Association between intraoperative carotid blood flow reductions and postoperative delirium in older patients undergoing surgery: a prospective observational study Authors: Xingyang Liu, Chunyan Ye, Jinghan Wu, Yingqi Li, E Wang Published: 2026 ABSTRACT AbstractPurposePostoperative delirium (POD) is common among older patients undergoing surgery and is associated with impaired cerebral perfusion. In this study, we sought to investigate the relationship between intraoperative carotid artery blood flow changes and POD risk in older patients undergoing abdominal surgery.MethodsWe conducted a prospective observational analysis of patients older than 65 yr who underwent elective abdominal surgery in a single tertiary care hospital from September 2021 to May 2022. Common carotid artery (CCA) blood flow was continuously monitored intraoperatively using ultrasound vector flow imaging (VFI). We assessed the association between CCA flow changes and POD with multivariable logistic regression models. We explored additional models incorporating intraoperative mean arterial pressure (MAP) reductions to evaluate relative predictive performance. Postoperative delirium was assessed within 7 days postoperatively.ResultsAmong 210 patients analyzed (from 242 screened), POD occurred in 41 (19.5%). In the primary analysis, longer intraoperative reductions in CCA blood flow > 40% from baseline were independently associated with increased POD risk (adjusted odds ratio [aOR] per 10-min increment, 2.24; 95% confidence interval [CI], 1.51 to 3.31; Bonferroni-adjusted P < 0.001), which remained robust in sensitivity analyses (aOR, 1.78; 95% CI, 1.45 to 2.19; Bonferroni-adjusted P < 0.001). In secondary analyses, prolonged CCA blood flow increases were associated with reduced POD (aOR, 0.58 per 10-min increment; 95% CI, 0.40 to 0.83; Bonferroni-adjusted P < 0.001). Common carotid artery flow reductions showed limited concordance with MAP-based hypotension (Cohen’s κ ≤ 0.08). Models incorporating CCA flow showed better discrimination and calibration metrics for predicting POD than those based on MAP.ConclusionsIn older patients undergoing abdominal surgery, persistent intraoperative CCA blood flow reductions > 40% from baseline were associated with an increased risk of POD Intermittent CCA flow measurement with VFI may help identify patients not identifiable through blood pressure monitoring alone. Nevertheless, because the results derive from an observational design with interval-based sampling, the findings should be interpreted with caution and warrant prospective validation.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12630-026-03139-x.Keywords: carotid artery blood flow, intraoperative monitoring, older adults, postoperative delirium, vector flow imaging Postoperative delirium (POD) is a common perioperative complication in older patients, with an incidence ranging from 5% to 50% depending on surgery type.1 It manifests as acute confusion, inattention, and altered consciousness2 and is associated with adverse outcomes such as prolonged hospitalization, cognitive decline, and increased 1-year mortality.3–5 Despite its clinical significance, the pathophysiology of POD remains poorly understood. Intraoperative cerebral hypoperfusion is considered a key mechanism,6 yet direct cerebral perfusion monitoring during general anesthesia remains challenging. The carotid arterial system supplies approximately 82% of cerebral blood flow7 intraoperative changes in carotid flow may reflect cerebral perfusion status. Doppler ultrasound and more recently vector flow imaging (VFI) have emerged as practical, noninvasive modalities for real-time assessment of carotid blood flow dynamics intraoperatively.8,9 Prior research has identified various physiologic predictors of POD, such as intraoperative hypotension10 and cerebral oxygen saturation.11 Recent evidence indicates that patients undergoing abdominal surgery frequently experience significant reductions in carotid artery blood flow.12 Nevertheless, limited data exist concerning the association between fluctuations in carotid blood flow and POD occurrence. Furthermore, the clinical implications of both reduced and increased carotid blood flow during surgery have yet to be systematically explored. In this prospective observational study, we enrolled older adults undergoing elective abdominal surgery under general anesthesia. Intermittent intraoperative monitoring of carotid blood flow was performed using VFI. Our primary aim was to determine whether intraoperative carotid flow reductions were associated with POD. The secondary aim was to investigate the relationship between increased or mean carotid blood flow and the occurrence of POD. Methods Study design and population This single-centre prospective observational cohort study was approved by the Ethics Committee of Xiangya Hospital Central South University (Changsha, China) on 9 March 2021 (ref. No.: 20210386) and registered with the Chinese Clinical Trial Registry (ChiCTR2100048599) on 11 July 2021. We enrolled older patients (aged 65–80 yr) who underwent elective gastrointestinal surgery at Xiangya Hospital, Central South University, between September 2021 and May 2022. Patients were excluded if they met any of the following criteria: 1) history of severe traumatic brain injury, intracranial infection, Alzheimer’s disease, other forms of dementia, or psychosis; 2) imaging-confirmed cerebral infarction; 3) severe preoperative hypertension, diabetes mellitus, or coronary artery disease (for detailed definitions, see Electronic Supplementary Material [ESM] eAppendix 1); 4) alcohol dependence or withdrawal syndrome; 5) severe vision or hearing impairment; 6) hospitalization within the previous 3 months or terminal-stage illness; and 7) presenting with delirium at admission. Data were collected from electronic medical charts, anesthesia records, surgical reports, and laboratory results throughout the entire hospital stay. We obtained written informed consent from all participants or their legal representatives. Vector flow imaging measurement of carotid blood flow The primary exposure variable was intraoperative common carotid artery (CCA) blood flow. Preoperative baseline carotid artery blood flow was measured bilaterally, whereas intraoperative carotid artery blood flow