The posterior dominant rhythm: an electroencephalographic biomarker for cognitive recovery after general anaesthesia

Abstract

Background

The posterior dominant rhythm (PDR) was the first oscillatory pattern noted in the EEG. Evoked by wakeful eyelid closure, these oscillations dissipate over seconds during loss of arousal. The peak frequency of the PDR maintains stability over years, suggesting utility as a state biomarker in the surveillance of acute cognitive impairments. This EEG signature has not been systematically investigated for tracking cognitive dysfunction after anaesthetic-induced loss of consciousness.

Methods

This substudy of Reconstructing Consciousness and Cognition (NCT01911195) investigated the PDR and cognitive function in 60 adult volunteers randomised to either 3 h of isoflurane general anaesthesia or resting wakefulness. Serial measurements of EEG power and cognitive task performance were assessed relative to pre-intervention baseline. Mixed-effects models allowed quantification of PDR and neurocognitive trajectories after return of responsiveness (ROR).

Results

Individuals in the control group showed stability in the PDR peak frequency over several hours (median difference/inter-quartile range [IQR] of 0.02/0.20 Hz, P=0.39). After isoflurane general anaesthesia, the PDR peak frequency was initially reduced at ROR (median difference/IQR of 0.88/0.65 Hz, P<0.001). PDR peak frequency recovered at a rate of 0.20 Hz h⁻¹. After ROR, the PDR peak frequency correlated with reaction time and accuracy on multiple cognitive tasks (P<0.001).

Conclusion

The temporal trajectory of the PDR peak frequency could be a useful perioperative marker for tracking cognitive dysfunction on the order of hours after surgery, particularly for cognitive domains of working memory, visuomotor speed, and executive function.

Clinical trial registration

NCT01911195.

Editor's key points.

• •The posterior dominant rhythm (PDR) is an oscillatory pattern in the EEG evoked by wakeful eyelid closure that dissipates over seconds during loss of arousal.
• •The peak frequency of the PDR is stable over years, and might be a biomarker in defining acute cognitive impairment.
• •In a volunteer study of 60 healthy adults, after isoflurane general anaesthesia the PDR peak frequency was initially reduced at return of responsiveness and slowly recovered over a period of hours, which correlated with cognitive recovery.
• •This EEG biomarker offers a potential marker for prediction of acute cognitive impairment.

The posterior dominant rhythm (PDR), commonly known as the posterior alpha oscillation, has unclear perioperative utility.¹ These EEG oscillations emerge after eyelid closure during relaxed wakefulness² and reflect thalamocortical connectivity,³ with maximal amplitude in occipital channels.⁴ Over the course of development, PDR peak frequency, or individual alpha frequency,⁵ arises around 3 months of age in the 3–5 Hz range,⁶ then gradually increases until plateauing around age 16 yr, where the peak frequency lies within the 8–13 Hz alpha band⁷ during adulthood. Although it fluctuates within a 1 Hz range over years,⁸ it also slowly decrements with ageing.⁹

Relative changes in the alpha band EEG activity are associated with acute changes in arousal and cognitive function. A decline in the PDR peak frequency occurs during descent into sleep.¹⁰ The PDR peak frequency has been linked to performance on tasks requiring somatosensory attention¹¹ and visuospatial attention.^12,13 Lower PDR peak frequencies emerge during reversible acute illnesses such as hepatic encephalopathy¹⁴ or sepsis-associated encephalopathy.¹⁵ Augmentation of PDR peak frequencies can be induced by nicotine¹⁶ and caffeine.¹⁷ GABAergic anaesthetics induce loss of power in the PDR and anteriorisation of alpha activity during general anaesthesia,^18,19 with re-emergence of occipital alpha power within minutes after return of responsiveness (ROR).²⁰ However, changes in the PDR peak frequency in relationship to cognitive function have not been systematically characterised in the hours after general anaesthesia.

