Preoperative cognitive status is a major contributing factor to postoperative delirium in cardiac surgery: A post-hoc analysis of a prospective study
Postoperative delirium, Cognitive function, Serum neurofilament light, alpha-band power
Published online: Aug 31 2025
Abstract
Background: We previously showed that in cardiac surgery preoperative cognitive status of the patient was the most predictive factor of postoperative delirium (POD), independently of other clinical, biological and neurophysiological variables included in different multivariable adjusted logistic regression models.
Objective: We aimed to find the best fitted model among our previously reported models.
Design and setting: Prospective study performed in a tertiary hospital
Methods: This is a post-hoc analysis of a study including 220 patients undergoing elective cardiac surgery with cardiopulmonary bypass. The day before surgery, patients were submitted to a battery of neurocognitive tests. Blood sample was collected before the start of anesthesia for neurofilament light (NfL) measurement. After induction of anesthesia a 32-channel electroenecephalogram (EEG) was recorded. Alpha-band signal was recorded over frontal electrodes. The mean power of alpha-band activity was extracted from individual frequency spectra. Patients were screened for POD until discharge from the hospital.
Main outcome measures: Association between the studied exposures and the incidence of POD were assessed by fitting multivariable adjusted logistic regression models. Comparisons between models were assessed using the Likelihood Ratio Test and the Akaike Information Criterion.
Results: In a model where patient’s baseline cognition was not evaluated, postinduction mean frontal alpha power and baseline serum NfL concentration both significantly predicted POD. However, in multivariable regression models where baseline cognition was available, it was consistently the only variable that predicted POD. Addition of postinduction mean alpha power and baseline serum NfL concentration to a model where baseline cognition was known, did not statistically improve the predictive value.
Conclusions: Whenever baseline cognition is included in a multivariable model predicting POD, it remains the only significant variable. This highlights the importance of the evaluation of baseline cognition whatever clinical, neurophysiologic or biological information would be used.
Trial registration: NCT03706989