was measured unilaterally using ultrasound VFI (linear array transducer, 3–9 MHz, Mindray Resona 7 ultrasound machine, Shenzhen Mindray Bio-Medical Electronic Co., Shenzhen, China). Vector flow imaging is an angle-independent technique that quantifies blood flow velocity and direction. This technology is increasingly available on modern high-end commercial ultrasound systems, although it is not yet standard on all machines. Preoperative baseline carotid artery blood flow was reported after patients entered the operating room. Following a 10-min rest period in the supine position, bilateral carotid artery blood flow was assessed. For each side, three consecutive flow measurements were obtained and averaged to determine the baseline carotid blood flow for the patient. Electronic Supplementary Material eTable 1 presents the intraobserver variability of carotid blood flow measurements. The mean (standard deviation [SD]) left-sided flow was 317.7 (67.5) mL·min−1, while the right-sided flow was 325.0 (66.9) mL·min−1. The Pearson correlation coefficient between the two sides was 0.724 (P < 0.001), indicating a strong linear relationship. Intraoperatively, carotid artery blood flow was recorded every 10 min from anesthesia induction until the end of surgery. All patients underwent consistent intraoperative measurement of the left CCA owing to operating room layout constraints. Intraoperative carotid artery blood flow reduction was defined by the degree of reduction relative to baseline values. The thresholds used to define blood flow reduction were < 30%, 30–40%, 40–50%, and > 50% relative to baseline. Blood flow measurements at each time point were used to give an estimate of blood flow over a 10-min period. Intraoperative carotid artery blood flow reduction was quantified using the duration of blood flow reduction within different thresholds throughout the surgery. Mean intraoperative carotid blood flow was calculated by dividing the sum of all reported carotid blood flow values by the total intraoperative duration of mechanical ventilation. Common carotid artery blood flow measurements adhered to a standardized protocol: The measurement site was 1.5–2 cm below the carotid bulb8 at a depth of approximately 2 cm. The arrow lifecycle was set to 25 msec, with arrow density maintained at 10%. The luminal diameter was measured as the distance between the inner layers of the parallel vessel walls. Electronic Supplementary Material eFig. 1 presents the specific illustration of carotid artery VFI measurement. Anesthesia and perioperative management Intraoperative monitoring included electrocardiography, noninvasive blood pressure monitoring, pulse oximetry, Bispectral Index™ (BIS™; Medtronic, Dublin, Ireland), end-tidal carbon dioxide (EtCO2) concentrations, anesthetic gas concentrations, and urine output. Invasive arterial pressure and central venous pressure were monitored when clinically indicated. The same anesthesia protocol was applied to all patients. Anesthesia induction was achieved using midazolam, etomidate, propofol, sufentanil, and cisatracurium, with dosages adjusted according to patient weight and age. Anesthesia maintenance involved propofol, remifentanil, cisatracurium, and sevoflurane, with sevoflurane discontinued 30 min before surgery completion. Routine intraoperative management included controlling anesthesia depth, fluid administration, and transfusions. Anesthesia depth was targeted to maintain a BIS value between 40 and 60, managed by adjusting anesthetic agents and administering vasoactive drugs such as ephedrine, keeping blood pressure fluctuations within 20% of baseline values. Additionally, mean arterial pressure (MAP) was kept above 65 mm Hg, and heart rate was maintained above 50 min−1. Electronic Supplementary Material eAppendix 2 details the specific hemodynamic management strategies. Mechanical ventilation parameters included a tidal volume of 6–8 mL·kg−1 body weight and respiratory rate adjustments to keep EtCO2 concentrations between 30 and 35 mm Hg. Upon completion of surgery, patients were extubated and initially observed in the postanesthesia care unit. Stable individuals were then transferred to general wards, whereas those deemed high risk or clinically unstable were admitted to the intensive care unit for continued monitoring and management. Postoperative analgesia was provided with self-controlled intravenous analgesia, using a combination of sufentanil and dezoxine. Other perioperative managements were provided per routine. Postoperative delirium The primary outcome was the occurrence of POD within 7 days following surgery. Postoperative delirium was diagnosed using the 3-Minute Diagnostic Interview for Confusion Assessment Method-Defined Delirium (3D-CAM) administered twice daily—morning (08:00–10:00 a.m.) and afternoon (4:00–6:00 p.m.)—from postoperative day 1 to day 7.13 The 7-day diagnostic window was chosen in accordance with clinical practice and is consistent with definitions supported by the International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9/10-CM)14 and prior studies.15 Delirium was further evaluated through review of clinical notes, nursing documentation, and psychiatric consultation records. Electronic Supplementary Material eTable 2 provides further details. To minimize assessment bias, delirium was evaluated under blinded conditions. Statistical analysis A predefined data analysis and statistical plan was submitted to the institutional review board and is publicly accessible at the Open Science Framework (10.17605/OSF.IO/YV5QH). Normally distributed continuous variables were presented as mean (SD) and compared using the independent samples t test. Non-normally distributed continuous variables were expressed as median (interquartile range) and compared using the Mann–Whitney U test. Categorical variables were reported as frequencies (percentages) and analyzed using the Chi square test or Fisher’s exact test, as appropriate. To examine the association between the duration of intraoperative reductions in CCA blood flow and the occurrence of POD, multivariable logistic regression models were constructed. Potential confounders were identified on the basis of a combination of prior studies and clinical reasoning.16,17 As previously described, CCA flow reductions were quantified as