We evaluated the evolution of the PDR peak frequency over several hours during recovery from 3 h of isoflurane general anaesthesia. We hypothesised that the PDR peak frequency initially declines and gradually recovers during the hours after ROR. In addition, we hypothesised that deviations from the baseline PDR peak frequency correlate with impairments in cognitive task performance. Our rationale is that this neurophysiologic marker of thalamocortical integrity may be useful for longitudinal monitoring of cognitive dysfunction during acute and critical illness.

Methods

The previously described^21,22 experimental methods for participant recruitment and data acquisition for the Reconstructing Consciousness and Cognition study (ClinicalTrials.gov NCT01911195) are summarised here.

Participants

This study was approved by the institutional review boards at University of Michigan, University of Pennsylvania Perelman School of Medicine, and Washington University School of Medicine in St. Louis. Twenty healthy adults were recruited from each site, and written informed consent was obtained. Each site randomised 10 participants to each study arm, with a total of 30 participants in the Control arm and 30 participants in the Isoflurane arm.

The following inclusion criteria were applied: 20–40 yr old, ASA physical status 1 or 2 (healthy), BMI <30 kg m⁻², and an easily visualised uvula. The following exclusion criteria were used: physical signs suggestive of a difficult airway, such as a mouth opening <3 cm, short distance between the chin and neck, poor mandibular subluxation, or a thick neck; history of oesophageal reflux, family history of problems with anaesthesia (including, but not limited to, malignant hyperthermia), or history of postoperative nausea/vomiting or motion sickness; reactive airway disease; hypertension or current medication for blood pressure control; cardiovascular disease or arrhythmias; neuropsychiatric disorders or sleep disorders; history or current use of psychotropic medications; history of positive urine toxicology screen; current tobacco or alcohol use; history of obstructive sleep apnoea; pregnancy; known allergy to eggs, egg products or soy; and/or dentures, bridges, or crowns.

Study protocol and data acquisition

Participants were randomised to a 3-h period of either resting wakefulness (Control arm) or general anaesthesia (Isoflurane arm) compatible with surgical procedures. Those in the Control arm remained awake for 3 h while watching television or reading a book (Fig. 1a). In contrast, general anaesthesia for those in the Isoflurane arm was induced by a 15 min propofol infusion followed by 3 h of isoflurane inhalation (at 1.3 times minimum alveolar concentration [MAC]; Fig. 1b). After isoflurane was discontinued, an audio recording was started to assess ROR to verbal command objectively. At 30-s intervals, the recording instructed the participant to squeeze either the left or right hand twice. ROR was defined as the moment the participant was able to comply with verbal commands by squeezing a bulb in the instructed hand twice, which marked the beginning (T0) of cognitive performance testing and eyes closed EEG recording.

Fig 1.

Overview of study procedures. Eyes-closed EEG and cognitive task performance were assessed at baseline before intervention (Pre-). (a) Participants in the Control arm underwent a 3-h period of resting wakefulness while watching videos on a television. After this period, participants underwent serial cognitive testing and 5-min epochs of eyes-closed EEG recordings at 30-min intervals up to 180 min (T0 to T180). (b) Participants in the Isoflurane arm underwent a general anaesthetic protocol with an initial 15-min propofol infusion for induction followed by a 3-h isoflurane inhalation. After ROR (return of responsiveness to verbal command), participants underwent serial cognitive testing and 5-min epochs of eyes-closed EEG recordings at 30-min intervals up to 180 min (T0 to T180). PDR, posterior dominant rhythm.

