TMS-EEG responses across the lifespan: Measurement, methods for characterisation and identified responses

The combination of transcranial magnetic stimulation (TMS) and electroencephalography (EEG) allows probing of the neurophysiology of any neocortical brain area in vivo with millisecond accuracy. TMS-EEG is particularly unique compared with other available neurophysiological methods, as it can measure the state and dynamics of excitatory and inhibitory systems separately. Because of these capabilities, TMS-EEG responses are sensitive to the brain state, and the responses are influenced by brain maturation and ageing, making TMS-EEG a suitable method to study age-specific pathophysiology. In this review, we outline the TMS-EEG measurement procedure, the existing methods used for characterising TMS-EEG responses and the challenges associated with identifying the responses. We also summarise the findings thus far on how TMS-EEG responses change across the lifespan and the TMS-EEG features that separate typical and atypical brain maturation and ageing. Finally, we give an overview of the gaps in current knowledge to provide directions for future studies.


Introduction
Transcranial magnetic stimulation (TMS) is a non-invasive and painless brain stimulation method that can activate cortical neurons (Barker et al., 1985). By itself, the applicability of TMS to the study of the brain is limited. When combined with simultaneous electroencephalography (EEG), TMS evolves into a sophisticated method (TMS-EEG) capable of studying the neurophysiological characteristics and brain dynamics of any neocortical brain area or network (Ilmoniemi and Kičić, 2010). As such, TMS-EEG can have various applications, ranging from basic science to clinical research (Tremblay et al., 2019). TMS-EEG is also applicable in animal studies with the same setup as in human studies, making it an excellent translational study tool (Rotenberg et al., 2008).
The uniqueness of TMS-EEG is based on the capability of TMS to activate cortical excitatory (glutamatergic) and inhibitory (gammaaminobutyric acid-ergic, GABAergic) neurotransmitter systems differently and at different time scales (Ferreri et al., 2011;Premoli et al., 2014;Cash et al., 2017;Hui et al., 2020). These different neurotransmitter activations create separate components or peaks that construct TEP. The specific neurotransmitter system associated with a TEP peak can be evaluated with pharmacological studies, studies using paired-pulse techniques, and studies combining magnetic resonance spectroscopy (MRS) with TMS-EEG. Previous evidence suggests that the peaks within the first 30 ms reflect excitatory neurotransmission (Ferreri et al., 2011, Rogasch and Fitzgerald, 2013, but also contradictory evidence exists (Rogasch et al., 2020). Later peaks have been linked to the balance between glutamatergic excitatory and GABAergic inhibitory neurotransmission (Ferreri et al., 2011;Premoli et al., 2014;Du et al., 2018;Belardinelli et al., 2021). More specifically, the N45 is regulated by a balance of GABA A ergic inhibition and N-methyl--D-aspartate (NMDA) receptor-mediated glutamatergic excitation as dextromethorphan (an NMDA receptor antagonist), benzodiazepines and zolpidem (positive allosteric modulators of the GABA A receptor) have been demonstrated to increase its' amplitude (Premoli et al., 2014;Belardinelli et al., 2021). In contrast, the specific GABA B receptor agonist baclofen increases the N100 TEP amplitude (Premoli et al., 2014), and using MRS, N100 TEP has been linked to the excitation-inhibition balance mediated by local GABA and glutamate circuits, with higher glutamate/GABA ratio predicting larger N100 amplitude (Du et al., 2018). Moreover, perampanel [an α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor antagonist] reduces the amplitude of the P60 in the non-stimulated hemisphere, suggesting that the P60 TEP amplitude reflects glutamatergic (interhemispheric) signal propagation mediated by AMPA receptor activation (Belardinelli et al., 2021). However, also contradictory evidence exists (Ferreri et al., 2012). This is because GABA A -and cholinergic short-latency afferent inhibition (SAI, for more details, see 2.3 Different TMS-EEG pulse paradigms) inhibits motor-evoked potential (MEP) amplitudes and attenuates cortical P60 and N100 responses in the stimulated hemisphere (Ferreri et al., 2012). The neural origin of P180 is still mostly unknown, but antiepileptic drugs, such as lamotrigine, levetiracetam and carbamazepine, suppress the P180 amplitude (Premoli et al., 2017;Darmani et al., 2019). None of the studies has evaluated the mechanisms influencing N280 yet. Due to these connections between TEP peaks and neurotransmitters, changes in specific TEP peaks are thought to reflect changes in the corresponding neurotransmitter system. More studies investigating the neural mechanisms of TEPs are, however, still needed, as recent evidence suggests that especially the later peaks might originate or be partly modulated by TMS-induced artefacts , for more details, see 4.2 Physiological artefacts).
In addition to evoking a visible TEP in the time domain, TMS induces oscillations that can be quantified with EEG frequency domain analyses (Rosanova et al., 2009;Thut et al., 2012). These oscillations are commonly specific to the brain area. For example, TMS targeted to the occipital cortex evokes alpha oscillations (7.5-12.5 Hz), parietal cortex TMS evokes beta oscillations (12.5-30 Hz), and frontal cortex TMS evokes fast beta and gamma oscillations (21-50 Hz) (Rosanova et al., 2009). Hence, it is thought that instead of eliciting new neural responses, TMS mostly synchronises pre-existing and ongoing oscillations (Ferreri et al., 2014), but the possibility of TMS adding new energy to the system cannot be excluded (Pellicciari et al., 2017). When the induced oscillations are time and phase-locked to the TMS pulse, they create a TEP. TMS also induces non-phase-locked responses that average out in the TEP, but they can be seen with specific signal-processing methods (Pellicciari et al., 2017).
The TMS pulse sequences applied in TMS-EEG are safe if appropriate safety guidelines, which are similar for those for children (Rajapakse and Kirton, 2013;Zewdie et al., 2020) and adults (Rossi et al., 2021), are followed. TMS without EEG has been applied in as young as three-month-old infants (Kowalski et al., 2019) and up to 90-year-old geriatrics (Mosimann et al., 2004), although no specific safety guidelines for TMS applications in infants or geriatrics have been established. However, all available data suggest that harm to brain tissue from TMS applications is extremely unlikely. Some scalp discomfort may occur during the measurement, but this is transient and can be adjusted by changing the stimulation location.
Although TMS-EEG can be well applied across the lifespan, most studies to date have focused on adults, and studies in children and geriatrics are only just emerging; those in infants do not exist yet. In this review, we provide an overview of TMS-EEG across the lifespan, namely, how to measure TEPs, methods for characterising TEPs, challenges hindering TEP characterisation and how TEPs change with age and differentiate typical and atypical brain maturation and ageing. We also provide perspectives on future research on TMS-EEG across the lifespan.

