Auditory attention measured by EEG in neurological populations: systematic review of literature and meta-analysis

Sensorimotor synchronization strategies have been frequently used for gait rehabilitation in different neurological populations. Despite these positive effects on gait, attentional processes required to dynamically attend to the auditory stimuli needs elaboration. Here, we investigate auditory attention in neurological populations compared to healthy controls quantified by EEG recordings. Literature was systematically searched in databases PubMed and Web of Science. Inclusion criteria were investigation of auditory attention quantified by EEG recordings in neurological populations in cross-sectional studies. In total, 35 studies were included, including participants with Parkinson’s disease (PD), stroke, Traumatic Brain Injury (TBI), Multiple Sclerosis (MS), Amyotrophic Lateral Sclerosis (ALS). A meta-analysis was performed on P3 amplitude and latency separately to look at the differences between neurological populations and healthy controls in terms of P3 amplitude and latency. Overall, neurological populations showed impairments in auditory processing in terms of magnitude and delay compared to healthy controls. Consideration of individual auditory processes and thereafter selecting and/or designing the auditory structure during sensorimotor synchronization paradigms in neurological physical rehabilitation is recommended.


Quality assessment
The quality assessment of the included articles was based on the STROBE checklist 20 .

Data extraction
The following data were extracted from the selected articles: participant population (healthy or neurological disease), descriptive characteristics of the participants (age, disease information), neuropsychological information about the participants (neuropsychological test results), descriptive characteristics of the EEG paradigm used (frequency of stimuli, inter-stimulus-interval, decibels (dB) of stimuli, task instructions, stimulus length and probability of the deviant sound), electrophysiological measures (P3 ERP results (amplitude and latency) measured at Pz location).

Data analysis
A meta-analysis comparing healthy controls and neurological populations was performed on P3 amplitude and latency data using Review Manager version 5.4.1 for a meta-analysis using random effects and 95% confidence intervals (CIs).Subgroup analyses were performed stratifying the data into neurological populations.Studies were included in the meta-analysis when P3 amplitude and latency values were provided.

Terminology
Specialized terminology used throughout the manuscript can be found in Appendix 1.

Quality assessment
The Supplementary Table 1 shows the results of the STROBE checklist for all included studies.Overall, the quality of the studies was acceptable.The articles had a clear explanation of their scientific background and provided clear explanations of the aims, hypothesis, and experimental design of their study.
All studies but three, compared the patient group with healthy controls based on neurophysiological measures (ERPs) and neuropsychological measures (cognitive outcome measure).When a healthy control group was included, they were age-matched to the patient group.

Neuropsychological test results
A variety of neuropsychological tests were used across studies.An overview of all these tests can be found in supplementary Table 2.Not all studies compared neuropsychological test results statistically between healthy controls and the patient groups.When a comparison was made, significant results were found for persons with PD on cognitive screening 23,26,31,33 , verbal fluency 21,25,30,33 , visuospatial skills 21 , visual memory 23 , recognition abilities 23 , intelligence screening 25 , working memory 33 and sustained attention 30 , indicating better scores for healthy controls.For persons with stroke, significant impairments were found in cognitive screening 36,37 , verbal fluency 38 and working memory 38 .For persons with ALS, a lower score compared to healthy controls was found for cognitive screening [52][53][54] , verbal fluency 53,54 , intelligence screening 52 , visual attention 53 and working memory 54 .For TBI, lower scores were found for intelligence screening 43,44 and working memory 44 .Last, for MS, only significant results are found for visual memory 48 .
Overall, the results indicate that cognitive screening, verbal fluency and working memory are the cognitive functions that were most impaired within the neurological populations included in this review.www.nature.com/scientificreports/

Experimental paradigm
All experiments applied an auditory oddball paradigm.The mean length of auditory stimuli was 135 ms ranging from 40 ms to 500 ms.The frequency of the deviant sound also varied between studies ranging from 500 Hz to 2000 Hz (Mean: 1604.57Hz) with 2000 Hz as the frequency used in 41% of the studies.The difference between the deviant and frequent sounds frequency ranged from 500Hz up to 1750 Hz with an inter-stimulus-interval of an average of 1461.48 ms.However, we should note that inter-stimulus-interval was not always reported in all studies.Of all studies included, 21 studies (61.76%) instructed participants to mentally count the number of deviant sounds and report them after each trial.While 14 studies (41.18%) instructed participants to press a button when a deviant sound was presented.The mean (decibels) dB used in all studies was 71.19 dB with SD = 9.84.The probability of the deviant sounds ranged from 10 to 30% with 20% as most used in the included studies.For most studies, participants were instructed to sit silently on a chair and to keep head movements as minimal as possible to control for muscle artifacts.An overview of all descriptive information regarding the paradigm can be found in Table 2.