Neurocognitive testing and 5 min of eyes-closed wakefulness EEG were acquired before the anaesthetic/control period and at 30 min intervals after ROR (T0, T30, T60, T90, T120, T150, and T180). As the actual times of testing diverged depending on individual clinical recovery (presence of nausea, vomiting, or lightheadedness), precise testing times for each task were subsequently used in analytical modelling. Six neurocognitive tests were administered with the Cognition test battery on a Dell Latitude E5430 Laptop with a 14-in liquid crystal display (Dell, Inc., Round Rock, TX, USA) to evaluate different cognitive domains such as working memory, abstract thinking, attention, and reaction time²³ including: the Psychomotor Vigilance Test (PVT), the Digit Symbol Substitution Task (DSST), the Motor Praxis task (MP), a Fractal-2-Back task (F2B), the Visual Object Learning Task (VOLT), and the Abstract Matching task (AM). The PVT assesses vigilance by measuring the time needed to press a space bar in response to recognising an incrementing timer on the screen. In addition to sensorimotor speed, performance on the DSST requires visual search, spatial memory, and paired associate learning in matching digits to symbols. The MP evaluates speed of sensorimotor performance in moving a cursor to the centre of a square with varied locations and sizes. The VOLT entails the memorisation of three-dimensional figures and tests their recall amongst decoy figures. The F2B evaluates working memory of fractal images, in which the task is to indicate when the current pattern matches that presented two previously. The AM assays executive function; classification of visual objects is tested through the identification of implicit abstract rules based on characteristics, such as filling or shape. The median values for accuracy and speed were computed for each testing session.

EEG acquisition, processing, and quantification of the posterior dominant rhythm

EEG was recorded using 32-, 64- or 128-electrode Geodesics Sensor Net (Electrical Geodesics, Inc., Eugene, OR, USA) and Net Amps 400 amplifiers. Five-minute epochs of wakeful eye closure were identified for analyses. EEG data underwent 1–50 Hz band-pass filtering with a first-order Butterworth filter and subsequent resampling to 250 Hz. Spectral analyses were performed using the Chronux toolbox,²⁴ with spectral estimates between 1 and 30 Hz based on 6 s non-overlapping time windows, a time–bandwidth product of 3, and 5 tapers.

Custom-written scripts were used to quantify PDR peak frequency by analysing the spectral peak within the 8–13 Hz alpha band in channel P4-O2. To address motion artifact, eye opening protocol violations, and poor detectability, we used the Better OSCillation (BOSC) approach²⁵ to assist detection of spectral peaks above the background spectral power. BOSC output was then used as the input to the custom-written peak finder algorithm to identify the frequency with peak power using the following steps: (i) outlier removal based on five standard deviation rules applied to the spectral content over 6-s windows; (ii) smoothing of the power spectral density using a moving average with a window length of 2; (iii) utilisation of a noise-tolerant derivative-based peak finder, with quadratic interpolation around each extrema to estimate accurate position of each peak; and (iv) calculation of median frequency across all 6-s windows. Participants were excluded if they had <1 min of artifact-free data, a lack of a detectable peak frequency in the alpha band, or a lack of baseline PDR to quantify changes in the PDR throughout the experiment. We assessed the difference in median peak frequency relative to pre-intervention baseline. As we were not analysing the variance and assumed stability in the PDR over the course the 5-min epochs of eye closure, we did not correct for the total amount of data included at each time point. Quantitative changes in the PDR frequency are more easily assayed from power spectra of occipital parietal EEG (P4-O2) signals compared with frontal central EEG (F4-Cz) (Supplementary Fig. S1).

Statistical analysis

Wilcoxon signed-rank tests were utilised to test for paired differences in medians. Linear mixed-effects models were used to assess the relationship between changes in PDR peak frequency and time after ROR as a function of the study arm. Marginal modelled responses and 95% confidence intervals (CIs) were calculated to compare the recovery between the two groups. Baseline PDR peak frequency was included as a fixed effect to assess a relationship between individual recovery and baseline PDR peak frequency. Repeated-measures analysis of variance (anova) was used to assess the stability of PDR peak frequency for study arms. Non-linear mixed-effects models based on a damped exponential were generated for each cognitive task to account for the recovery over time relative to pre-treatment baseline performance:

$$y_{\text{t}}-y_{\text{baseline}}=a+b{e}^{-ct}$$ (1)

where y_baseline is pre-treatment baseline performance, a + b is initial decrement, and c is the decay rate.