Measurement of TMS-EEG responses
A basic TMS-EEG system consists of a TMS stimulator and coil, a TMS-compatible EEG amplifier and TMS-compatible EEG electrodes. Additional pieces of equipment that can be included are a neuronavigation system and an electromyography (EMG) device with EMG electrodes. The difference between conventional and TMS-compatible EEG systems is that the latter must have technology that avoids amplifier saturation, which results from the TMS pulse (Virtanen et al., 1999) and electrodes that are not susceptible to movement under the TMS coil or heating from the magnetic pulse (Varone et al., 2021). Some systems allow measurement of electro-oculography (eye movement and blinks), electrocardiography and facial muscle EMG, which might be useful Fig. 1. Typical TMS-evoked potential (P30-N45-P60-N100-P180-N280) in a healthy, young adult. The red line reflects data from the Cz electrode, whereas the black lines show data from all channels. A suprathreshold TMS pulse was targeted to the left primary motor cortex and given at a time of 0 ms. depending on the study design and data analysis protocol.
The strength of TMS pulses can be adjusted on a percentage scale between 0% and 100% of the maximum stimulator output (MSO). The MSO is stimulator dependent, so the values between two different manufacturers' devices may not be directly comparable given the different pulse parameters (Novey, 2019). Commonly, the pulse strength, i.e. the stimulation intensity, is normalised to the resting motor threshold (rMT) (Rossini et al., 1994;Ferreri and Rossini, 2013), which is unique in each individual and thought of as a general representative measure of how easily TMS can activate the brain. This relative stimulation intensity reflects individual cortical excitability levels Rossini et al., 2015). The rMT is estimated as the minimum stimulation intensity capable of inducing MEPs in a peripheral muscle with an amplitude of at least 50 µV in 50% of stimulations if an EMG device is available (Rossini et al., 2015). If not, the rMT can also be estimated using visible muscle movements as an acceptable MEP criterion, but this approach may overestimate the rMT (Westin et al., 2014). Although the rMT is sufficient to induce motor responses on the primary motor cortex (M1), it may not be enough on other brain areas, as the rMT is dependent on the coil-to-cortex distance (Herbsman et al., 2009). In TMS-EEG applications, the rMT should always be calculated with the EEG cap on, and the final stimulation intensity should be decided based on the target cortical area. However, EEG activity is induced with lower stimulation intensities than EMG activity is, and intensities as low as 40% of the rMT may already evoke TEPs (Komssi et al., 2007). Another commonly applied method to set the stimulation intensity is using the estimated electric field calculation at the target, which can be obtained with some neuronavigation systems. Neuronavigation is a system that combines real-time TMS coil location with a head model constructed from subject-specific magnetic resonance images (Ruohonen and Karhu, 2010). In addition to potential electric field calculation, neuronavigation allows tracking of the location of the TMS coil during measurement, targeting TMS pulses to some pre-set target, such as those derived from neuroimaging (Cash et al., 2020), and ensuring that the TMS coil is stable during TMS-EEG data acquisition. Small TMS coil location and orientation changes may affect the induced TEPs (de Goede et al., 2018). The coil can also be targeted with scalp-based methods, such as using standardised EEG electrode locations (Herwig et al., 2003).
A higher TMS pulse intensity, in relation to rMT, produces larger amplitude TEPs (Komssi et al., 2007;Saari et al., 2018). Higher intensity is also associated with greater auditory and somatosensory components, as the higher the intensity, the louder the TMS pulse click sound and the greater the cranial muscle activation near the stimulated location (Korhonen et al., 2011). These auditory and somatosensory components interfere with the neural signal of interest (see 4.2 Physiological artefacts). To reduce the TMS pulse-associated click sound effect, researchers may use white noise or similar noise during measurements (Nikouline et al., 1999). However, it may be challenging to attenuate the click sound entirely, as part of it travels along the skull via bone conduction (Nikouline et al., 1999). Attenuating somatosensory responses is also challenging because cranial muscles cover the scalp, and if the magnetic field travels through them, there will always be some muscle activation (Korhonen et al., 2011). Hence, the optimal approach to reduce these artefacts would be adding a sham condition (Gordon et al., 2018) that would induce similar auditory and somatosensory sensations as real stimulation but without the studied neural effects.

Special measurement considerations in paediatric populations
In paediatric populations, there are some limitations in TMS-EEG measurements associated with brain maturation and currently available TMS-EEG technology. The rMT is significantly higher in children than in adolescents and adults because of ongoing brain maturation (Määttä et al., 2017). Thus, a potential challenge is that the rMT may be higher than the MSO, precluding measurement with the TMS device. This issue may be overcome by using other TMS intensity normalisation methods instead of the rMT, such as active MT, which is estimated similarly to the rMT but during voluntary muscle contraction. This pre-TMS muscle activation significantly lowers the MT compared with that of a resting muscle (Ah Sen et al., 2017). Using active MT instead of rMT to normalize the stimulation intensity leads to smaller amplitude TEPs and oscillations (Saari et al., 2018). However, it could still be feasible as EEG responses can be induced at much lower intensities than MEPs (Komssi et al., 2007). Another potential limitation is the smaller head circumference in children than in adults, as most TMS coils have been designed for adult population head sizes. TMS-compatible EEG caps, on the other hand, are readily available for paediatric populations. Additional confounding factors may be the increasing of the scalp-to-cortex distance with maturation (Määttä et al., 2017(Määttä et al., , 2019 and the study subject's engagement, as EEG is susceptible to the level of attention (Herring et al., 2015;Cardone et al., 2021). Children also have a larger and slower TEP component (N100) compared with adolescents and adults on M1 (Määttä et al., 2017, for more details, see 5. TMS-EEG responses in children), and this needs to be considered when selecting data analysis protocols, especially low-frequency filtering.

Special measurement considerations in geriatric populations
Because of atrophy in geriatric populations, TMS intensity may need to be increased, especially in frontal lobe TMS-EEG studies, compared with that in other study populations (Fjell et al., 2009). In TMS, the stimulating electric field is transferred to the brain via magnetic fields (Rossini et al., 1994;Ilmoniemi et al., 1999). If the coil-to-cortex distance increases because of atrophy, the magnetic fields need to travel a longer distance, making them weaker when they reach the surface of the cortex . This will also lead to weaker induced electric fields. Thus, a higher relative stimulation intensity might be needed to achieve the same effects as in a non-atrophy brain, leading to higher auditory and sensorimotor components affecting the EEG.

Different TMS-EEG pulse paradigms
Currently, two TMS-EEG approaches exist in which the administration of TMS varies with no changes in EEG acquisition. These two approaches are the single-and paired-pulse protocols. In the single-pulse mode, single TMS pulses are delivered with a several second interstimulus interval (ISI). With a single pulse, the goal is to activate a brain area and evaluate the evoked response. By contrast, paired-pulse TMS-EEG consists of applying two consecutive TMS pulses with a short ISI ranging from a few milliseconds to hundreds of milliseconds, followed by a several second pause before the pulse pair is repeated (Ferreri et al., 2011). Depending on the protocol, the paired-pulse sequence results in either facilitation or inhibition of the TEP evoked in the single-pulse mode. There are four paired-pulse protocols: short-interval intracortical facilitation (ISI = 1.1-1.5, 2.3-3.0 and 4.1-4.5 ms [I-wave timings], glutamatergic, TEP facilitation), long-interval intracortical facilitation (ISI = 8-30 ms, glutamatergic, TEP facilitation), short-interval intracortical inhibition (SICI, ISI = 1-6 ms, GABA A ergic, TEP inhibition) and long-interval intracortical inhibition (LICI, ISI = 50-200 ms, GABA B ergic, TEP inhibition) (Chen et al., 2008;Cash et al., 2017).
Aside from the above-mentioned protocols, a special group of pairedpulse protocols includes both peripheral and central nervous system stimulations. A peripheral nerve is first stimulated with electricity, followed by cortical TMS given after a specified delay. ISIs around 18-21 ms result in short-latency afferent inhibition (SAI), and those around 200-1000 ms lead to long latency afferent inhibition (LAI) (Chen et al., 2008;Turco et al., 2018). SAI reflects the inhibitory activity of the cholinergic and GABA A ergic systems, whereas the system mediating LAI is more unclear, but some evidence shows a similar GABA A ergic association as in SAI (Turco et al., 2018). TEP studies with SAI are still rare and in LAI do not exist yet; however, both are expected to inhibit TEP peak amplitudes. For example, there are indications that SAI would at least decrease the amplitudes of P60 and N100 peaks (Ferreri et al., 2012).
Overall, independent of the protocol, the effects of paired-pulse protocols on TEP components depend on the applied TMS parameters (Opie et al., 2017;Rawji et al., 2021), and not all components are modulated (Ferreri et al., 2012;Cash et al., 2017;Premoli et al., 2018). For example, in LICI, a decrease in P30, N45, and P180 amplitudes is dependent on the applied ISI, whereas N100 decreases in amplitude regardless of the ISI (Opie et al., 2017). In SICI, the induced amplitude reduction is strongest when the stimulation intensity of the first pulse is clearly sub-threshold . With an intensity of 70% of rMT, there is an amplitude reduction in N45, P60 and P180 peaks, whereas, with an intensity of 80% of rMT, there are no significant reductions in any of the TEP peaks .

