Delirium, Caffeine, and Perioperative Cortical Dynamics

Abstract

Delirium is a major public health issue associated with considerable morbidity and mortality, particularly after surgery. While the neurobiology of delirium remains incompletely understood, emerging evidence suggests that cognition requires close proximity to a system state called criticality, which reflects a point of dynamic instability that allows for flexible access to a wide range of brain states. Deviations from criticality are associated with neurocognitive disorders, though the relationship between criticality and delirium has not been formally tested. This study tested the primary hypothesis that delirium in the postanesthesia care unit would be associated with deviations from criticality, based on surrogate electroencephalographic measures. As a secondary objective, the impact of caffeine was also tested on delirium incidence and criticality. To address these aims, we conducted a secondary analysis of a randomized clinical trial that tested the effects of intraoperative caffeine on postoperative recovery in adults undergoing major surgery. In this substudy, whole-scalp (16-channel) electroencephalographic data were analyzed from a subset of trial participants (n = 55) to determine whether surrogate measures of neural criticality - (1) autocorrelation function of global alpha oscillations and (2) topography of phase relationships via phase lag entropy - were associated with delirium. These measures were analyzed in participants experiencing delirium in the postanesthesia care unit (compared to those without delirium) and in participants randomized to caffeine compared to placebo. Results demonstrated that autocorrelation function in the alpha band was significantly reduced in delirious participants, which is important given that alpha rhythms are postulated to play a vital role in consciousness. Moreover, participants randomized to caffeine demonstrated increased alpha autocorrelation function concurrent with reduced delirium incidence. Lastly, the anterior-posterior topography of phase relationships appeared most preserved in non-delirious participants and in those receiving caffeine. These data suggest that early postoperative delirium may reflect deviations from neural criticality, and caffeine may reduce delirium risk by shifting cortical dynamics toward criticality.

Keywords: cognitive aging; cognitive dysfunction; delirium; electroencephalography; postoperative cognitive changes.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2021 Kim, McKinney, Brooks, Mashour, Lee and Vlisides.

Figures

FIGURE 1

Whole-scalp, 16-channel electroencephalographic (EEG) data were obtained from all participants. Following preprocessing, autocorrelation function and phase lag entropy (PLE) were analyzed from the EEG waveform data. The autocorrelation of global synchronization was calculated with various time lags. For PLE, the phase difference was first calculated and binarized to compute the distribution of the patterns. PLE was then defined with the Shannon entropy of the probability of the patterns (see text for references).

FIGURE 2

Study flow diagram presented. In total, 118 participants were screened for inclusion, with 65 total participants enrolled and randomized in the parent trial. No EEG data were collected for nine participants, and one participant was withdrawn from the study prior to EEG data collection. EEG and delirium data were thus available for a subset of 55 participants (27 in the placebo group and 28 in the caffeine group).

FIGURE 3

(A) Autocorrelation function presented for non-delirious participants (n = 42) subtracted from delirious participants (n = 11) across study epochs. EEG data were obtained in the PACU after initial admission and nursing evaluation. During PACU evaluation, alpha (8 – 12) and high alpha (10 – 14 Hz) autocorrelation function were significantly reduced in delirious participants, and theta (4 – 8 Hz) autocorrelation function was significantly increased with delirium. (B) Autocorrelation function presented for caffeine (n = 28) minus dextrose 5% in water placebo (n = 27) participants. Participants receiving caffeine demonstrated significantly higher alpha (8 – 12 Hz) autocorrelation function during PACU recovery. Color bars represent median autocorrelation function differences between groups. Pre-oxy, pre-oxygenation; LOC, loss of consciousness; Maint, maintenance anesthesia phase; PACU, postanesthesia care unit; PLE, phase lag entropy.

FIGURE 4

Topography of phase lag entropy (2 – 20 Hz) presented. (A) At preoperative baseline, participants demonstrate relatively high frontal and central phase lag entropy compared to posterior regions. This anterior-posterior asymmetry, with relatively high phase lag entropy in anterior regions, appears more preserved in non-delirious participants compared to those with delirium during PACU recovery. (B) Relatively low phase lag entropy is present throughout anterior, central, and posterior regions in participants randomized to placebo, whereas relatively high anterior phase lag entropy is present with caffeine during PACU recovery.

FIGURE 5

(A) Phase lag entropy presented for non-delirious participants (n = 42) subtracted from delirious (n = 11) participants bandwidths and study epochs. High alpha (10 – 14 Hz) phase lag entropy appears increased during delirium in the PACU based on spectral comparisons (red shading). Conversely, delta (0.5 – 4 Hz) phase lag entropy appears lower with PACU delirium (blue shading). Color bars represent median phase lag entropy differences between groups. (B) Alpha phase lag entropy appears increased in anterior regions during delirium. Color bars represent phase lag entropy for the left and middle panels and median difference between groups in the right panel. (C) Although high alpha phase lag entropy was higher in patients with delirium, this increase did not reach statistical significance [5.38 (5.20 – 5.60) vs. 5.34 (5.25 – 5.39); p = 0.258]. (D) Delta phase lag entropy appeared globally reduced during delirium. Color bars represent median phase lag entropy for the left and middle panels and median difference between groups in the right panel. (E) Global delta phase lag entropy was significantly reduced from non-delirious to delirious participants (5.49 ± 0.200 vs. 5.32 ± 0.243, respectively; p = 0.023). Pre-oxy, pre-oxygenation; LOC, loss of consciousness; Maint, maintenance anesthesia phase; PACU, postanesthesia care unit; PLE, phase lag entropy.

FIGURE 6

(A) Phase lag entropy presented for participants randomized to placebo (n = 27) subtracted from those randomized to caffeine (n = 28) across bandwidths and epochs. Alpha (8 – 12 Hz) appears reduced with caffeine during postanesthesia care unit recovery (blue shading). Color bars represent median phase lag entropy differences between groups. (B) The anterior-posterior alpha phase lag entropy gradient appears diminished in the placebo group compared to caffeine during PACU recovery, and alpha phase lag entropy appears particularly reduced posteriorly in the caffeine group. Color bars represent median phase lag entropy for the left and middle panels and median difference between groups in the right panel. (C) Overall, alpha phase lag entropy is relatively lower in the caffeine group [5.18 (5.09 – 5.23)] compared to placebo [5.37 (5.25 – 5.42); p = 0.005] during PACU recovery. Pre-oxy, pre-oxygenation; LOC, loss of consciousness; Maint, maintenance anesthesia phase; PACU, postanesthesia care unit; PLE, phase lag entropy.