Median recovery curves were calculated from the modelled responses with 95% CI derived using bootstrapping (2000 iterations). Power analysis was carried out only for the primary study outcomes.²¹ There were no corrections of alpha for multiple null hypothesis statistical testing.

Results

Posterior dominant rhythm peak frequency is stable during wakefulness and undergoes a linear recovery after isoflurane anaesthesia

To serve as a reliable EEG marker for acute cognitive dysfunction, robust measures of the PDR with serial testing are required. We assessed the stability of the PDR peak frequency over a period of hours among individuals in the Control group. Representative data for an individual in the Control group showed stability in the peak frequency after a 3-h period of resting wakefulness (Fig. 2a and b). This demonstrates the reproducibility of the PDR peak frequency over ∼6 h of measurements (median difference/IQR of 0.02/0.20 Hz; Wilcoxon signed-rank test, P=0.39) (repeated-measures anova: F_6,154=0.17, P=0.98).

Fig 2.

Peak frequency of the posterior dominant rhythm (PDR) remains stable in non-anaesthetised controls and recovers after isoflurane general anaesthesia. (a) Representative Control participant's time–frequency power spectrograms of occipital (P4-O2) EEG demonstrate minimal change in peak frequency of the PDR over six 30-min intervals when compared with pre-control period baseline. This illustrates the reproducibility of an individual's dominant frequency in the PDR over approximately 3 h of measurements. (b) Power spectral density comparing periods of resting wakefulness (T0 to T150) and pre-intervention baseline (Pre-, black) showing that the peak frequency remains relatively consistent despite changes in power. (c) Compared with pre-intervention baseline (eyes-closed), spectrograms from a representative participant in the Isoflurane arm demonstrate a decrement in the peak frequency of the PDR during the first testing session after return of response (T0, red), and steadily increases in frequency and power throughout subsequent 30-min intervals. (d) Power spectral density comparing periods of resting wakefulness (T0 to T150) and baseline (Pre-, black) showing that the PDR peak frequency decreases initially before recovering towards baseline. (e) Box plot illustrating the deviation from pre-intervention baseline PDR peak frequency as a function of time relative to ROR (return of responsiveness) for individual participants in both the Control (blue) and Isoflurane (red) arms. Peak frequency of the PDR reverts towards baseline over the hours during recovery from isoflurane general anaesthesia. For quantitative linear mixed-effects modelling, actual times of measurements were ascertained. For panels (a)–(e), data are shown only up to T150 owing to the limited number of acquired EEG measures at T180. (f) To compare the differences between and within groups, bootstrapping was performed on modelled recovery curves (with 2000 bootstraps). The solid lines represent the bootstrapped median of each group and the shaded sections represent the 95% confidence intervals (CIs). The reduction in PDR peak frequency was significant between groups by 180 min after ROR (no overlap between two groups at t=180 min, with median −0.29 (95% CI, −0.37 to −0.16) for the Isoflurane arm and −0.02 (95% CI, −0.10 to 0.03) for the Control arm.

In contrast to non-anaesthetised awake control subjects, individuals in the Isoflurane group showed variability in the peak frequency and spectral bandwidth of occipitoparietal EEG power as a function of time after ROR (Fig. 2c and d). Using time-averaged data at each time point, the peak frequency increased monotonically over the span of hours (Fig. 2e), demonstrating that at an individual level, the PDR peak frequency varies over the hours after emergence from general anaesthesia (repeated-measures anova: F_6,147= 0.17, P=8.36×10⁻¹⁰).

