) | |
| Blood gases analysis, median (IQR) | |||
| | 7.432 (7.402, 7.459) | 7.426 (7.392, 7.450) | |
| | 38 (36, 41) | 40 (38, 42) | |
| | 210 (166, 240) | 189 (151, 238) | |
| | 12.9 (12.2, 13.5) | 12.9 (12.6, 13.7) | |
| | 37.7 (33.3, 40.7) | 38.6 (35.4, 42.6) | |
| | 6.0 (5.2, 6.7) | 5.6 (5.0, 6.5) | |
| | 0.9 (0.6, 1.2) | 0.8 (0.7, 1.3) | |
| Fluid intake, mL, median (IQR) | 2000 (1500, 2000) | 1600 (1500, 2000) | |
| Urine volume, mL, median (IQR) | 800 (550, 1400) | 600 (400, 1400) | |
| Blood loss, mL, median (IQR) | 50 (50, 100) | 50(50, 100) | |
| Midazolam, mg, median (IQR) | 2 (1,2) | 2 (0,2) | |
| Anticholinergic agents, n (%) | 53 (43.8) | 7 (36.8) | |
| Sufentanil, μg, median (IQR) | 35 (30, 40) | 35 (30, 40) | |
| Remifentanil, μg, median (IQR) | 960 (680,1280) | 1040 (800, 1400) | |
| Propofol, mg, median (IQR) | 630 (480, 820) | 720 (610, 920) | |
| Without neurological complications | With neurological complications | | |
| (N = 121) | (N = 19) | ||
| Reoperation, n (%) | 2 (1.7) | 2 (10.5) | |
| ICU admission, n (%) | 5 (4.1) | 2 (10.5) | |
| Post wound infections, n (%) | 4 (3.3) | 1 (5.3) | |
| Post wound hematoma, n (%) | 2 (1.7) | 1 (5.3) | |
| MoCA (B) score, median (IQR) | 20 (18, 22) | 21 (18, 22) | |
| Duration of hospitalization, median (IQR) | 10 (8, 13) | 13 (9, 15) | |
©2024. Elsevier Inc.