Methods for characterising TMS-EEG responses
The analysis of TMS-EEG does not generally differ from the study of other EEG responses . However, TMS pulses introduce artefacts (for more details, see 4. Challenges in characterising TMS-EEG responses), often substantially immense, in the EEG signal, which must be removed in the pre-processing step. The naive approach is to remove the trials with artefacts or replace the high-amplitude TMS pulse artefacts with zeroes or averages from the data preceding the pulse. More sophisticated methods exist, such as independent component analysis (ICA) (Hyvärinen and Oja, 2000;Hyvärinen, 2012). ICA applied to TMS-EEG seeks to decompose the recorded EEG data into statistically independent components in order to separate the TMS pulse and other artefacts, such as eye blinks, into one or more components. The artefactual components can then be removed from the decomposition, leaving only the neural components intact. Other options include source-based artefact rejection techniques, which suppress unwanted signal components by rejecting topographies that best capture the artefact (Mutanen et al., 2016(Mutanen et al., , 2018. Currently, there are many freely available toolboxes for TMS-EEG pre-processing and analysis, such as EEGLAB (Delorme and Makeig, 2004), FieldTrip (Oostenveld et al., 2011), TMSEEG (Atluri et al., 2016), TESA , and ARTIST (Wu et al., 2018). There are two main approaches to identifying TMS-EEG responses: the time domain and frequency domain methods. These two can also be combined. Below are short descriptions of the most frequently applied signal analysis methods in TMS-EEG studies.

Time domain methods
Time domain methods evaluate the data to find when and where TMS-induced events happen. A commonly used time domain method is characterising the time evolution of a TEP by looking at the latencies and amplitudes of the TEP peaks averaged across trials. The waveform can be studied in a single electrode or multiple electrodes as a spatial average of neighbouring electrodes in a region of interest. However, this time domain method to characterise the TEP peaks includes some ambiguity that the investigator needs to deal with. For instance, the investigator should decide how a peak is determined, as a TEP may consist of significant components across the participants, but they are small or part of another peak. Many studies in the field have not specified their exact peak detection strategies, and there have been no set standards for this.
In addition to individual peak latencies and amplitudes, the TEP can be used to evaluate the level of global neuronal activity, i.e. cortical excitability to TMS. This can be done with a measure known as the global mean field power (GMFP; sometimes referred to as the global mean field amplitude or global field power) (Lehmann and Skrandies, 1980). It is calculated at each time point as the standard deviation across all the electrodes. The overall global post-stimulus brain activity can be calculated as the area under the GMFP curve (Komssi and Kähkönen, 2006). Time points corresponding to TEP peaks produce high GMFP peaks, whereas flatter TEP components produce low GMFP peaks. Some other less-used methods also exist, such as significant current density (SCD). SCD measures cortical excitability by evaluating the absolute current value of significantly activated sources (Casali et al., 2010).
Time domain methods also allow the evaluation of how the TEP spreads in the brain. These methods are the topographical, source location and signal propagation methods. Topographical techniques present the spreading of a TEP in a 2D voltage map of the brain (Nuwer et al., 1999). Although the EEG electrodes are separate, the maps are made continuous using linear or quadratic interpolation methods between neighbouring electrodes. This method gives a general overview of the spreading at each time point and can be used, for example, to track the TEP components or to locate the activity in different frequency bands. As a result of the volume conduction on the EEG signal, the topographical maps present only how the signal behaves on the scalp but give no indication of the source location. More advanced methods are needed to evaluate the source location, such as dipole modelling methods, which aim to find a solution to the inverse problem (Hämäläinen and Ilmoniemi, 1984).
To obtain an estimate of how a specific TEP component changes between two different brain locations, either intra-or inter-hemisphere, the TEP can be analysed via signal propagation analysis (Voineskos et al., 2010;Määttä et al., 2017). This method evaluates the signal by taking the area under a rectified curve of the TEP component in the time domain and comparing it with the area under a rectified curve of the TEP component at a different location but by considering the time taken for the component to transfer to this location. Signal propagation reflects signal spreading efficiency and is affected by white matter characteristics (Voineskos et al., 2010). (Paus et al., 2001;Fuggetta et al., 2005;Rosanova et al., 2009;Ferreri et al., 2012;Gordon et al., 2018) can be quantified using frequency domain methods. An increase in oscillatory EEG power relative to a pre-stimulus baseline is called event-related synchronisation (ERS). By contrast, a decrease is called event-related desynchronisation (ERD) (Pfurtscheller and Lopes Da Silva, 1999). Straightforward estimation of ERD/ERS in main EEG frequency bands, including delta (0.5-3.5 Hz), theta (3.5-7.5 Hz), alpha (7.5-12.5 Hz), beta (12.5-30 Hz) and gamma (> 30 Hz), can be achieved by taking a Fourier transform of the EEG signal.

Time-frequency methods
The time and frequency domain methods can also be combined. One commonly utilised time-frequency analysis method is event-related spectral perturbation (ERSP) (Delorme and Makeig, 2004), which estimates the evolution of EEG power over frequency and time, typically achieved with a power calculation by a short-time Fourier transform or a wavelet transform averaged across trials. Thus, ERSP expresses the magnitude (in decibels) of EEG oscillations and is not sensitive to the EEG phase. Analogous to ERSP, inter-trial coherence (ITC) estimates the level of cortical synchronisation revealed by the EEG signal phase. In TMS-EEG, an ITC value of 1 at a given frequency and a time point relative to the TMS pulse would indicate complete consistency of the EEG signal phase across trials, whereas a value of 0 would mean that the phase is random (Delorme and Makeig, 2004). Aside from ERSP and ITC, synchrony, i.e. consistency in amplitude and phase as a function of frequency between two brain areas, can be analysed with coherence methods. TEP-induced coherence can be used to recognise effective connections between regions of interest (Bortoletto et al., 2015).

Challenges in characterising TMS-EEG responses
Two main phenomena hinder the characterisation of TEPs: nonphysiological artefacts caused by the measurement setup and artefacts caused by unwanted physiological sources. With careful measurement procedures and data analysis protocols, however, their influence can be minimised. Here, we outline the artefacts that could potentially affect TEP characterisation. For extensive reviews of TMS-EEG artefacts and how they can be best prevented and removed, see Farzan et al. (2016), Rogasch et al. (2017) and Varone et al. (2021).

Non-physiological artefacts
As TMS-EEG is an electrical measurement technique, it is vulnerable to environmental electrical interference. Power line noise is the main constant electrical interference; it is a sinusoidal wave at either 50 or 60 Hz, depending on power network characteristics. Artefacts that have specific timings include those associated with the TMS pulse. The most extensive and complicated of these is the high-voltage TMS-induced electromagnetic pulse artefact that occurs immediately after the TMS pulse (Korhonen et al., 2011). TMS coils could also press or move the EEG electrodes and cause vibrations during the pulse (Varone et al., 2021). As EEG electrode impedances are achieved by applying conductive gel, TMS pulses could accumulate charges at the skin-electrode interface, creating artefacts (Varone et al., 2021).