To account for changes in the PDR peak frequency systematically as a function of study intervention, participant, and time, we utilised linear mixed-effects models to fit experimental measures. We maximised precision of our modelling by using the actual starting times of our EEG acquisition instead of designated time points (e.g. T0, T30, T60). Measurements acquired after ROR from general anaesthesia showed an initial decrement (median difference/IQR of 0.88/0.65 Hz; Wilcoxon signed-rank test, P=4.01×10⁻⁵). A linear return towards baseline was observed at both the individual and group levels (Fig. 2f; Red, Isoflurane arm: intercept at t=0, –0.90 Hz; 95% CI, –1.13 to –0.67; df=16; P=2.33×10⁻¹²; slope=0.20 Hz h⁻¹; 95% CI, 0.10 to 0.29; P=3.79×10⁻⁵). In contrast, the peak frequency did not vary over time for individuals in the Control arm, as evidenced from the marginal responses (Fig. 2g; Blue, Control arm: intercept at t=0: –0.05 Hz; 95% CI, –0.15 to 0.04; df=136; P=0.26; slope=0.01 Hz h⁻¹; 95% CI, –0.03 to 0.06; P=0.48). Overall, the PDR peak frequency recovers over a period of hours after general anaesthesia when compared with an individual's baseline measurements.

Recovery of cognitive task performance and posterior dominant rhythm peak frequency are correlated after general anaesthesia

We assessed the recovery of cognitive task performance after ROR from Isoflurane general anaesthesia arm compared with individuals in the Control arm (Fig. 3). We used the actual start time of the specific cognitive testing battery instead of the designated times after ROR and accounted for a damped exponential trajectory during recovery. We first compared this recovery of cognitive task performance with that of the PDR peak frequency for the two tasks that showed a large decrement in both accuracy and speed after isoflurane anaesthesia²² (Fig. 4). For the Isoflurane group, changes in sensorimotor task performance correlated with changes in the PDR peak frequency for both measures of the PVT and DSST (linear mixed-effects models, all P<0.001 and r>0.3). In contrast, the Control group had minimal changes in PDR peak frequency and sensorimotor task performance measures (repeated-measures anova: all P>0.05). For the remaining four tasks, changes in performance measures correlated with changes in PDR peak frequencies in varied extents (Fig. 5). Model comparison using Akaike information criterion (AIC) verified that decrements in task performance across all tasks correlated non-linearly with changes in PDR peak frequency for the Isoflurane cohort (Supplementary Material). Although change in task performance could be predicted based on time alone, the addition of change in PDR peak frequency (Supplementary Table S1) yielded better predictions for both speed (DSST, MP, VOLT, and AM) and accuracy (DSST and AM). Thus, the temporal trajectory of the PDR peak frequency may be a useful marker for tracking cognitive function after general anaesthesia, particularly for cognitive domains of working memory, visuomotor speed, and executive function.

Fig 3.

Cognitive task performance improves towards baseline over hours after isoflurane general anaesthesia. For quantitative modelling, cognitive task performance measures (circles) were fitted with realisations for either linear mixed-effects models (Control arm, blue) or non-linear mixed-effects models that incorporated a damped exponential function (Isoflurane arm, red). Performance measures were standardised as z-scores using the baseline (distribution of pre-intervention scores). All measures are normalised such that a higher score corresponds to a higher performance. (a) Change in speed measures during the Psychomotor Vigilance Task (PVT) vs time after return of responsiveness. (b) Change in accuracy measures for PVT vs time for modelled scores using mixed-effects models, similar to panel (a). (c, d) Speed and accuracy measures during the Digit Symbol Substitution Test (DSST) vs time for modelled scores using mixed-effects models, respectively.

Fig 4.

Changes in PVT and DSST performance vs changes in posterior dominant rhythm (PDR) peak frequency. (a) Correlation between PVT speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.35, F-test F=20.92, P=1×10⁻⁵). In all plots, markers represent the median of each group and the grey error bars represent the 25th and 75th percentiles. (b) Correlation between PVT accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.33, F-test F=17.36, P=5×10⁻⁵). (c) Correlation between DSST speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.37, F-test F=23.06, P=4×10⁻⁶). (d) Correlation between DSST accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.31, F-test F=15.85, P=1×10⁻⁴). DSST, Digit Symbol Substitution Test; PVT, Psychomotor Vigilance Task.