the cumulative duration within specific reduction ranges relative to baseline: < 30%, 30–40%, 40–50%, and > 50%. These variables were treated as continuous predictors in the regression model, with all patients contributing either zero or non-zero values for each range. All analyses were conducted using the full cohort (N = 210) with no loss of sample size for any predictor. To assess the linearity assumption of the logistic regression models, we used restricted cubic spline functions (with three knots) to model each continuous predictor. Likelihood ratio tests were used to compare spline-based models with linear models. None of the tested variables showed statistically significant nonlinearity and were therefore modeled as linear terms (ESM eTable 3). Adjusted odds ratios (aORs) with 95% confidence intervals (CIs) are reported after adjusting for relevant confounders, including American Society of Anesthesiologists (ASA) Physical Status classification and blood transfusion (categorical variables), as well as preoperative Montreal Cognitive Assessment (MoCA) scores, intraoperative blood loss, and intraoperative dose of ephedrine (continuous variables). Variance inflation factors (VIFs) were calculated for all covariates, and all values were below 5, indicating no concerning multicollinearity. Electronic Supplementary Material eTable 4 presents detailed VIF results. We also used restricted cubic spline models (with knots placed at the 10th, 50th, and 90th percentiles) to visualize the relationship between the duration of CCA blood flow reduction and POD risk.18 To formally test whether the effect of CCA flow reduction > 50% differed significantly from that of 40–50%, we conducted a postestimation comparison of the regression coefficients. To verify the robustness of the effect estimates, sensitivity analyses were conducted using both nonparametric bootstrap resampling and the Huber–White robust variance estimator. We used receiver operating characteristic (ROC) curve analysis to assess the discriminative ability of the logistic regression model. The area under the curve (AUC) and its 95% CI are reported. To mitigate potential residual confounding, we performed additional sensitivity analyses by expanding the original multivariable model to include the covariates age, preoperative serum albumin concentration, and intraoperative hypotension, quantified as the AUC for MAP < 65 mm Hg. Although these variables did not reach statistical significance in univariate analysis, they have been consistently associated with POD in prior studies.16,17,19 The sensitivity analyses were intended to assess the robustness of the primary findings. We conducted subgroup analyses on the basis of the following prespecified criteria: age (65–70 yr vs > 70 yr; 70 yr selected on the basis of clinical relevance and literature-based risk stratification), sex (male vs female), ASA Physical Status classification (II vs III, as assessed during preoperative evaluation), and duration of mechanical ventilation (≤ 240 vs > 240 min, with 240 min chosen on the basis of the median duration). Adjusted odds ratios and their corresponding 95% CIs were calculated for each subgroup analysis. An exploratory analysis was conducted to evaluate the association between intraoperative systemic hypotension and POD. Several blood-pressure-based metrics were examined, including the AUC for MAP < 65 mm Hg and the cumulative duration of MAP reductions from baseline (< 30%, 30–40%, 40–50%, and > 50%). Concordance between each hypotension definition and carotid hypoperfusion was evaluated using the Cohen’s κ statistic. Subsequently, we performed multivariable logistic regression analyses using each MAP-based parameter as the primary exposure. All models were adjusted for prespecified confounders. Model performance was assessed using the area under the ROC curve (AUC-ROC), Akaike information criterion (AIC), the Hosmer–Lemeshow test, and Brier scores. The Brier score ranges from 0 to 1, with lower values indicating better overall model performance in terms of both discrimination and calibration. All statistical analyses were performed using SPSS software version 25.0 (IBM Corp., Armonk, NY, USA) and R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). Electronic Supplementary Material eAppendix 3 provides the specific R packages and code. All statistical tests were two-tailed, and we considered a P value of < 0.05 statistically significant. To control for family-wise error rate due to multiple comparisons, we applied the Bonferroni correction to both P values and 95% CIs in the main analysis. Specifically, this correction was applied for at least six multiple comparisons involving different ranges of carotid blood flow changes. Sample size calculation Based on previous studies, we assumed the incidence of POD in older patients undergoing abdominal surgery to be approximately 15%.20 A preliminary pilot study conducted by our group indicated that patients who developed delirium experienced a mean duration in CCA blood flow reduction from baseline of 190 min, compared with 140 min among patients without delirium. Assuming an SD of 62 min, we calculated the required sample size to detect this 50 min difference with a power of 85% at a two-sided significance level of 0.05, using PASS 2025 software (NCSS, LLC, Kaysville, UT, USA). The calculated required sample size was 180 patients. To account for a potential dropout rate of approximately 15%, we planned to enroll 210 participants. Results Study population and patient characteristics Of the 242 patients initially screened, 210 met the eligibility criteria and were included in the final analysis. Figure 1 illustrates the screening and inclusion process, and Table 1 summarizes baseline characteristics of the study population. The mean (SD) age of the cohort was 69 (4) yr, and 98 patients (46.7%) were female. Of these, 141 (67.1%) were classified as ASA Physical Status III. Surgery was performed laparoscopically in 176 (83.8%) patients, while the remaining 34 (16.2%) underwent open laparotomy procedures. Table 1 summarizes additional baseline and intraoperative variables, and ESM eTable 5 details anesthesia-related parameters. Study flow chart OR = operating room Table 1Baseline and surgical characteristics of patientsVariableTotal cohort N = 210Non-POD