Neurophysiological results quantified by the EEG recordings
Below, we describe P3 amplitude and latency differences between healthy controls and neurological populations presented as a meta-analysis.Forest plots for random-effects meta-analysis stratified by neurological population comparing amplitude and latency outcomes between neurological populations and healthy controls and for all studies combined are presented in Fig. 2. Noteworthy, the meta-analysis of the P3 amplitude contained only one study for MS 50 and SCI 55 , two studies for stroke 35,38 and PD 25,26 and six studies for TBI 40,42,43,[45][46][47] .The meta-analysis of P3 latency included one study for stroke 55 and SCI 35 , two for ALS 53,54 , three for MS 48,50,51 , five for PD 23,25,26,29,30 and six for TBI 40,42,43,[45][46][47] .
Additionally, we compared P3 amplitude and latency of each neurological population separately with healthy controls.The results showed that: • Lower P3 amplitudes were found for the following neurological populations compared to healthy con- trols: persons with stroke (p = 0.007) (mean difference - www.nature.com/scientificreports/A visual illustration of the mean amplitude and latency collapsed across the different populations can be found in Fig. 3, and a complete overview of P3 amplitude and latency values and the analysis time-windows can be found in Table 3.

Discussion
The aim of this systematic review was to investigate auditory attention differences between neurological populations and healthy controls.Consistent with literature, the studies included in this review applied the auditoryoddball paradigm for these investigations, as the P3 ERP component is frequently used to investigate attentional resources 19 .
Our results show an overall longer P3 latencies and lower amplitudes for neurological populations compared to healthy controls.When comparing each neurological population, we saw that this overall effect in terms of amplitude was present for persons with stroke, TBI, MS and SCI, indicating lower amplitudes for these neurological populations compared to healthy controls.However, this was not the case for PD and ALS.In terms of latency, the overall effect was seen for stroke, MS, PD and ALS indicating longer latencies for the latter neurological populations compared to healthy controls.However, this effect was not seen for TBI and SCI.
The amplitude of the P3 is proportional to the level of attentional resources activated in the processing a stimulus 17 , and in our study, this is specific to the auditory stimulus.The P3 has been reported to be decreased in the presence of attentional deficits 56,57 .The P3 latency reflects the time needed for stimulus evaluation 58 .When latencies are longer, more time is needed to evaluate and process the stimulus 59 .Noteworthy, some factors could influence P3 amplitude and latency such as stimulus significance 60 , global target probability [61][62][63][64][65] , inter-stimulusinterval (ISI) 66,67 , the time-window used 68 and task-instruction 69 .These are some important aspects to consider when looking at the existing literature.The studies included in this review used a time-window ranging from 200 to 700 ms, with the range usually set between 250 and 600ms 70 .
Our results show inconsistencies in terms of P3 amplitude and latency across neurological populations, mainly in the pathologies of ALS and PD.These results can be explained by either, the limited number of studies that could be included in the meta-analysis, or due to the underlying pathophysiology of the diseases.Below, the latter is elaborated for the different neurological populations.
PD is characterized by lesions within the basal ganglia caused by degeneration of dopaminergic neurons 71 .Within this population, studies have shown that the basal ganglia show preferential activation by perception of rhythms with a steady beat without deviations 72 .In terms of P3 amplitude, a systematic review by De Groote and colleagues (2020) 73 has shown that auditory perception deficits seen in PD attribute to the impaired central auditory processing; however, sample size and the similarity between deviant sounds and frequent sounds could largely affect results.Additionally, studies show that persons with PD show impaired timing of isochronous intervals 74 causing the perception of oddballs or changes in rhythms to be impaired.
ALS is an idiopathic progressive neurodegenerative disorder that affects nerve cells in the brain and spinal cord 75 .It primarily targets the motor neurons, which are responsible for controlling voluntary muscle movements 75 .However, no clear studies could be found on the processing of deviances in rhythmic sequences for persons with ALS.This could be explained by the pathophysiology of the disorder as it largely affects motor neurons responsible for muscle control and movement, rather than sensory processing areas of the brain which could explain the lack of differences between persons with ALS and healthy controls.
In persons with TBI, perception of deviances in rhythmic sequences can be impaired as a result of the alteration in brain function due to the trauma caused by an external force 76 .Greater impairment in rhythmic perception is seen for patients with right hemisphere damage compared to the left hemisphere 76 .However, lesion location can highly impact possible processing difficulties of sounds.
In persons with stroke, studies have shown impaired rhythmic perception 77 .This is not always the case and is influenced on the location of the stroke-related lesions.More impairments with rhythmic perception difficulties are reported when damage is found in the basal ganglia and supplementary-motor-area 77 .Evidence suggests a relation between the stroke lesion and acquired amusia, indicating that the ability to perceive rhythms can be impaired within this population 78 .
In persons with MS, an overall consensus could be seen in terms of lower P3 amplitudes and longer P3 latencies compared to healthy controls.Impaired information processing capacities within this population due to impaired connectivity between critical brain regions caused by demyelination is often reported.Studies have shown that up to 50% of persons with MS experience difficulties with information processing 79 .However, www.nature.com/scientificreports/evidence shows the capability of persons with MS to synchronize their steps to beats in music and metronomes at different tempi 4 .The results of this review provide insights that auditory processing is present but impaired in the neurological populations compared to healthy controls: both in terms of magnitude (amplitude) and delay (latency).These insights should be considered when composing the auditory stimuli in strategies using sensorimotor synchronization.For example, considerations of the tempi are required: too fast or too slow tempi would hamper the auditory processing in the presence of the impairment.Another aspect when considering these impairments is the application of adaptive rhythmic systems.Studies have shown that an alignment strategy that continuously adapted the music to the participants' walking pattern showed the best results in terms of synchronization 80 , and these effects have been shown to be favorable in persons with PD as well 81 .Thus, the delay in attentional processes of individual participants need to be considered when developing such alignment strategies.Building on the theme of adaptation, we hereby address the recent development of methodologies designed to capture the dynamic nature of attending 7,82 .In particular, measuring variations in the frequency of oscillatory brain components attuned to the rhythmic stimulus has the potential for future fundamental research on the clinical populations investigated in the present work 83,84 .Among these developments, we point at event-related frequency adjustments (ERFAs) as a viable alternative to traditional ERPs paradigms, to investigate how different pathologies selectively impair oscillatory dynamics underlying auditory attention and sensorimotor synchronization (for details on the experimental paradigm, see 84 ).The impact of designing the stimuli to fit the individual attentional capacities can be seen in anticipating the provision of precision medicine with heightened benefits in terms of longer training durations, or training at higher intensities.