Fig 5.

Changes in MP, VOLT, F2B, and AM performance vs change in posterior dominant rhythm (PDR) peak frequency. (a) Correlation between MP speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.41, F-test F=30.35, P=2×10⁻⁷). (b) Correlation between MP accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.12, F-test F=2.06, P=0.15). (c) Correlation between VOLT speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.37, F-test F=22.45, P=5×10⁻⁶). (d) Correlation between VOLT accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.31, F-test F=15.75, P=1×10⁻⁴). (e) Correlation between F2B speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.33, F-test F=17.36, P=5×10⁻⁵). In all plots, markers represent the median of each group and the grey error bars represent the 25th and 75th percentiles. (f) Correlation between F2B accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.53, F-test F=56.12, P=6×10⁻¹²). (g) Correlation between AM speed and changes in PDR peak frequency for the Isoflurane arm (red, r=0.33, F-test F=18.31, P=3×10⁻⁵). (h) Correlation between AM accuracy and changes in PDR peak frequency for the Isoflurane arm (red, r=0.21, F-test F=6.77, P=0.01). AM, Abstract Matching; F2B, Fractal-2-Back task; MP, Motor Praxis; VOLT, Visual Object Learning Task.

Discussion

We compared a unique data set of EEG and task performance measures acquired from healthy volunteers exposed to surgical levels of isoflurane anaesthesia and awake controls. The PDR peak frequency did not change significantly over several hours of passive wakefulness. In contrast, an initial decline in the PDR peak frequency after 3 h of isoflurane anaesthesia was followed by a linear reversion towards baseline. The recovery of cognitive task performance and PDR peak frequency correlated in a non-linear manner. Changes in the PDR peak frequency were predictive for tasks requiring working memory, visuomotor speed, and executive function. Our findings complement prior investigations documenting a shift in alpha power from occipital to frontal regions during the descent into anaesthetised states. Overall, these data advance the PDR peak frequency as an EEG biomarker that could be tracked as a perioperative surrogate for serial assessments of cognitive performance.

Posterior dominant rhythm recovery over time and interventions

Our data on systematic fluctuations in the PDR peak frequency over the hours after isoflurane general anaesthesia complement the characterisation of this marker across different time scales. The peak frequency of the PDR has been viewed as a ‘trait marker’ that is stable over years,²⁶ and previously referenced as the ‘individual alpha frequency’,²⁷ with approximate heritability of 0.8.²⁸ Although longitudinal studies over decades have not been reported, cohort studies suggest that the PDR peak frequency is low during childhood development,⁶ reaches a plateau in young adulthood, and subsequently declines with age.⁹ Test–retest reliability suggests that these oscillations are sufficiently robust such that 20 s may suffice to assess the dominant EEG frequency in the context of noise.²⁹ Fluctuations in both the amplitude and dominant frequency are in clinical use as a ‘state marker’ to assess transitions between wakefulness and sleep (AASM).¹⁰ Furthermore, augmentation of the PDR peak frequency or power after treatment with psychostimulants^16,17 suggests that it may be a useful marker for states of both hypervigilance and impaired arousal. Changes in the PDR peak frequency appears to slow around 90 min after ROR, concordant with return of power in these oscillations.¹⁸ Thus, circuit mechanisms regulating both the strength and frequency of these oscillations may be linked. Overall, the temporal stability of the PDR across time scales spanning seconds to years confers usefulness as a biomarker in varied clinical and research settings.