Physiological artefacts
In addition to neural activity, other physiological electrical signals could be mixed in the EEG, such as those originating from eye movements, blinks, heartbeats and facial or scalp muscle contractions (Varone et al., 2021). This is because the function of the eyes, heart and muscles is electrical. If the study participant sweats a lot, this could cause a slow wave artefact in the EEG (Varone et al., 2021).
A novel group of physiological artefacts called non-transcranial TMSevoked potentials was recently identified (Conde et al., 2019). These artefacts are of neural origin and appear because of the measurement setup. Non-transcranial TMS-evoked potentials are neural activities caused by somatosensory and auditory processing of the TMS pulse because often, the participant can feel and hear the TMS pulse. Not much research exists on these artefacts, especially in different age groups or neuropsychiatric disorders, as all studies so far have been conducted in healthy adult populations (Biabani et al., 2019;Conde et al., 2019;Rocchi et al., 2021). As somatosensory and auditory processing in neuropsychiatric disorders may differ from healthy individuals and vary with age (Hagenmuller et al., 2014;Uppal et al., 2016), these artefacts may be unique in each study group.

TMS-EEG responses in children
The central nervous system undergoes major structural and connectional changes during the first two decades of life (Gerber et al., 2009). The main findings related to anatomical changes during development are linear increases in white matter volumes throughout the brain (Giedd et al., 1999) and regionally specific inverted U-shaped trajectories of grey matter (Giedd et al., 1999;Sowell et al., 2003Sowell et al., , 2004Gogtay et al., 2004). These structural changes involve the reorganisation of excitatory and inhibitory circuits (Kilb, 2012); together, these modifications result in developmental refinement in sensory, motor and cognitive abilities. TMS-EEG enables the non-invasive assessment of cortical excitability and inhibition, the functional state of the brain and the functional connectivity of the neural circuits, providing information with high temporal precision that is inaccessible with any other method (Ilmoniemi and Kicic, 2010). These qualities make TMS-EEG a potential tool in developmental neuroscience. The results of the reviewed publications on TMS-EEG in healthy children and adolescents and typical and atypical brain development are summarised in Table 1.

TMS-EEG responses in typically developing children
The first study to explore TMS-evoked EEG responses in children was conducted by Bender and co-workers (Bender et al., 2005). These authors measured TEPs to the right M1 stimulation in 6-to 10-year-old children and young adults in the baseline condition and during a contingent negative variation (CNV) task [CNV is a slow negativity in EEG during periods preparatory to perception and action, and for obtaining the CNV, two stimuli are usually presented: a warning stimulus and, after a short interval, i.e. time of preparation, an imperative stimulus (Walter et al., 1964)]. Suprathreshold stimulation intensity was used or 100% MSO in the case of the rMT exceeding the MSO. In addition, a stimulation intensity of 50% MSO was used to investigate maturational effects at the same absolute intensity in children and adults. The results showed that children had a gigantic TEP component in the N100 latency range (later called N100) with an amplitude over 100 µV; in children, this response was easily distinguishable, even in single trials. Importantly, even though N100 was dependent on the stimulation intensity (stronger stimulations eliciting larger N100), the response was also larger in children when using the same absolute stimulation intensity. Thus, the children's higher rMT did not fully explain their N100 response with a larger amplitude. The topography of N100 lateralised ipsilaterally to the side of stimulation with a centroparietal negative maximum and frontopolar positivity, and its amplitude decreased with increasing age. Furthermore, during late CNV (i.e. during motor preparation and stimulus anticipation), the N100 amplitude was reduced. In summary, the authors concluded that the N100 component was easily distinguishable in 6-to 10-year-old children, showed age-dependent maturation, and was reduced during cortical pre-activation, suggesting the inhibitory nature of this component. No other TEP peaks were described in this study.
Later studies in both neurotypical children and children with neurodevelopmental disorders have confirmed that the children's TEP in response to M1 stimulation is strongly dominated by a large N100 component (Bruckmann et al., 2012;Helfrich et al., 2012;D'Agati et al., 2014;Määttä et al., 2017;Moliadze et al., 2018;Baumer et al., 2020). This is in contrast to results from studies investigating TEP in response to M1 TMS in adults who have consistently described a sequence of positive and negative peaks P30, N45, P60, N100, P180 and N280 (Bonato et al., 2006;Ferreri et al., 2011). In children, the topography of N100 has shown ipsilateral centroparietal negativity with frontopolar positivity, with a source in or near the stimulated M1 (Bruckmann et al., 2012;Helfrich et al., 2012;Määttä et al., 2017). Results regarding N100 latency, on the other hand, have been mixed, with other reports showing N100 latency decrease with increasing age (Bender et al., 2005;Bruckmann et al., 2012). In contrast, others have found no significant differences in N100 latency between children and adults (Määttä et al., 2017). N100 in children has also been shown to be sensitive to non-invasive neuromodulators (Helfrich et al., 2012;Moliadze et al., 2018;Baumer et al., 2020), reducing in amplitude after inhibitory 1 Hz repetitive TMS (rTMS) (Helfrich et al., 2012) and cathodal transcranial direct current stimulation (tDCS) (Moliadze et al., 2018).
A study comparing several aspects of TEPs to M1 stimulation between healthy children, adolescents and adults found that the complexity of the TEP waveform increased with increasing age (Määttä et al., 2017). In children and adolescents, the TEP waveform was characterised by a large N100, and in children, the amplitude of this component was significantly larger than that in older age groups. The amplitude augmentation is located mainly under the stimulated area and in its vicinity. In children and adolescents, the N100 component was preceded by positivity at 50 ms, followed by positivity at 300 ms, instead of the typical P30-N45-P60-N100-P180-N280 sequence detected in adults. Thus, in children and adolescents, there were no distinguishable N45 and P60 peaks. In dipole analysis, the positivity at around The mean area of GMFP declined with age Changes in GMFP amplitude paralleled improvement in manual dexterity Intra-and interhemispheric transmission increased with age The TEP waveform was less complex in children and adolescents than in adults N100 declined with increasing age The topography of N100 differed between the age groups ITC in alpha-and the beta range was increased in children and adolescents compared to adults →Cortical motor circuits develop from strong local connectivity towards more distributed connectivity and more robust integration Moliadze et al.   In both groups, the N100 amplitude was reduced equally during the response preparation task compared to the control situation During the go-and nogo tasks, however, the N100 amplitude was differently modulated in the ADHD-and control group →Findings partly confirm the N100 as a motor cortical marker of inhibition in children Jarczok et al. ISP did not differ between the ASD participants and controls TMS-evoked activity in the left or right motor cortex did not differ between the study groups but correlated negatively with age ISP correlated positively with age →Effective interhemispheric connectivity matures from childhood to young adulthood, no global deficit in connectivity in ASD Baumer et al.  50 ms topographically was located more posteriorly than the adult N45 and more anteriorly and centrally than the typical P60 in adults. Centrally maximal TEP at around 300 ms was negative in adults but positive in younger age groups. The results from this study also demonstrated that the mean area of the GMFP declined with age and correlated with motor dexterity and that interhemispheric signal propagation (ISP) (Voineskos et al., 2010) increased significantly with age. Of note, the measures of GMFP and ISP were strongly influenced by the huge N100 component in children. The results from frequency domain analyses demonstrated that in children and adolescents, TMS of M1 produced a robust and highly significant increase in phase-locking (evaluated with ITC), which was not frequency-specific and covered a large time range. In summary, the authors concluded that children have a different pattern of activation in response to M1 TMS than adults, suggesting that with development, the excitability of the motor system increases, and signal spreading facilitates (Määttä et al., 2017). Age-typical EEG responses to M1 stimulation in representative individuals are seen in Figs. 2 and 3. Although TMS-EEG may be a particularly useful tool for evaluating the development of any neocortical brain area, there is currently only one published study describing TMS-EEG responses in children and adolescents outside of M1 (Määttä et al., 2019). In that study, the right superior frontal cortex was stimulated in three age groups (children, adolescents and adults). From a methodological standpoint, this study differed from previous developmental TMS-EEG studies in the definition of stimulation intensity; in this study, TMS intensity was based on the estimated electric field in the stimulated cortex instead of the conventionally used rMT. This was done to ensure that the stimulation intensities were comparable among age groups despite developmental differences in neuroanatomy. The main findings from this study were that, unlike in the TEP elicited by M1 stimulation, there were no group differences in TEP waveform complexity or in the amplitude of most of the peaks between the age groups (Fig. 4). Instead, the results revealed that the topographical distribution of TEP peaks and the sources computed using minimum norm estimation at different time points differed between the age groups. These findings were interpreted to reflect developmental changes in frontal cortex effective connectivity. Age-typical GMFP responses to M1 and premotor cortex stimulation in representative individuals are seen in Fig. 4.