N = 169POD N = 41P valueAge (yr), mean (SD)69 (4)69 (3)71 (5)0.08*Female, n/total N (%)98/210 (46.7%)79/169 (46.7%)19/41 (46.3%)0.62†BMI (kg·m−2), mean (SD)22.6 (2.9)22.7 (2.8)22.4 (3.6)0.30*Smoking history, n/total N (%)64/210 (30.5%)52/169 (30.8%)12/41 (29.3%)0.85†Alcohol use history, n/total N (%)53/210 (25.2%)38/169 (22.5%)15/41 (36.6%)0.06†Education years (yr), median [IQR]9 [6–12]9 [6–11]9 [6–12]0.83‡MMSE score, median [IQR]††23.5 [20–26]24 [20–26]23 [19.5–25]0.08‡MoCA score, median [IQR]††20 [17–23]20 [17–23]19 [14.5–21]0.03‡HADS score, median [IQR]††1 [0–3]0 [0–3]1 [0–5]0.16‡Complication, n/total N (%) Hypertension79/210 (37.6%)59/169 (34.9%)20/41 (48.8%)0.10† Diabetes mellitus36/210 (17.1%)32/169 (18.9%)4/41 (9.8%)0.18** Coronary heart disease13/210 (6.2%)8/169 (4.7%)5/41 (12.2%)0.14† Anemia29/210 (13.8%)23/169 (13.6%)6/41 (14.6%)0.87†Preoperative albumin (g·L−1), mean (SD)35.4 (4.0)35.6 (3.9)34.5 (4.3)0.13*Carotid plaque, n/total N (%)56/210 (26.7%)41/169 (24.4%)15/41 (36.6%)0.11†Carotid artery stenosis, n/total N (%)‡‡ Normal202/210 (96.2%)163/169 (96.5%)39/41 (95.1%)0.66** Mild stenosis8/210 (3.8%)6/169 (3.6%)2/41 (4.9%)ASA Physical Status, n/total N (%) Ⅱ69/210 (32.9%)63/169 (37.3%)6/41 (14.6%)0.01† Ⅲ141/210 (67.1%)106/169 (62.7%)35/41 (85.4%)NYHA class, n/total N (%) Ⅱ149/210 (71.0%)115/169 (68.0%)34/41 (82.9%)0.08† Ⅲ61/210 (29.0%)54/169 (32.0%)7/41 (17.1%)Operative approaches, n/total N (%) Laparotomy34/210 (16.2%)25/169 (14.8%)9/41 (22.0%)0.34† Laparoscopy176/210 (83.8%)144/169 (85.2%)32/41 (78.0%)Duration of surgery (min), mean (SD)206 (68)204 (68)215 (67)0.36*Duration of mechanical ventilation (min), mean (SD)238 (74)234 (72)256 (80)0.09*Blood loss (mL), median [IQR]100 [50–120]100 [50–100]100 [50–200]0.03‡Urine output (mL), median [IQR]200 [100–300]200 [100–300]250 [150–325]0.21‡Fluid volume (mL), median [IQR]1,800 [1,100–2,100]1,700 [1,100–2100]2,000 [1,600–2,600]0.09‡Blood transfusion, n/total N (%)13/210 (6.2%)4/169 (2.4%)9/41 (22.0%)< 0.001†Ephedrine (mg), median [IQR]0 [0–3]0 [0–4]0 [0–0]0.01‡Atropine use, n/total N (%)9/210 (4.3%)7/169 (4.1%)2/41 (4.9%)0.69**Malignant tumor, n/total N (%)129/210 (61.4%)100/169 (59.2%)29/41 (70.7%)0.17†Intraoperative complications, n/total N (%) Hypertension***32/210 (15.2%)26/169 (15.4%)6/41 (14.6%)0.91† Hypotension†††43/210 (20.5%)37/169 (21.9%)6/41 (14.6%)0.30† Arrhythmia‡‡‡17/210 (8.1%)13/169 (7.7%)4/41 (9.8%)0.66**CCA blood flow measurement Baseline CCA blood flow (mL·min−1), mean (SD)321.3 (62.4)318.3 (61.6)333.9 (64.8)0.15* Intraoperative minimum value (mL·min−1), mean (SD)145.3 (62.3)157.7 (60.9)94.1 (36.9)< 0.001* Maximum reduction (mL·min−1), mean (SD)174.4 (67.4)160.6 (59.4)231.4 (69.1)< 0.001* Maximum percentage reduction (%), mean (SD)53.5 (15.3)50.6 (14.9)65.8 (9.8)< 0.001*Values in bold are considered statistically significant*Student’s t test†Chi square test‡Mann-Whitney U test**Fisher’s exact test††The scale scores were collected before the surgery‡‡Mild carotid stenosis was defined as a luminal narrowing of less than 50% without significant hemodynamic compromise. No patients in this cohort had moderate or severe carotid stenosis; only normal and mild stenosis are reported in the table.***Hypertension was defined as systolic blood pressure > 160 mm Hg or 20% of the baseline value, or mean blood pressure > 110 mm Hg for 5 min that required treatments†††Hypotension was defined as systolic blood pressure < 90 mm Hg or 70% of the baseline value, or mean blood pressure < 65 mm Hg for 5 min that required treatments‡‡‡Arrhythmia was defined as the new onset of any abnormal cardiac rhythm, including bradycardia, tachycardia, atrial fibrillation, or ventricular arrhythmias, documented by intraoperative monitoringASA = American Society of Anesthesiologists; CCA = common carotid artery; BMI = body mass index; HADS = Hospital Anxiety and Depression Scale; IQR = interquartile range; MMSE = Mini-Mental State Examination; MoCA = Montreal Cognitive Assessment; NYHA = New York Heart Association; POD = postoperative delirium; SD = standard deviation Baseline and surgical characteristics of patients Values in bold are considered statistically significant *Student’s t test †Chi square test ‡Mann-Whitney U test **Fisher’s exact test ††The scale scores were collected before the surgery ‡‡Mild carotid stenosis was defined as a luminal narrowing of less than 50% without significant hemodynamic compromise. No patients in this cohort had moderate or severe carotid stenosis; only normal and mild stenosis are reported in the table. ***Hypertension was defined as systolic blood pressure > 160 mm Hg or 20% of the baseline value, or mean blood pressure > 110 mm Hg for 5 min that required treatments †††Hypotension was defined as systolic blood pressure < 90 mm Hg or 70% of the baseline value, or mean blood pressure < 65 mm Hg for 5 min that required treatments ‡‡‡Arrhythmia was defined as the new onset of any abnormal cardiac rhythm, including bradycardia, tachycardia, atrial fibrillation, or ventricular arrhythmias, documented by intraoperative monitoring ASA = American Society of Anesthesiologists; CCA = common carotid artery; BMI = body mass index; HADS = Hospital Anxiety and Depression Scale; IQR = interquartile range; MMSE = Mini-Mental State Examination; MoCA = Montreal Cognitive Assessment; NYHA = New York Heart Association; POD = postoperative delirium; SD = standard deviation Postoperative delirium occurred in 41 (19.5%) patients, whereas the remaining 169 (80.5%) patients did not develop delirium. Electronic Supplementary Material eTable 2 summarizes the daily postoperative incidence of delirium. As Table 1 summarizes, baseline carotid artery blood flow was similar between the two groups, but intraoperative reductions in carotid blood flow were significantly greater among patients who developed POD (P < 0.001). No patients in the cohort experienced stroke or persistent neurologic deficits during hospitalization. Primary analysis As Table 2 shows, after adjusting for ASA Physical Status classification, preoperative MoCA scores, intraoperative blood loss, blood transfusion, and cumulative ephedrine dose, both 40–50% and > 50% reductions from baseline were significantly associated with an increased risk of POD. Electronic Supplementary Material eTable 6 provides unadjusted odds ratios