Relationship of cognitive function to the posterior dominant rhythm peak frequency

The PDR peak frequency has been linked to somatosensory attention,¹¹ visuospatial attention,^12,13 and performance on semantic encoding and retrieval tasks.⁵ PDR peak frequency varies inversely with age-related decrements in the speed of information processing measured by reaction time in a recognition task.^30,31 PDR peak frequency is also postulated to reflect the integrity of neural networks during development.^32,33 However, the exact relationship of the PDR to cognitive processes remains unclear.

Since Berger³⁴ demonstrated a slow rhythm in senile patients, a reduction in the PDR peak frequency during wakefulness has shown promise as a marker of cognitive impairment incurred through depression or dementia.35, 36, 37 This premise is supported by previous research demonstrating alterations in PDR peak frequency associated with reversible cognitive impairment related to hepatic encephalopathy^14,38 or sepsis.^15,39 Overall, changes in the peak frequency of the PDR have been related to measures of cognition on both short- and long-time scales. Future longitudinal investigations are necessary to validate the potential utility of both PDR peak frequency and power for tracking postoperative cognitive dysfunction in the perioperative setting.

Clinical applications

EEG offers robust, inexpensive monitoring capable of improving future clinical care in the context of prediction and intervention for acute cognitive impairments. Ongoing investigations are assessing whether baseline measures of the PDR may be useful in assessing patient risk for postoperative delirium,⁴⁰ analogous to preoperative electrocardiograms that document baseline myocardial conduction, structure, and prior injury. Such measures would serve as objective baseline readings to facilitate detection of impending altered mental status in a continuous manner similar to cardiac telemetry. In the context of surgical procedures, differences from preoperative baselines may serve as a useful marker in predicting perioperative neurological complications that involve disorders of consciousness. Monitoring for acute changes in brain activity that may precede clinical decrements in cognitive function may provide early data to target interventions towards relieving systemic metabolic or inflammatory immune responses that are secondary to sepsis and organ failure. As a non-specific biomarker that is noninvasively and inexpensively acquired, this robust and well-studied signature has the potential for exploration over short time scales of acute illness.

Study limitations

A few limitations of our study motivate future investigation, as these secondary analyses were not prespecified before study initiation. Although study participants attempted to remain awake during eyes-closed periods, drowsiness may have contaminated our analysis epochs. We were also unable to control for whether suppression of the PDR may have occurred because of anxiety or mental effort. Furthermore, our algorithms to ‘scrub’ out these periods and quantify the peak frequency of alpha oscillations may have included periods of drowsiness. In addition, the remote possibility exists that residual propofol from induction of anaesthesia or use of the antiemetic ondansetron may have contaminated EEG measures or cognitive task performance metrics. Finally, the findings in this study may not generalise to patient populations with underlying cognitive impairments.

Conclusions

The posterior dominant rhythm peak frequency, stable on the order of hours, shows decrements after isoflurane general anaesthesia in healthy volunteers. Recovery towards baseline correlates in a varied manner with the return of working memory, visuomotor speed, and executive function. Perioperative utility for monitoring post-anaesthesia cognitive recovery and guiding mitigation of postoperative neurocognitive dysfunction deserves further investigation.

Authors' contributions

Study conceptualisation: BJP, ERH, SBM

Data acquisition: GAM, MBK, MSA, SBM, MB, BJP

Manuscript drafting: BJP, ERH, MK, AKL

Data analysis: MK, AKL, MB, BJP, ERH

Grant funding: GAM, MBK, MSA, BJP

Final review of manuscript: all authors

Handling editor: Hugh C Hemmings Jr

Declarations of interest

MSA is an editor of the British Journal of Anaesthesia. The other authors declare no conflicts of interest.

Funding

The James S. McDonnell Foundation (to GAM, MBK, MSA), the McDonnell Center for Systems Neuroscience at Washington University in St. Louis (BJP), and the US National Institutes of Health National Institute on Aging (NIA) (grant R01 AG057901 to BJP).

Appendix A. Supplementary data

The following is the Supplementary data to this article: