Authors: Kang M et al.
Anesthesiology. December 2025.
Summary
This prospective observational study evaluated whether eye movement–based saccade tasks can serve as a noninvasive and accurate tool to predict postoperative delirium (POD) in elderly patients undergoing arthroplasty. Given the limitations of existing cognitive screening tests and molecular biomarkers, the authors sought to determine whether saccadic performance could better identify patients at risk for POD.
The study enrolled 316 patients aged 65 years or older undergoing elective arthroplasty. Preoperative assessments included standard cognitive screening tools (MMSE and MoCA), serum and cerebrospinal fluid biomarkers (including neurofilament light chain), and a battery of saccadic eye movement tasks. POD was diagnosed postoperatively using the Confusion Assessment Method by trained evaluators. The predictive performance of these assessment modalities was compared.
Postoperative delirium occurred in 8.2% of patients. Patients who developed POD demonstrated significantly worse performance on multiple saccadic parameters, including prolonged reaction times, increased primary saccade error, altered saccadic gain, and reduced peak velocity in anti-saccade and memory-guided saccade tasks. While MMSE, MoCA, and serum NfL levels also differed between POD and non-POD patients, their predictive accuracy was substantially lower.
In predictive modeling, saccadic parameters outperformed traditional cognitive screening tests and molecular biomarkers. Logistic regression models using saccadic variables achieved good discrimination, while a multilayer perceptron machine learning model further improved prediction accuracy. These findings suggest that saccadic testing captures subtle deficits in executive function, attention, and neural network integrity that may precede clinically apparent delirium.
The authors conclude that saccadic eye movement testing represents a promising, objective, and noninvasive behavioral biomarker for identifying elderly surgical patients at elevated risk for postoperative delirium, with potential applications in preoperative risk stratification and targeted prevention strategies.
Key Points
Postoperative delirium occurred in approximately 8% of elderly arthroplasty patients.
Saccadic eye movement parameters differed significantly between patients with and without POD.
Saccade-based models predicted POD more accurately than MMSE, MoCA, or neurofilament light chain levels.
Machine learning models further enhanced predictive performance using saccadic data.
Saccadic testing may serve as a practical, noninvasive tool to improve preoperative delirium risk stratification.
Thank you for allowing us to review and summarize this innovative predictive study from Anesthesiology.