and corresponding 95% CIs. Building upon the primary findings, further analysis showed that, after adjustment for confounding variables, the duration of intraoperative carotid blood flow reduction > 40% from baseline was associated with an increased risk of POD (aOR, 2.24 per 10-min increment; 95% CI, 1.51 to 3.31; Bonferroni-adjusted P < 0.001), as Table 3 (Model 1) shows. Figure 2 shows duration-dependent association between carotid blood flow reduction > 40% baseline and higher risk of POD. Table 2Association between intraoperative carotid blood flow variables and postoperative deliriumVariableNon-POD N = 169*POD N = 41*aOR†95% CI†P valueBonferroni-adjusted CI‡Bonferroni-adjusted P value**Duration of reduced CCA flow (% from baseline) < 30%105 [80–140]90 [60–140]0.980.82 to 1.070.080.80 to 1.110.50 30–40%20 [10–4]30 [20–45]1.000.73 to 1.360.150.61 to 1.480.92 40–50%20 [0–30]40 [20–60]1.621.18 to 2.22< 0.0011.07 to 2.43< 0.001 > 50%0 [0–20]60 [50–80]2.251.59 to 3.20< 0.0011.38 to 3.44< 0.001Duration of increased CCA flow (above baseline)55 [20–92.5]0 [0–20]0.580.40 to 0.83< 0.0010.35 to 0.92< 0.001Mean intraoperative CCA flow††246.7 [220.3–296.8]229.9 [190.5–277.2]0.910.86 to 0.970.0040.83 to 0.990.03CCA blood flow duration variables were modeled as continuous variables, analyzed per 10-min increment. Each patient contributed a value representing the total time spent in the specified range during surgery. Values in bold were considered statistically significant. aORs with 95% CIs, and corresponding P values were derived from multivariable logistic regression models. The dependent variable was the occurrence of POD, and covariates included ASA Physical Status classification, preoperative MoCA score, intraoperative blood loss, blood transfusion, and cumulative intraoperative ephedrine dose. Risk differences are for the POD group relative to the non-POD group.*Values for carotid blood flow reduction or increase represent total duration in minutes. Mean carotid blood flow values are reported in mL·min−1. Data are presented as median [IQR].†The aORs and 95% CIs for duration-related and flow-related variables were calculated per 10-min and per 10 mL·min−1 increments, respectively‡Bonferroni correction was applied assuming six comparisons; the confidence level was set at 1 − (0.05 ÷ 6) = 99.17% for each interval**Bonferroni correction was applied assuming six comparisons, adjusted P values were calculated as Padj = min (P × 6, 1). A Bonferroni-adjusted P value < 0.05 was considered indicative of statistical significance††Mean intraoperative carotid blood flow was calculated by dividing the sum of all reported carotid blood flow values by the total duration of mechanical ventilationaOR = adjusted odds ratio; ASA = American Society of Anesthesiologists; CCA = common carotid artery; CI = confidence interval; IQR = interquartile range; MoCA = Montreal Cognitive Assessment; OR = odds ratio; POD = postoperative delirium Association between intraoperative carotid blood flow variables and postoperative delirium CCA blood flow duration variables were modeled as continuous variables, analyzed per 10-min increment. Each patient contributed a value representing the total time spent in the specified range during surgery. Values in bold were considered statistically significant. aORs with 95% CIs, and corresponding P values were derived from multivariable logistic regression models. The dependent variable was the occurrence of POD, and covariates included ASA Physical Status classification, preoperative MoCA score, intraoperative blood loss, blood transfusion, and cumulative intraoperative ephedrine dose. Risk differences are for the POD group relative to the non-POD group. *Values for carotid blood flow reduction or increase represent total duration in minutes. Mean carotid blood flow values are reported in mL·min−1. Data are presented as median [IQR]. †The aORs and 95% CIs for duration-related and flow-related variables were calculated per 10-min and per 10 mL·min−1 increments, respectively ‡Bonferroni correction was applied assuming six comparisons; the confidence level was set at 1 − (0.05 ÷ 6) = 99.17% for each interval **Bonferroni correction was applied assuming six comparisons, adjusted P values were calculated as Padj = min (P × 6, 1). A Bonferroni-adjusted P value < 0.05 was considered indicative of statistical significance ††Mean intraoperative carotid blood flow was calculated by dividing the sum of all reported carotid blood flow values by the total duration of mechanical ventilation aOR = adjusted odds ratio; ASA = American Society of Anesthesiologists; CCA = common carotid artery; CI = confidence interval; IQR = interquartile range; MoCA = Montreal Cognitive Assessment; OR = odds ratio; POD = postoperative delirium Table 3Association between duration of intraoperative carotid blood flow reduction > 40% and postoperative deliriumVariableNon-POD N = 169*POD N = 41*aORb†95% CIb†P valueBonferroni-adjusted CI‡Bonferroni-adjusted P value**Model 1Duration of reduced CCA flow > 40% from baseline25 [10–45]110 [80–140]2.241.51 to 3.31< 0.0011.38 to 3.62< 0.001Model 2Duration of reduced CCA flow > 40% from baseline25 [10–45]110 [80–140]1.781.45 to 2.19< 0.001——Model 1 is the primary multivariable logistic regression adjusting for ASA Physical Status classification (II as reference), preoperative MoCA score, intraoperative blood loss, blood transfusion, and ephedrine dose. Model 2 is a sensitivity analysis incorporating additional covariates, including age, preoperative serum albumin level, and intraoperative hypotension (quantified as the AUC for MAP < 65 mm Hg), to assess the robustness of the association. As this model assessed the robustness of the primary findings, no Bonferroni correction was applied.*Values for carotid blood flow reduction or increase represent total duration in minutes. Data are presented as median [IQR].†The odds ratios and 95% CI for duration-related variables were calculated per 10 min‡Bonferroni correction was applied assuming six primary comparisons; the confidence level was set at 1 − (0.05 / 6) = 99.17% for each interval**Bonferroni correction was applied assuming six primary comparisons, adjusted P values were calculated as Padj = min (P × 6, 1). A Bonferroni-adjusted P value < 0.05 was considered indicative of statistical significance.aOR = adjusted odds ratio; ASA = American Society of Anesthesiologists; AUC = area under the curve; CCA = common carotid artery; CI = confidence interval; IQR = interquartile range; MAP = mean arterial pressure; MoCA = Montreal Cognitive Assessment; POD = postoperative delirium Association between duration of intraoperative carotid blood flow reduction > 40% and postoperative delirium Model 1 is the primary multivariable logistic regression adjusting for ASA Physical Status classification (II as reference), preoperative MoCA score, intraoperative blood loss, blood transfusion, and ephedrine dose. Model 2 is a sensitivity analysis incorporating additional covariates, including age, preoperative serum albumin level, and intraoperative hypotension (quantified as the AUC for MAP < 65 mm Hg), to assess the robustness of the association. As this model assessed the robustness of the primary findings, no Bonferroni correction was applied. *Values for carotid blood flow reduction or increase represent total duration in minutes. Data are presented as median [IQR]. †The odds ratios and 95% CI for duration-related variables were calculated per 10 min ‡Bonferroni correction was applied assuming six primary comparisons; the confidence level was set at 1 − (0.05 / 6) = 99.17% for each interval **Bonferroni correction was applied assuming six primary comparisons, adjusted P values were calculated as Padj = min (P × 6, 1). A Bonferroni-adjusted P value < 0.05 was considered indicative of statistical significance. aOR = adjusted odds ratio; ASA = American Society of Anesthesiologists; AUC = area under the curve; CCA = common carotid artery; CI = confidence interval; IQR = interquartile range; MAP = mean arterial pressure; MoCA = Montreal Cognitive Assessment; POD = postoperative delirium Predicted risk of postoperative delirium by total cumulative time of reduced common carotid artery blood flow. Predicted risk of postoperative delirium with 95% confidence intervals as a function of total cumulative minutes of common carotid artery blood flow reduction > 40% from baseline. The curve was based on logistic regression models using restricted cubic splines. The model was adjusted for preoperative Montreal Cognitive Assessment score, intraoperative blood loss, blood transfusion, and intraoperative dose of ephedrine. CCA = common carotid artery; POD = postoperative delirium Model discriminative ability was evaluated for the primary models and showed excellent performance, with an AUC-ROC of 0.96 (95% CI, 0.93 to 0.98; P < 0.001). Figure 3 provides further details. The robustness of the findings was confirmed through bootstrap resampling and robust variance estimation (ESM eTables 7–11). Comparison of receiver operating characteristic curves for different definitions of carotid hypoperfusion and intraoperative hypotension. Receiver operating characteristic curves were constructed from multivariable logistic regression models predicting postoperative delirium, each using one of four primary exposure variables: 1) duration of carotid blood flow reduction > 40% from baseline, (2) duration of carotid blood flow reduction > 50% from baseline, 3) area under the curve for mean arterial pressure < 65 mm Hg, and 4) duration of mean arterial pressure reduction > 50% from baseline. All models were adjusted for ASA Physical Status classification, preoperative Montreal Cognitive Assessment score, intraoperative blood loss, blood transfusion, and ephedrine use. The grey dashed diagonal line represents the reference line for random classification (AUC = 0.5). ASA = American Society of Anesthesiologists; AUC = area under the curve; CCA = common carotid artery; MAP = mean arterial pressure; ROC = receiver operating characteristic An exploratory comparison suggested that reductions greater than 50% may be associated with a stronger effect than 40–50%, although this difference was not statistically significant after Bonferroni correction (adjusted P = 0.15). Secondary analysis In the secondary analysis, a longer duration of intraoperative increases in carotid blood flow relative to baseline was associated with a lower risk of POD (Table 2). Specifically, for every 10-min increase in the duration of increased carotid blood flow was associated with an aOR of 0.58 (95% CI, 0.40 to 0.83; Bonferroni-adjusted P < 0.001). Additionally, each 10 mL·min−1 increase in mean carotid blood flow corresponded to an aOR of 0.91 (95% CI, 0.86 to 0.97, Bonferroni-adjusted P = 0.03). Electronic Supplementary Material eTable 12 provides further details on associations between increased blood flow at different thresholds and POD. The robustness of the findings was confirmed through bootstrap resampling and robust variance estimation (ESM eTables 13 and 14). Subgroup analysis Electronic Supplementary Material eFig. 2 presents the post hoc subgroup analyses according to age, sex, ASA Physical Status classification, and duration of mechanical ventilation. Consistent with the findings from the primary analysis, the duration of intraoperative CCA blood flow reduction greater than 40% from baseline remained independently associated with increased POD risk across all examined subgroups. Sensitivity analysis As Table 3 (Model 2) shows, additional covariates including age, preoperative serum albumin level, and intraoperative hypotension were incorporated into the regression model. After adjustment, the duration of CCA blood flow reduction > 40% from baseline remained associated with the risk of POD (aOR, 1.78 per 10-min increment; 95% CI, 1.45 to 2.19; adjusted P < 0.001). Exploratory analysis of model metrics: carotid flow reduction versus mean arterial pressure reduction As ESM eTables 15 and 16 present, concordance analyses showed no agreement between CCA flow and MAP reduction, except when MAP fell by > 40% and CCA flow by > 50%, where Cohen’s κ indicated only slight agreement (κ = 0.08; P = 0.01). Notably, among the 175 hypoperfusion events identified in our cohort, 80% (n = 139) occurred when MAP remained > 65 mm Hg (ESM eTable 6), suggesting that significant carotid flow reduction frequently occurred despite normotension. In ESM eTable 7, only the duration of MAP reduction > 50% was associated with POD before correction (aOR, 2.00; 95% CI, 1.06 to 3.73; P = 0.03), but this did not remain significant after Bonferroni adjustment (P = 0.16). Electronic Supplementary Material eTable 18 shows that, under acceptable calibration (Hosmer–Lemeshow P > 0.05), CCA-based models outperformed MAP-based models with lower AIC and Brier scores. Consistently, ROC curve analyses (Fig. 3) further showed the superior discriminative ability of CCA-based models compared with MAP-based models. Discussion Our study found that a longer reduction in carotid blood flow during surgery, particularly a reduction greater than 40% from baseline, was associated with an increased risk of POD. Conversely, prolonged periods of increased carotid blood flow above baseline were associated with a reduced risk of delirium. These associations underscore the potential importance of intraoperative cerebral perfusion patterns in POD development. Several previous studies have highlighted the importance of intraoperative hemodynamic fluctuations, including alterations in cerebral perfusion, as a key factor contributing to POD.21,22 By specifically examining carotid artery blood flow, our findings provide quantitative evidence regarding how the magnitude and duration of intraoperative perfusion changes may influence delirium risk. Supporting our results, a prior study reported significantly lower carotid artery blood flow velocities in patients who developed delirium following cardiac surgery compared with patients without delirium.23 Additionally, reduced CCA blood flow has been associated with cognitive impairment among older patients with heart failure,24 indicating a broader relationship between carotid artery perfusion and neurocognitive outcomes. Our study expands upon these findings by showing that intraoperative reductions in CCA blood flow may serve as a risk factor for POD among older patients undergoing noncardiac surgery. The observed link between reduced carotid artery blood flow and POD may be attributed to impaired cerebral perfusion and subsequent disturbances in brain function. Cognition-related regions such as the medial temporal lobes and frontal-subcortical circuits are particularly vulnerable to hypoperfusion and oxygen deficiency.25 Even brief intraoperative reductions in cerebral blood flow can trigger neuronal stress and cognitive impairment.26 Additionally, hypoperfusion may increase oxidative stress via reactive oxygen species production,27 and activate microglia-driven neuroinflammation, further contributing to neuronal injury.28,29 Ageing amplifies these effects through neuronal loss, impaired vascular regulation, and blood–brain barrier disruption.30 Nonetheless, prior intraoperative monitoring studies report mixed associations with POD. Transcranial Doppler studies generally align with our findings and link reduced cerebral flow with delirium, although most evidence derives from cardiac surgery with limited data in noncardiac surgical cohorts.31,32 In contrast, cerebral oximetry has yielded inconsistent associations with delirium.33–35 In this context, our study used an alternative approach by measuring extracranial carotid inflow and identified an association between cerebral hypoperfusion and increased POD risk. Given the multifactorial nature of delirium, these observations should be interpreted with caution. Additionally, longer durations of intraoperative increases in carotid blood flow relative to baseline were associated with a decreased risk of delirium in this study, suggesting a potential neuroprotective effect. Evidence on intraoperative cerebral hyperperfusion and delirium remains limited and inconsistent. One cohort study found no significant association using cerebral oxygen saturation monitoring,11 while another study of cardiac surgery linked a MAP above the upper limit of autoregulation to higher delirium risk, possibly due to hyperperfusion-related damage.36 In our cohort, only 3.8% of patients showed > 130% increases from baseline, which likely remained within autoregulatory limits and may have mitigated ischemic injury. Further studies are needed to clarify the neurologic effects of more substantial cerebral perfusion increases. These findings underscore the importance of intraoperative cerebral perfusion monitoring. Although no cases of overt stroke or persistent neurologic deficits were observed in this cohort, the absence of such events should not be interpreted as evidence that intraoperative carotid hypoperfusion lacks clinical relevance. Longer-term follow-up studies are needed to evaluate potential subclinical or delayed neurologic consequences. Nevertheless, real-time intraoperative monitoring of carotid perfusion remains challenging in routine practice. Carotid blood flow during surgery is affected by multiple factors, including anesthesia, ventilation settings, and surgical manipulation.12 Ultrasound technology provides dynamic, noninvasive bedside evaluation of carotid blood flow parameters for surgical patients. Additionally, during laparoscopic surgery, fluctuations in arterial CO2 partial pressure can lead to cerebral vasodilation.37,38 Simply measuring blood flow velocity may not accurately reflect the true perfusion status, and therefore, we have achieved more precise measurement of blood flow volume through the application of VFI technology.9 Vector flow imaging is an advanced ultrasound technique that captures velocity components from multiple angles to reconstruct actual blood flow vectors.39,40 Prior studies have reported that VFI offers acceptable accuracy for quantifying carotid blood flow, with in vivo measurement errors ranging from 10% to 20% when compared with the gold standard, phase-contrast magnetic resonance imaging. Vector flow imaging has also shown superior consistency relative to conventional Doppler methods, particularly in complex flow environments such as near the carotid bifurcation.41 Vector flow imaging has been proven to have significant clinical value in flow measurement. An additional finding was that, although most patients were not diagnosed with intraoperative hypotension using noninvasive arterial blood pressure measurement, a considerable proportion exhibited carotid hypoperfusion. This dissociation suggests impaired cerebrovascular regulation, particularly in older adults. Age-related factors such as reduced baseline cerebral blood flow, increased vascular resistance, and