rSVO₂ and Postoperative Delirium Prediction

Overview

This prospective observational cohort study will enroll 300 patients aged ≥65 years undergoing elective noncardiac, non-neurosurgical procedures . Baseline rSVO₂ will be measured preoperatively using Near-Infrared Spectroscopy (NIRS). Intraoperative rSVO₂ will be monitored, recording the absolute lowest value and the rate of rSVO₂ decrease (\[baseline - lowest\]/baseline × 100). Postoperative delirium (POD) will be assessed daily for 7 days post-surgery using the Confusion Assessment Method (CAM). Frailty will be assessed using the Fried Frailty Phenotype. Logistic regression models, adjusted for age, ASA physical status, Charlson Comorbidity Index (CCI), hemoglobin, and intraoperative factors, will evaluate rSVO₂ metrics as predictors of POD. Linear regression will assess factors influencing the rate of rSVO₂ decrease. Pearson correlation and multivariate regression will analyze associations between POD risk factors. Receiver Operating Characteristic (ROC) analysis will compare the predictive accuracy of rSVO₂ metrics.

Study Design and Population This prospective observational cohort study will be conducted at Wonkwang University School of Medicine Hospital from May 2025, upon approval by the Institutional Review Board (IRB#2025-04-010-0021). Written informed consent will be obtained from all participants. Patients aged ≥65 years scheduled for elective noncardiac, non-neurosurgical procedures (e.g., general, orthopedic, urologic, vascular) under general anesthesia with an expected hospital stay ≥2 nights will be included. Exclusion criteria will include dementia, stroke, cerebral tumors, chronic use of antipsychotics or anticholinergics, inability to perform cognitive assessments, or non-English first language. Data Collection Baseline rSVO₂ Measurement: Baseline rSVO₂ will be measured preoperatively using the O3® Regional Oximetry system (Masimo Corp., Irvine, CA, USA). Sensors will be placed on the patient's forehead while seated, breathing room air, and resting. The stable rSVO₂ value after 60 seconds will be recorded. Intraoperative rSVO₂ Monitoring: Intraoperative rSVO₂ will be continuously monitored using the same O3® system. Two metrics will be extracted: (1) the absolute lowest intraoperative rSVO₂ value, and (2) the rate of rSVO₂ decrease, calculated as: $$\\text{rate of rSVO₂ decrease (%)} = \\frac{\\text{Baseline rSVO₂} - \\text{Absolute lowest intraoperative rSVO₂}}{\\text{Baseline rSVO₂}} \\times 100$$ Postoperative Delirium Assessment: POD will be assessed daily for 7 days post-surgery using the Confusion Assessment Method (CAM) \[12\]. A positive CAM on any day will indicate POD. Postoperative Patient Destination: This will be recorded as transfer to the ICU or general ward post-surgery. Covariates: These will include age, sex, American Society of Anesthesiologists (ASA) physical status, Charlson Comorbidity Index (CCI), preoperative hemoglobin, frailty, and intraoperative factors: vital signs (mean arterial pressure \[MAP\], heart rate), blood loss (mL), transfusion (yes/no), fluid administration (crystalloid and colloid volumes in mL), and vasopressor use (norepinephrine, phenylephrine, or none). Frailty will be assessed using the Fried Frailty Phenotype, defined by five criteria: unintentional weight loss (\>4.5 kg in the past year), self-reported exhaustion, weak grip strength (measured by dynamometer), slow walking speed (timed 4.5-meter walk), and low physical activity (assessed via patient-reported activity levels) \[16\]. Patients meeting 0 criteria will be classified as robust, 1-2 as pre-frail, and ≥3 as frail. Frailty assessments will be conducted preoperatively by trained study personnel. Sample Size Calculation The sample size will be determined to address our primary objective of comparing the predictive performance of three rSVO₂ metrics (baseline rSVO₂, absolute lowest intraoperative rSVO₂, and rate of rSVO₂ decrease) for postoperative delirium (POD) in patients aged ≥65 years. Two complementary approaches will be used: ROC AUC Comparison: From pilot data, baseline rSVO₂ yielded an AUC of 0.78, while the rate of rSVO₂ decrease had an AUC of 0.70-defining a minimum clinically relevant ΔAUC of 0.08. Using a two-sided α of 0.05, 80% power, and an anticipated POD incidence of 30% (event:non-event ratio ≈0.43:1), investigators will employ the power.roc.test() function in R's pROC package. This analysis indicates that 75 total patients (approximately 22 with POD and 53 without POD) will be sufficient to detect the specified AUC difference. Multivariable Logistic Regression (EPV Rule): To ensure stable adjustment for covariates in multivariable models, investigators plan to include eight predictors: the three rSVO₂ metrics plus intraoperative blood loss, hypotension, vasopressor use, patient age, and ASA physical status. Applying the "events per variable" rule (minimum 10 events per predictor) will require 80 POD events. Assuming a 30% POD rate, this corresponds to 267 total patients (80 ÷ 0.30). Finally, to accommodate an anticipated 10% rate of missing data or exclusions, investigators will inflate the larger of these estimates by dividing by 0.90, yielding a target enrollment of 297 patients, which will be rounded to 300 participants. Statistical Analysis Continuous variables will be reported as mean (SD) or median \[IQR\], and categorical variables as number (%). Normality will be assessed using the Shapiro-Wilk test. Comparisons will use Student's t-test or Mann-Whitney U test for continuous variables and chi-square or Fisher's exact test for categorical variables. Logistic regression models, adjusted for age, ASA status, CCI, hemoglobin, frailty, and intraoperative factors, will evaluate rSVO₂ metrics as predictors of POD. Linear regression will assess factors influencing the rate of rSVO₂ decrease. Pearson correlation and multivariate regression will analyze associations between POD and risk factors. Receiver Operating Characteristic (ROC) curve analysis will compare the predictive accuracy of rSVO₂ metrics. A P-value \<0.05 will be considered significant. Analyses will be performed using SPSS v29.0 (IBM, NY) and R v4.3.0 (R Foundation for Statistical Computing).