Limitations
The amount of studies included in this systematic review both reporting on amplitude and latency measures is rather limited, and thus the meta-analysis included a limited number of studies.Within the included studies, no differences were made between P3a and P3b components, making the interpretation of novelty and habituation  www.nature.com/scientificreports/difficult.Additionally, further sensitivity analysis on the effect of task instruction (i.e.mental counting or buttonpressing), on P3 amplitude and latency could not be performed as well.However, studies have shown that motor responses can occlude P3 differences resulting in smaller P3 amplitudes and shorter P3 latencies 69 .Building on the concept of embodied cognition, defined as the body's interactions with the environment that contribute to cognition 85 , where a motor action-here a button-press-can offload cognitive processing and thus facilitate it.On the other hand, mentally counting the deviant sounds adds a layer of attention and working memory to the task, which might make the task more cognitively difficult compared to a button press, possibly resulting in longer processing times 86 .Additionally, studies did not all report on the cognitive or motor characteristics (or impairments) of the included participants, and thus, these factors could not be assessed within our investigations.Last, the studies included in this review focus on the processing of auditory deviations in rhythmic sequences to better understand how possible processing delays can impact auditory stimulation in rehabilitation settings.However, one could consider that higher order auditory processing is not accounted for (e.g., dichotic listening tasks), where a person is asked to selectively shadow or repeat information presented in one ear while ignoring information presented in the other ear to understand right or left ear advantage 87,88 .To move forward in understanding higher order auditory processing differences between neurological populations and healthy controls, a thorough review of this literature is needed.Further, the current review does not consider the robustness of auditory object formation needed to correctly attend and differentiate between target and non-target auditory stimuli 89 .This could have important implications as the evolution of a sound can impact auditory processing and lead to differences in P3 latency and amplitude, rather than being the result of a neurological condition.

Conclusion
Overall, neurological populations showed impairments in auditory processing in terms of magnitude (P3 amplitude) and delay (P3 latency) during auditory oddball paradigms compared to healthy controls.Discrepancies in the direction of change of P3 amplitude and latency was found only in persons with PD and ALS for amplitude and in PD and TBI for latency when compared across the neurological pathologies.Consideration of individual auditory processes and thereafter selecting and/or designing the auditory structure during sensorimotor synchronization paradigms in neurological physical rehabilitation is recommended.

Figure 1 .
Figure 1.Flowchart over the search strategy and article selection process (according to the PRISMA guidelines).

Figure 2 .
Figure 2. (A) Forest plot of random effects for amplitude outcomes.(B) Forest plot of random effects for latency outcomes.

Figure 3 .
Figure 3. (A) Mean amplitude and latency collapsed across studies reporting on these measures, divided by health controls and neurological populations.*All references are indicated at each datapoint using the reference number listen in the reference list.HC = healthy controls.(B) Mean amplitude and latency collapsed across studies reporting on these measures, divided by health controls and different neurological populations.*HC healthy controls.

Table 1 .
Descriptive information of the studies.HC healthy control.

Table 2 .
Descriptive information of auditory oddball paradigms applied in the included studies.

Table 3 .
Preprocessing of ERPs and results.HC healthy control.