Introduction
Delirium is a serious neurocognitive disorder that is seen in 20%–40% of older adults undergoing surgery.1–3 It is characterised as a fluctuating disturbance in attention and awareness over a short period (hours–days) accompanied by a disturbance in cognition.1–3 Delirium is associated with many adverse outcomes in older adults, including a ninefold increased risk of incident dementia,4 41% increased likelihood of long-term cognitive impairment5 and a three-fold increased risk of mortality at 1 year.6 Delirium is a considerable financial burden worldwide, costing the Australian healthcare system AUD$8.8 billion in the 2016/2017 financial year7, and between US$143 and US$152 billion in the USA as reported in 2011.8
Delirium subtypes include hypoactive, hyperactive and mixed .9 Hypoactive delirium is characterised by decreased activity and amount or speed of speech, along with reduced awareness, while hyperactive delirium presents with increased activity, agitation and hallucinations.10 11 Displaying features of both hypoactive and hyperactive delirium characterises mixed delirium.10 Hypoactive delirium, as compared with other motor subtypes, has generally been associated with increased mortality and worse long-term cognition.12 Along with different prognoses, each subtype demands different hospital care.13
Delirium is conceptualised as a disorder of brain disintegration,14–16 and delirium vulnerability (ie, high risk of incident delirium) is thought to be driven by reduced baseline functional connectivity.17 There has been a recent call for subtype features to be assessed, with implications for better understanding the underlying neurobiology of delirium.1
Electroencephalography (EEG) is a portable and non-invasive functional neuroimaging technique, which can be used during rest and cognitive tasks. It measures summed postsynaptic excitatory and inhibitory potentials from the scalp with excellent temporal resolution.18 19 Spectral analysis is a standard measure reflecting the amount of periodic activity (sometimes termed oscillatory) in predefined frequency bands, for example, delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz) and gamma (30–100 Hz).14 Event-related potentials (ERPs) can be extracted from EEG data and reflect deflections in voltage time-locked to events.20 ERPs provide dynamic information about sensory and cognitive processes and can index activity before, during and after the onset of a stimulus on a millisecond-by-millisecond basis.21 EEG has successfully provided neural markers of numerous clinical disorders, including schizophrenia,22–24 coma,25 26 psychosis27 28 and depression,29 30 and is a promising approach to capture delirium vulnerability.31
Our recent systematic review summarised EEG associations with delirium relative to time, that is, before (vulnerability for delirium), during, and after delirium.31 These time-points are relevant as the EEG can be affected by surgical and situational factors. For example, EEG recorded during surgery is known to be affected by anaesthesia and other events including hypothermia.32 In our review, EEG at the time of a delirium episode was consistently associated with slowing, predominantly characterised by higher delta and theta power, along with lower alpha power.31 Only two studies measured EEG before both the precipitant and the manifestation of delirium, a time-point unaffected by surgical factors, with neither reporting significant differences in relative delta power and EEG hemispheric symmetry between those who did and did not go on to develop delirium.31
A recent prospective study collected preoperative and postoperative EEG, along with preoperative MRI33 . They reported that those who went on to develop delirium had higher preoperative alpha power, increased alpha-band functional connectivity and increased radial diffusivity.33 Increased functional connectivity was interpreted as a compensatory mechanism for maintaining cognitive function in the presence of underlying structural degeneration, which was then overwhelmed by mechanisms of delirium.33 The EEG recording consisted of 15 min of eyes-closed resting state, and the possibility of periods of sleep cannot be excluded.34 We consider it essential to assess both eyes-open and eyes-closed states, given the arousal systems are key in delirium neurobiology, and to control for baseline cognitive function, given cognitive impairment is a known delirium risk factor.35–39
EEG delirium markers are the result of underlying neurobiological processes. Multiple neurotransmitter systems (and their interactions) are implicated in the development of delirium, including acetylcholine, gamma-aminobutyric acid (GABA), norepinephrine, serotonin and dopamine.15 40 41 The role of acetylcholine is heavily involved in two key features of delirium: attention42 and arousal.40 Acetylcholine abnormalities can disrupt sensory input, giving rise to delirium symptoms, including inattention, disorganised thinking and perceptual disturbances.40 Increases in dopamine may lead to hyperactive symptoms, including hallucinations, agitation and irritability, due to the inhibition of the ability for catechol-O-methyl transferase to break down dopamine in the prefrontal cortex.43 44 GABAergic medications, including benzodiazepines, are a precipitant of delirium.17 Neurotransmitter levels correlate with EEG indices, for example, early ERP components appear to be modulated by cholinergic medication, and low levels of cholinergic acetyltransferase have been associated with increased delta power.45 46 GABAergic, glutamatergic and cholinergic neurotransmission are important for predictive attentional processes, such as those indexed by the mismatch negativity (MMN) ERP component during the auditory oddball paradigm.42 47
Delirium is a whole-brain disorder, representing an extensive failure of normal brain function. This failure is undoubtedly the result of widespread network disintegration with disturbances within and between arousal systems and cognitive networks.37 48 EEG spectral profiles and patterns of ERP componentry between delirium subtypes have not been investigated. Increases in neurotransmitters such as norepinephrine have been thought to contribute to symptoms characteristic of hyperactive and mixed delirium such as hypervigilance, while changes in GABA and serotonin may potentially be predominantly involved in hypoactive delirium.41 44 In other disorders where hypervigilance is a defining characteristic (eg, post-traumatic stress disorder), EEG changes have been observed, including increased MMN amplitude.49 In contrast, states of hypo-arousal characterising disorders such as attention deficit hyperactivity disorder, are associated with EEG slowing and attenuated ERP components.50–52 It is not yet clear whether EEG spectral profiles and patterns of ERP components differ between delirium subtypes. Still, given the interactions between neurotransmission, EEG changes and behavioural symptoms, we expect that delirium subtypes (hyperactive, hypoactive and mixed) and no delirium will relate differently to EEG power and ERP indices.
The DIVULGE study aims to characterise the neural mechanisms underlying vulnerability to delirium and its subtypes using EEG and ERPs. We will employ a prospective observational design, measuring EEG in older adults prior to elective cardiac surgery that may precipitate delirium. We will use state-of-the-art EEG data processing and visualisation methods to assess differences in EEG power along with ERP amplitudes and latencies between those who do and do not go on to develop delirium. We will also determine the effects of delirium subtype, severity and duration. We will extract EEG data in the form of periodic and aperiodic power spectra from resting data (eyes open and eyes closed) and ERP component amplitude and latencies from an auditory oddball paradigm.53 54 It is hypothesised, based on previous literature,31 33 55–57 that those who go on to develop delirium will display increased EEG slowing and attenuated ERP amplitudes as compared with those who do not. How this pattern varies as a function of subtype, severity and duration is an exploratory aim.
Characterising neural mechanisms of delirium vulnerability will lead to significant theoretical advances in the field of delirium neurophysiology. Furthermore, findings could feed into a delirium risk tool using EEG to identify individuals at high risk prior to surgery, a time during which preventative efforts can be employed.58 59 Such a tool could differentiate between risk of different subtypes, which have different care pathways and prognoses.12 13 Prevention of delirium is more effective than treatment once the delirium has occurred,59 and a recent meta-analysis reported that non-pharmacological multicomponent interventions reduced the incidence of delirium (risk ratio: 0.53; 95% CI 0.41 to 0.69).60