TMS-EEG in children with neurodevelopmental disorders
Children's N100 TEP to M1 stimulation may represent a promising marker of cortical inhibition/excitability in children with attention deficit hyperactivity disorder (ADHD) (Bruckmann et al., 2012;Helfrich et al., 2012;D'Agati et al., 2014). Bruckmann and co-workers (Bruckmann et al., 2012) were the first to show altered N100 TEP in children with ADHD. They recorded N100 TEP to suprathreshold M1 stimulation in the baseline condition and during response preparation (during the CNV task) in children with ADHD and control children of the same age. In a subsample of subjects, the test asked the participants to press the button that triggered TMS (self-paced movements). In brief, the results showed that response preparation decreased the N100 amplitude and that there was a strong reduction in N100 during movement execution.  However, the main finding of the study was that N100 was smaller in children with ADHD than in the controls, and the response preparationand movement execution-related N100 amplitude decrements were smaller in that group. These results were partly confirmed in a subsequent study, in which N100 to left M1 stimulation was measured in children with ADHD and in healthy controls in a baseline condition and during cued go and no-go tasks (D'Agati et al., 2014). In both groups, the N100 amplitude was reduced during the response preparation task. There were no differences in the N100 amplitude in the baseline condition between children with ADHD and the controls. However, the N100 amplitude was modulated differently in the ADHD and control groups during the go and no-go tasks.
Other than studies on ADHD, up to now, only one study has investigated alterations in TMS-EEG responses in youth with autism spectrum disorder (ASD) (Jarczok et al., 2016) and one study in children with benign childhood epilepsy with centrotemporal spikes (BECTS) (Baumer et al., 2020). In the ASD study, the authors stimulated left M1 using individual 1 mV-threshold stimulation intensity to quantify ISP in boys with ASD and typically developing controls. While there were no significant differences in ISP between the study groups, an effect of age was observed when the groups were pooled together, with the ISP values being higher in the older group (Jarczok et al., 2016). In the study with children with BECTS, the TEP to suprathreshold left M1 stimulation was recorded in the baseline and after a 1 Hz rTMS plasticity protocol (Baumer et al., 2020). Similar to previous reports of healthy children (Määttä et al., 2017), children with BECTS displayed high-amplitude TEPs 50 and 100 ms after stimulation. The results demonstrated that rTMS-induced plasticity varied across children and that there was a strong correlation between baseline N100 amplitude and plastic changes induced by 1 Hz rTMS. That is, rTMS induced an increase in the N100 amplitude in children with larger baseline N100 peaks and a decrease in those with smaller peaks. In addition, rTMS induced changes in the N100 amplitude correlated with language learning in neuropsychological testing, decreased N100 amplitude correlating with worse language learning and increased N100 amplitude correlating with better language learning. In conclusion, the authors suggested that TMS-EEG biomarkers could be developed to identify children at the highest risk for learning problems.

Summary
Taken together, the results of the reviewed studies demonstrate that TMS-EEG can be applied to paediatric populations and may be feasible even in children under school age. TMS-EEG indices change from childhood to adulthood, suggesting that this methodology can track the continued maturation of neural systems. Notably, none of the reviewed studies reported significant adverse effects.
The majority of published developmental TMS-EEG studies have focused on the M1 TMS-evoked N100 component, which has been suggested to be a valuable marker for neurodevelopmental disorders in children and adolescents. Although experimental paradigms have linked this component to inhibitory functioning, studies have not yet evaluated its neural origins and whether it represents neural processes similar to the N100 response of adults.
Overall, developmental and paediatric TMS-EEG studies are still scarce. In addition, all published studies have used cross-sectional designs. Further work with broad and continuous age ranges and longitudinal designs is needed to better understand typical and atypical development.

TMS-EEG responses in physiological and pathological ageing
Physiological and pathological brain ageing are characterised by changes that directly affect brain plasticity, such as changes in cellular connectivity Ward et al., 2015). Yet, a growing body of evidence supports the notion that the central nervous system is capable of changing and adapting throughout life to cope with normal and abnormal experiences and inputs . At any age, the brain is thus capable of acquiring new skills, adapting to new contexts and even recovering functions after nervous system injury or along neurodegenerative processes, when the brain attempts to compensate for lost activity, leading to positive but also negative or aberrant reorganisation ). An increasing number of studies have examined the potential of TMS-EEG in investigating physiological brain ageing and in understanding the neurobiology of dementing illnesses. The results of the reviewed publications on TMS-EEG and ageing are summarised in Table 2.

TMS-EEG responses in physiological ageing
Healthy ageing does not seem to affect TEP morphology. In healthy older adults, the occurrence of the typical TEP waveform, composed of P30-N45-P60-N100-P180-N280 peaks, is well documented (Ferreri et al., 2017a(Ferreri et al., , 2017bNoda et al., 2017Noda et al., , 2021Opie et al., 2018Opie et al., , 2020. However, there are several reported differences in the amplitude, timing and topography of the peaks, and experimental paradigms or tasks differently modulate the peaks in younger and older adults. In their seminal paper, Ferreri and co-workers (2017a) compared healthy younger and older adults using suprathreshold stimulation to the left M1. Their results demonstrated a significantly larger amplitude P30 and a smaller amplitude P180 in older adults compared with young adults globally. Furthermore, regional differences were seen in P30, N45, N100 and N280 amplitudes. P30 was larger in younger adults close to the stimulation site but larger in older adults frontally. The N45 component was smaller in the elderly in the left central and parietal regions, and N100 was larger frontally and smaller centrally in the younger adults, whereas N280 was larger at the stimulation site in the younger group and larger frontally in the elderly. The authors concluded that normal ageing may lead to changes in the excitability of specific neural circuits and that some other alterations may, in turn, result from compensatory activation induced by those changes. For example, the increased excitability seen in the elderly in the prefrontal cortex at 30 ms was suggested to represent a compensatory mechanism that counteracts the significantly decreased excitability of M1 (Ferreri et al., 2017a). Similar results were reported by Noda and co-workers (2021), who demonstrated that after M1 stimulation, older adults had significantly smaller N45 and P180 peaks compared with young adults in the left central region (Noda et al., 2021).
However, some of the abovementioned results were contrasted in another M1 TMS-EEG study (Opie et al., 2018). In that study, the authors recorded TMS-EEG responses to left M1 stimulation in younger and older adults. In addition to single-pulse stimulation, LICI was measured. In the single-pulse paradigm, P30 was unaffected by age, and N45 was increased in older adults, contrary to the results presented above (Ferreri et al., 2017a), and N100 and P180 showed altered spatial distribution. The latency of P30 was shorter, and that of P180 was longer in older adults. In addition, inhibition of N100 and P180 was increased in older adults following LICI. No differences were found in the oscillatory activity between the groups in any conditions (Opie et al., 2018).
Two studies reported TMS-EEG responses to frontal cortex TMS in physiological ageing. Casarotto et al. (2011) stimulated the left superior frontal cortex at an intensity that produced an electric field of approximately 110 V/m on the targeted cortex. TEP analysis focused on early responses in the 10-45 ms post-stimulus interval. In this study, TEP peaks were not analysed per se, but frontal cortex excitability was examined as the amplitude of early and local neural response to TMS and was measured by integrating the absolute current value of SCDs over a cortical region surrounding the stimulated site. No differences were observed between younger and older adults (Casarotto et al., 2011). In a more recent study, the dorsolateral prefrontal cortex (DLPFC) was stimulated with an intensity that produced 1 mV peak-to-peak amplitudes in MEPs and P30, N45, P60, N100 and P180 were analysed in five Table 2 TMS-EEG studies in physiological and pathological ageing.

Study
Methods Participants Analyses Main findings and conclusions (Ferreri et al., 2017a) Navigated single-pulse TMS with figureof-eight coil, left M1 Intensity: 120% rMT 12 older subjects (5 F, 7 M) mean age 67.6 years; range 59 -81 years 12 healthy young (7 F, 5 M), mean age 24.5 years; range 18 -30 years 19-channel EEG GMFP TEP amplitudes and latencies Current density analysis MEPs in the right FDI P30, N45, P60, N100, P180 and N280 peaks were identified in both groups GMFP was decreased in the elderly subjects compared to the young subject No latency differences P30 was globally increased in the older subjects N45 was decreased in the older subjects at C3, P3 and P7 N100 was larger at FP1 and F3 in the younger groups and smaller at Cz compared to the elderly P180 was globally larger in the younger subjects N280 was larger at C3 and smaller at Fp1 and Fp2 in the younger compared to older subjects The current density maps differed between the groups →Physiological ageing leads to alterations in cortical activity Ferreri et al. (2014), ( Ferreri et al., 2017b) Navigated single-pulse TMS, figure-ofeight coil, left M1 Intensity: 120% rMT 11 older subjects, mean age 67.7years, range 57 -81 years 8 younger subjects, mean age 24.6, range 18 -30 years 19-channel EEG MEPs in the right FDI MEP amplitudes were divided into two subgroups consisting of "high" and "low" MEPs, based on the 50th percentile of their amplitude distribution. The pre-stimulus EEG (power spectrum and spectral coherence) from M1 and correlated areas were compared for the "high" and "low" MEPs.
In both young and old subjects, significantly larger MEPs were evoked when the stimulated M1 was coupled in the beta-2 band in the ipsilateral prefrontal cortex. Conversely, the MEP size was modulated only in young participants when the M1 and ipsilateral parieto-occipital cortices were coupled in the delta band. This coupling was significantly higher in elderly brains than in young brains, both for high and low MEPs. →Direct evidence of an age-related stochastic linking of motor-related cortical areas via oscillatory synchronization in preferential EEG bands, which could contribute to the determination of M1 excitability Noda et al.
Single-pulse TMS, figure-of-eight coil, left M1 and DLPFC Intensity: individual 1 mV MEP intensity SAI protocol TMS-EEG in the baseline and after SAI Cognitive tests 12 older subjects (6 F, 6 M) aged 72 ± 9 years 12 younger subjects (6 F, 6 M) aged 29 ± 12 years 64-channel EEG TEP amplitude comparison before and after SAI protocol MEP in the right FDI P30, N45, P60, N100 and P180 TEPs were identified in both groups M1 stimulation: N45 and N100 components were less modulated by SAI protocol in older subjects DLPFC stimulation: N100 was less modulated by SAI in the older groups DLPFC: SAI-induced modulation of P60 and N100 correlated with cognitive functions in older adults →SAI in DLPFC may be a potential marker of cholinergic tone Opie et al. (2018) Paired-pulse TMS, with figure-of-eight coil, left M1 LICI protocol with conditioning and test stimuli at 120% rMT TMS-EEG before, during and after LICI 17 older adults aged 71.4 ± 1.4 years 17 younger subjects aged 24.2 ± 1.1 years Sex not specified 62-channel EEG GMFP TEP amplitudes Topographical maps TMS-evoked oscillations GMFP did not differ between the groups P30, N45, N100 and P180 were elicited in both groups After the test stimulus, in the GMFP, the latency of P30 was shorter and P180 longer in older adults, and the N45 amplitude was larger in the older subjects. The topography of N100 and P180 differed between the groups. LICI of N100 and P180 was stronger in older subjects No group differences in the oscillatory activity →Potentiation of cortical inhibition may accompany ageing Opie et al. (2020) Single-pulse TMS, figure-of-eight coil, left M1 Intensity: 120% rMT TMS-EEG before and after fatigue exercise 17 older adults (7 F, 10 M) aged 68.3 ± 5.6 years 23 young adults (11 F, 12 M) aged 22.3 ± 2.2 years 62-channel EEG TEP amplitudes MEP in the right FDI P30, N45, N100 and P180 were identified in both groups In older adults, the N45 peak was reduced during and after fatigue, and also N100 was modified In younger adults, P30 was increased after fatigue →Fatiguing exercise associated with GABA A mediated inhibition in older adults Noda et al.  P30, N100 and P200 peaks were identified in EEG GMFP was reduced in AD patients compared to controls at 30-90 and 280-400 ms post-stimulus MCI group had lower N100 amplitude in the GMFP, as well as a reduction between 300 and 370 ms post-stimulus P30 peak amplitude in AD patients was significantly reduced AD group had lower spectral density in the < 50 Hz frequencies →TMS-EEG may provide a new tool for examining the degree and progression of dementia Julkunen et al.
Navigated single-pulse TMS, figure-ofeight, left superior frontal cortex Intensity: 110 V/m on the targeted cortical surface Cognitive assessment 9 healthy young subjects (2 F, 7 M) aged 31 ± 4.5 years 9 healthy elderly subjects aged 72 ± 8.4 years (5 F, 4 M) 9 CE patients aged 72 ± 7.1 years (5 F, 4 M) 60-channel EEG Analyses in the 10-45 ms post-stimulus period Grand average TEPs SCD TEP waveform at FCI consisted of 2 peaks within the first 45 ms in healthy young and elderly subjects and one peak in AD patients The local mean SCD integrated over 10-45 ms was reduced in AD patients compared to other groups →Once the cortical electric field induced by TMS is standardized across subjects, TEPS are not significantly affected by ageing, TMS-EEG may detect pathological changes across the lifespan Ferreri et al.
Single-pulse TMS, figure-of-eight coil, left M1 Intensity: 120% rMT Cognitive assessment 12 right-handed AD patients (7 F, 5 M) aged 72.4 ± 5.9 years 12 healthy right-handed controls (6 F, 6 M) aged 68.6 ± 7.1 years 32-channel EEG GMFP TEP amplitudes and latencies Topographical maps Current density maps Control subjects displayed the typical P30, N45, P60, N100, P180 TEP waveform Patients with AD had an additional component at 80 ms GMFP amplitude was increased in AD patients between 20 and 150 ms P30 was increased in the sensorimotor region and P60 in the motor cortex in AD patients compared with controls N100 latency was longer in the AD patients compared with controls In the current density maps, the cortical activation in AD patients localized in the stimulated M1, whereas in controls, the activation propagated →Sensorimotor system is rearranged and hyperexcitable in AD Kumar et al.

Methods Participants Analyses Main findings and conclusions
Cognitive/neuropsychological assessment discriminating between mild and moderate AD →Prefrontal TMS-EEG can track disease progression in AD Kumar et al. (2020) Navigated single-pulse TMS figure-ofeight coil, left DLPFC Intensity: individual 1 mV MEP threshold intensity TMS-EEG before and after PAS (see Kumar et al., 2017) in the baseline and 1, 7, and 14 days after rPAS Cognitive/working memory assessment 32 CE patients (9 F) aged 76.5 ± 6.8 years, out which 16 received rPAS (9 F, 7 M) aged 76.5 ± 6.8 years, and 16 control PAS (7 F, 9 M) aged 76.4 years ± 6.0

64-channel EEG
Rectified area under the curve PAS induced a potentiation of the cortically evoked activity in both groups There were no significant differences between the active and control rPAS groups on DLPFC plasticity or working memory performance after the rPAS intervention On post hoc tests, with-in group analysis, only the active rPAS group demonstrated an increase in DLPFC plasticity from baseline comparison to post rPAS comparison →Active rPAS may enhance cognition in AD patients, further research is needed Assogna et al. Those who clinically converted to AD demonstrated lower ITC values than those who remained cognitively stable both in beta and gamma bands Cortical excitability (GMFP) and the alpha ITC were similar between groups Those who clinically converted to AD demonstrated more GMFP peaks and a lower parameter related to the TEP shape and reflecting timespecific alterations than those who remained cognitively stable Gamma ITC correlated positively with the TEP shape parameter and negatively associated with the number of GMFP peaks →A parameter related to the TEP shape and reflecting time-specific alterations in TMS-evoked activity and gamma ITC performed best in distinguishing between those who clinically converted to AD and those who remained cognitively stable Abbreviations: AD=Alzheimer´s disease; ADM=abductor digiti minimi muscle; DLPFC=dorsolateral prefrontal cortex; EEG=electroencephalography; ERSP=event-related spectral perturbation; F=female; FDI=first interosseous dorsal muscle; FTD=fronto-temporal dementia; GMFP=global mean field power; ITC= inter-trial coherence; LICI=long-interval intracortical inhibition; M=male; M1 =primary motor cortex; MEP=motorevoked potential; MCI=mild cognitive impairment; PAS=paired associative stimulation; rMT=resting motor threshold; rPAS=repetitive PAS; rTMS=repetitive TMS; SAI=short-latency afferent inhibition; SCD=significant current density; TEP=TMS-evoked potential; TMS=transcranial magnetic stimulation; cortical regions (Noda et al., 2021). As a result, the older subjects had a significantly smaller N45 in the left frontal region and prolonged N45 and P60 latencies in the right central region. Furthermore, in current density maps, the spatio-temporal patterns of cortical activation propagated differently between aged and young brains. In that study, the N45 amplitude attenuation was interpreted to be related to a relative GABA A decrease, and latency delays were suggested to reflect receptor-mediated inhibitory and/or its excitatory dysfunctions (Noda et al., 2021). Two other studies examined the age-related effects of experimental modulation on TMS-EEG responses. Noda and co-workers (2017) compared the effects of left M1 and left DLPFC SAI, and Opie et al. (2020) compared the effects of exercise on M1 TEPs between younger and older adults. The main findings from the SAI study demonstrated that older participants showed significantly less modulation of the N100 component in both M1 and frontal cortex experiments compared with the younger participants. The modulation of the P60 and N100 peaks by DLPFC SAI in older participants correlated with cognitive functions . In the exercise experiment, fatigue affected the TEP peaks differently in young and older subjects. For young participants, P30 post fatigue amplitude was increased, and P180 decreased relative to pre-fatigue, whereas these peaks were unaffected in the elderly; the amplitude of the N45 potential in older adults was significantly reduced by fatigue but unaffected in young participants (Opie et al., 2020). Of note, neither of these studies compared the baseline responses between the groups.
TMS-EEG recordings also clarified that the instantaneous state of M1 excitability influences MEP amplitude (Ferreri et al., 2014) and that this impact is age-dependent (Ferreri et al., 2017b). In their work, as physiological ageing is related to a decrease in motor performance and changes in excitability and connectivity strength within cerebral sensorimotor circuits, Ferreri and co-workers aimed to explore whether ageing also affects EEG-MEP interactions. In healthy aged subjects and young volunteers, they divided the MEP amplitudes into two subgroups consisting of high and low MEPs based on the 50th percentile of their amplitude distribution. Then, they analysed the characteristics of the pre-stimulus EEG from M1 and correlated areas separately for the high and low MEPs, comparing the two conditions. In both young and old subjects, significantly larger MEPs were evoked when the stimulated M1 was coupled in the beta-2 band with the ipsilateral prefrontal cortex. Conversely, the MEP size was modulated only in young participants when the M1 and ipsilateral parieto-occipital cortices were coupled in the delta band. The elderly did not show this kind of pattern. Importantly, this coupling was significantly higher in elderly brains than in young brains, both for high and low MEPs. These results suggest direct evidence of age-related stochastic linking of motor-related cortical areas via oscillatory synchronisation in preferential EEG bands and may contribute to the determination of M1 excitability in several physiological and pathological conditions.