a narrowed autoregulatory range, as well as chronic hypertension elevating the lower autoregulatory threshold, may render the brain more vulnerable to hypoperfusion despite normal systemic pressures.42–44 General anesthesia further impairs cerebrovascular responsiveness by reducing sympathetic tone and baroreflex sensitivity.45,46 Other factors such as positive pressure ventilation and reduced venous return may also contribute.47 In our study, models incorporating carotid flow reduction showed better predictive performance for delirium than MAP-based models, underscoring the limitations of standard blood pressure monitoring and supporting the value of direct carotid flow assessment. In addition to the primary findings, several baseline differences between the POD and non-POD groups warrant further consideration. One notable observation was that patients without POD received a higher cumulative dose of ephedrine during surgery. Nevertheless, the overall incidence of intraoperative hypotension did not significantly differ between groups (P = 0.30), suggesting this was not because of divergent blood pressure management. The discrepancy may be due to the modest sample size or individual variability in responsiveness to ephedrine. Moreover, cumulative ephedrine dose was significantly associated with POD in only a few adjusted models, suggesting that the role of vasopressor administration in the development of delirium should be interpreted with caution. This finding requires further validation in studies with larger sample sizes. Although several prior studies have reported an association between intraoperative blood transfusion and POD,16,48,49 this relationship was not consistently observed across all models in our study. Transfusion remained statistically significant in only a subset of models, and the wide CIs observed in other models suggest limited precision. These findings indicate that transfusion may reflect downstream physiologic deterioration, such as blood loss or metabolic stress, rather than serve as an independent determinant of delirium risk. In summary, our findings underscore the clinical value of individualized intraoperative management strategies aimed at preserving adequate carotid artery perfusion to mitigate the risk of POD among older patients undergoing surgery. Future studies are necessary to further validate these associations in broader surgical populations and to determine whether targeted interventions that maintain or enhance cerebral perfusion during surgery can improve other neurologic outcomes and long-term cognitive health. Limitations Several limitations of this study should be acknowledged. First, this was a single-centre study, potentially limiting the generalizability of our findings; additionally, the restricted setting may have constrained the diversity of patient characteristics and practice patterns, potentially introducing unrecognized or residual confounding. Second, only the left CCA was monitored owing to technical constraints. Although baseline concordance between sides was high, intraoperative asymmetry cannot be excluded. Third, although we averaged multiple measurements to reduce error, interobserver variability was not formally assessed. Fourth, carotid flow was reported every 10 min, potentially missing transient fluctuations. This intermittent sampling assumes a uniform risk within each interval, which may have contributed to the relatively large effect sizes observed (aOR, 1.78–2.24 per 10-min increment). While the association was robust across sensitivity analyses and thresholds, the magnitude of association should be interpreted with caution and not directly generalized as a clinical risk estimate. Future studies should adopt more frequent, real-time monitoring and evaluate active modulation strategies. Finally, while sample size estimation was based on group differences in a continuous variable, the final analysis used multivariable logistic regression for the binary outcome (POD). Although robustness was confirmed via bootstrap and robust variance estimation, this methodologic shift should be considered when interpreting results. Conclusions In older patients undergoing abdominal surgery, persistent intraoperative CCA blood flow reductions > 40% from baseline were associated with an increased risk of POD. Conversely, sustained increases in carotid blood flow were associated with a lower risk. These observed associations support the potential value of monitoring cerebral perfusion during surgery. Further research is required to confirm these findings in other surgical populations and to explore clinical approaches for optimizing cerebral perfusion management during surgery. Supplementary Information Below is the link to the electronic supplementary material. ELECTRONIC SUPPLEMENTARY MATERIAL (PDF 558 kb) Acknowledgments Author contributions E Wang contributed to the conception and design of the study, analysis and interpretation of data, and drafting of the manuscript. Xingyang Liu contributed to the conception and design of the study; acquisition, analysis, and interpretation of data; and the drafting of the manuscript. Chunyan Ye participated in consultation during the study design phase and contributed to drafting of the manuscript. Jinghan Wu and Yingqi Li were responsible for participant recruitment, data collection, and data processing. All authors critically revised the manuscript for important intellectual content, approved the final version, and agreed to be accountable for all aspects of the work. Acknowledgements The authors sincerely thank Shuang Liang for his invaluable help during the revision process of this manuscript. Disclosures The authors declare that they conducted the research without any commercial or financial relationships that could be considered a potential conflict of interest. Funding statement This work was supported by the National Natural Science Foundation of China (82170291) and the Hunan Provincial Key Natural Science Foundation of China (Grant 2024JJ3051). Generative AI tools use None. Editorial responsibility This submission was handled by Dr. Stephan K. W. Schwarz, Editor-in-Chief, Canadian Journal of Anesthesia/Journal canadien d’anesthésie. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials ELECTRONIC SUPPLEMENTARY MATERIAL (PDF 558 kb)