TMS-EEG findings in pathological ageing
In their pilot study, Julkunen and co-workers stimulated the left and right M1 with suprathreshold stimulation in subjects with Alzheimer's disease (AD), subjects with mild cognitive impairment (MCI) and agematched healthy controls. They found that GMFP was reduced in AD patients compared with the controls at 30-90 ms and 280-400 ms poststimulus periods. They also found a significantly reduced TMS-evoked P30 in the AD subjects, whereas subjects with MCI showed slightly increased activity in the P30 time window (Julkunen et al., 2008). In a subsequent study by the same group in which the data were further analysed, the authors found significant correlations between the P30 amplitude and dementia rating scales. The P30 amplitude also showed good specificity and sensitivity in identifying healthy subjects from Fig. 5. Global mean field power (GMFP) responses and the topography of the TMS-evoked potential peaks presented as scalp distribution maps in a representative Alzheimer's disease patient and a control subject. Note the additional peak in the Alzheimer's disease patient's GMFP around 80 ms that does not exist in the control subject. The TMS pulse was given at a time of 0 ms to left primary motor cortex. Representative subjects from (Ferreri et al., 2016). patients with AD and MCI (Julkunen et al., 2011). In accordance with the aforementioned M1 studies showing a reduction in the early part of TEP, a study analysing TEP after frontal cortex stimulation demonstrated that excitability as indexed by SCD over 10-45 ms post-stimulus was reduced in AD patients compared with healthy young and elderly controls (Casarotto et al., 2011). In conclusion, the results from these early studies suggested that TMS-EEG may present a new tool for examining cortical alterations and assessing the degree and progression of dementia and that TMS-EEG could provide a potential biomarker for identifying MCI and AD subjects (Julkunen et al., 2008(Julkunen et al., , 2011Casarotto et al., 2011).
Contrary to these results, a more recent TMS-EEG study of M1 demonstrated strong cortical hyperexcitability as an indexed increased activity at around the P30-P60 latency window in early AD, despite the lack of clinically evident motor manifestations (Ferreri et al., 2016). In this study, the GMFP analysis revealed an increase in TEP amplitude between 20 and 150 ms post-stimulus in AD patients. More specifically, the analysis showed a significant group difference in the P30 and P60 wave amplitudes. Intriguingly, the results also revealed an adjunctive peak in AD patients around 80 ms that was absent in the healthy controls, as shown in Fig. 5. The authors suggested that this additional peak may represent a reverberant local circuit in the sensorimotor system, possibly supported by neural degeneration leading to disconnection and/or aberrant connectivity, while the concomitant higher local sensorimotor activity at around the P30-P60 time range could represent a compensatory mechanism for the decreased functional connectivity (Ferreri et al., 2016). In support of these results, frontal cortex TMS-EEG research by Bagattini and co-workers found that the amplitude of the P30 potential was augmented with more severe cognitive decline, and its increase predicted disease severity (Bagattini et al., 2019).
Thus, some of the findings from the single-pulse TMS-EEG literature reviewed above speak to a decrease in early TEP peaks in AD patients. In contrast, other studies claim that AD is associated with cortical hyperexcitability, as expressed by increased early EEG activity. The reasons for this discrepancy could stem from the fact that the experiments in the reviewed articles vary across a number of important dimensions, such as stimulation target selection, stimulation intensity, TMS-EEG analysis methods and patient selection. For example, in a study by Ferreri and coworkers (Ferreri et al., 2016), the AD patients were newly diagnosed and not on medication that influences corticospinal excitability, whereas, in other studies, the experiment was conducted in patients receiving acetylcholinesterase inhibitor medication (Julkunen et al., 2008(Julkunen et al., , 2011. TMS-EEG research has also provided evidence of impaired cortical plasticity in AD. Kumar and co-workers (Kumar et al., 2017) administered paired associative stimulation (PAS, non-invasive electrical stimulation of the right median nerve in the wrist, followed by TMS of the left DLPFC after a 20 ms delay) and recorded TMS-EEG to left DLPFC stimulation before and after PAS administration (Kumar et al., 2017). The results demonstrated that AD patients experienced significantly less PAS-induced long-term potentiation (LTP; indexed as augmentation in the rectified area under the curve) in the EEG compared with the controls. In another study by the same authors (Kumar et al., 2020), repetitive PAS (rPAS) intervention was used on the DLPFC in AD patients, and PAS-induced LTP was measured similarly as in their previous study. Within-group analyses showed promising results, mainly that right after the intervention (i.e. post-day 1), active rPAS and not control rPAS results in enhanced DLPFC plasticity.
Some work has also shown that TMS-EEG can help evaluate treatment responses in patients with AD. Koch and co-workers (Koch et al., 2018) utilised TMS-EEG in evaluating neurophysiological modifications induced by rTMS treatment in AD. In their study, the authors administered subthreshold single-pulse TMS targeted over the precuneus and the left posterior parietal cortex, which served as a control site. The results revealed that cortical activity, measured using GMFP and TEP peak amplitudes after precuneus stimulation, increased 60-90 ms post-stimulus after rTMS but not after sham treatment. In addition, beta ERSP and ITC increased over the parietal region. TMS-EEG activity to parietal cortex stimulation did not change after the rTMS. In another study, TMS-EEG was utilised to assess the effects of palmitoylethanolamide combined with luteoline, an endocannabinoid with anti-inflammatory and neuroprotective effects in patients with frontotemporal dementia (Assogna et al., 2020). The authors applied single-pulse TMS at subthreshold stimulation intensity over the left DLPFC and posterior parietal cortices, and GMFP was used to evaluate the medication effects. The results demonstrated a higher overall left DLPFC cortical activity in the post-treatment evaluation compared with the pre-treatment evaluation, with no impact on specific GMFP peaks. In addition, gamma and beta ERSP of the left DLPFC showed a significant increment in the post-treatment evaluation compared with the pre-treatment. Analysis of GMFP elicited by parietal cortex stimulation did not reveal any significant effect.
Finally, a recent study investigated the neurophysiology of the sensorimotor cortex in pathological ageing. The 6-year follow-up study evaluated individuals with amnesic MCI who might convert to AD. The authors hypothesized that some individuals might represent-already early on-some plastic rearrangements induced by neurodegeneration that could be used to predict future conversion to AD . The results demonstrated that the M1 excitability was reduced in amnestic MCI subjects compared to controls. Moreover, a sensorimotor-specific abnormality of local EEG synchronization-specifically in beta and gamma bands-discriminated subjects who will convert to AD. In addition, a parameter related to the TEP waveform shape, reflecting time-specific alterations in TMS-induced activity, predicted the conversion from amnestic MCI to AD with high accuracy . The authors concluded that cortical changes reflecting synchronisation deficits within the cortico-basal ganglia-thalamic-cortical loop in amnesic MCI might reflect pathological processes underlying AD. These changes could be tested in larger cohorts as possible early neurophysiological biomarkers of AD .

Summary
While the developing nervous system may be more capable of modifications, dynamic plastic changes can also be documented in the adult nervous system. Indeed, the adult central nervous system is emerging as a dynamically adapting and changing system in which modifications are driven, amongst other things, by afferent input, behavioural influences of functional significance and pathophysiological modifications. In this respect, TMS-EEG offers a means to characterise significant deviations from physiological ageing and may hold promise for detecting markers of the early stage of disease or its progression. The emerging possibility of modulating cortical excitability in healthy and pathological brains, along with a complete comprehension of the mechanisms of ageing, would open wholly new and intriguing, even therapeutic, perspectives.

Conclusions and future directions
TMS-EEG is a versatile and unique method that can be used across the lifespan. The uniqueness of TMS-EEG is based on the possibility of studying the neurophysiology of any neocortical brain area noninvasively and its capability to evaluate the function and dynamics of glutamatergic, GABA A ergic and GABA B ergic systems separately. The main factors hindering the identification of TMS-EEG responses are physiological and non-physiological artefacts that interfere with the measured EEG signal. Both maturation and ageing impact TMS-evoked responses, but they do so distinctly. Furthermore, maturation influences TEPs differently based on the target cortical area. In M1, maturation decreases the amplitude of TEP. On the frontal cortex, on the contrary, maturation does not change the amplitude of TEP, but it alters effective connectivity. Moreover, children and adolescents are missing N45 and P60 peaks evident in adults.
Healthy ageing does not change TEP morphology, but conditions associated with ageing, such as MCI and AD, impact TEP amplitudes. Some of these alterations, such as P30, have shown promise in discriminating between healthy and demented subjects and in evaluating dementia severity. Interestingly, AD is also associated with an additional TEP component, P80, which does not appear in other age groups or controls. Furthermore, MCI subjects who will convert to AD show a sensorimotor-specific disruption of local EEG synchronization in beta and gamma bands and a global alteration of the TEP shape.
Overall, TMS-EEG is a safe method across the lifespan, as none of the reviewed studies reported any safety concerns. Most research to date has focused on adults, and in children, TMS-EEG is still an underused method. The studied age groups have been commonly selected so that they are clearly separate. Furthermore, studies have primarily targeted the M1 and frontal areas, and many other neocortical brain areas have not yet been examined. So far, sex differences have not been investigated in relation to maturation or ageing. Given that nearly all neuropsychiatric disorders have different prevalence and symptoms between the sexes, examining sex differences in neuromaturation could be particularly relevant for paediatric neuropsychiatry. Finally, most studies thus far are cross-sectional, and only one longitudinal study has been conducted. Longitudinal studies are essential for understanding normal development and disease progression.
To conclude, future studies should investigate other brain areas outside the frontal and motor areas, have continuous age ranges, include both sexes, and use a longitudinal study design to evaluate neurophysiological changes across the lifespan more accurately. More studies in children are also needed to understand typical and atypical development. Finally, the possible role and potential of TEP components as biomarkers separating typical and atypical maturation and healthy and pathological ageing should be investigated in extensive multi-site studies.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.