Motor cortex plasticity induced by theta burst stimulation is impaired in patients with obstructive sleep apnoea


  • George M. Opie,

    1. Discipline of Physiology, School of Medical Sciences, The University of Adelaide, Adelaide, SA, Australia
    Search for more papers by this author
  • Peter G. Catcheside,

    1. Discipline of Physiology, School of Medical Sciences, The University of Adelaide, Adelaide, SA, Australia
    2. Adelaide Institute for Sleep Health, Repatriation General Hospital, Daw Park, SA, Australia
    3. Faculty of Health Sciences, School of Medicine, Flinders University, Adelaide, SA, Australia
    Search for more papers by this author
  • Zafar A. Usmani,

    1. Adelaide Institute for Sleep Health, Repatriation General Hospital, Daw Park, SA, Australia
    Search for more papers by this author
  • Michael C. Ridding,

    1. Robinson Institute, School of Paediatrics and Reproductive Health, University of Adelaide, Adelaide, SA, Australia
    Search for more papers by this author
  • John G. Semmler

    Corresponding author
    • Discipline of Physiology, School of Medical Sciences, The University of Adelaide, Adelaide, SA, Australia
    Search for more papers by this author

Correspondence: Dr J. G. Semmler, as above.



Obstructive sleep apnoea (OSA) is a respiratory condition occurring during sleep characterised by repeated collapse of the upper airway. Patients with OSA show altered brain structure and function that may manifest as impaired neuroplasticity. We assessed this hypothesis in 13 patients with moderate-to-severe OSA and 11 healthy control subjects. Transcranial magnetic stimulation was used to induce and measure neuroplastic changes in the motor cortex by assessing changes in motor-evoked potentials (MEPs) in a hand muscle. Baseline measurements of cortical excitability included active (AMT) and resting motor thresholds (RMT), and the maximal stimulator output producing a 1-mV MEP. Intracortical inhibition (ICI) was investigated with short- and long-interval ICI paradigms (SICI and LICI, respectively), and neuroplastic changes were induced using continuous theta burst stimulation (cTBS). At baseline, differences were found between groups for RMT (9.5% maximal stimulator output higher in OSA) and 1-mV MEPs (10.3% maximal stimulator output higher in OSA), but not AMT. No differences were found between groups for SICI or LICI. The response to cTBS was different between groups, with control subjects showing an expected reduction in MEP amplitude after cTBS, whereas the MEPs in patients with OSA did not change. The lack of response to cTBS suggests impaired long-term depression-like neuroplasticity in patients with OSA, which may be a consequence of sleep fragmentation or chronic blood gas disturbance in sleep. This reduced neuroplastic capacity may have implications for the learning, retention or consolidation of motor skills in patients with OSA.


Obstructive sleep apnoea (OSA) is a respiratory condition occurring during sleep characterised by periods of upper-airway collapse resulting in reduced (hypopnoea) or completely absent (apnoea) airflow (Eckert & Malhotra, 2008). Most apnoeic/hypopnoeic periods end with arousal from sleep, resulting in sleep fragmentation and altered sleep architecture (i.e. alterations in the proportion of time spent in different sleep stages). Furthermore, apnoea-related reductions in airflow lead to hypoxia and hypercapnia. A range of neurocognitive impairments have been associated with OSA, including decreases in memory, attention and executive function (Campana et al., 2010). These symptoms negatively impact the daily life of patients, with reports of difficulty accomplishing routine work tasks (Ulfberg et al., 1996), and increased risk of motor vehicle (Tregear et al., 2009) and occupational (Ulfberg et al., 2000) accidents.

Although the pathophysiology of these cognitive deficits remains largely unknown, several studies have shown alterations in brain structure and function in patients with OSA. For example, patients with OSA show decreased grey matter in various brain regions (Joo et al., 2010b; Morrell et al., 2010; Torelli et al., 2011), and differences in neural activation of sensorimotor and autonomic brain regions during respiratory challenges (Zimmerman & Aloia, 2006). Furthermore, studies using transcranial magnetic stimulation (TMS) have shown increased motor thresholds (Joo et al., 2010a) and prolonged cortical silent periods (CSPs) in patients with OSA (Civardi et al., 2004; Grippo et al., 2005), reflecting cortical hypoexcitability (see Civardi et al., 2009). These changes may result from diminished corticospinal fibre integrity in patients (Macey et al., 2008), and are presumed to be a consequence of chronic intermittent hypoxaemia and sleep fragmentation (Morrell et al., 2003; Ohga et al., 2003).

Plasticity of cortical circuits is an important component of the brain's ability to adapt, learn and recover from injury. It is also known to be a fundamental process in memory function, which has been shown to be defective in OSA (Jackson et al., 2011). The application of repetitive trains of TMS (rTMS) is commonly used to non-invasively induce plasticity of neural circuits within the human motor cortex. A recently developed protocol known as continuous theta burst stimulation (cTBS) uses a specific pattern of rTMS that can suppress motor-evoked potentials (MEPs) for up to 1 h (Huang et al., 2005), and is thought to induce long-term depression (LTD)-like synaptic changes (Cardenas-Morales et al., 2010) within cortical circuits (Di Lazzaro et al., 2005).

The primary aim of this study was to examine motor cortex plasticity in patients with moderate-to-severe OSA using cTBS. As mouse models of OSA have shown impaired hippocampal plasticity (Xie et al., 2010) and sleep fragmentation could affect processes that promote plasticity (Diekelmann & Born, 2010), we hypothesised that cortical plasticity would be reduced in patients with OSA. Furthermore, as increased intracortical inhibition (ICI) can reduce neuroplastic capacity of cortical circuits (Ziemann et al., 2001), a secondary aim of this study was to quantify baseline ICI in patients with OSA compared with controls. Based on previous observations of increased CSP in OSA (Civardi et al., 2004; Grippo et al., 2005), which may reflect an increase in cortical inhibitory tone, we expected to see elevated levels of ICI in patients with OSA.

Materials and methods


Patients with moderate-to-severe OSA [apnoea–hypopnoea index (AHI) ≥ 20 events/h] who had not started continuous positive airway pressure (CPAP) treatment were recruited through Adelaide Institute for Sleep Health outpatient clinics (Repatriation General Hospital, South Australia). Control subjects were recruited from the University of Adelaide and wider community by advertisement.

All subjects were right handed (assessed with the Edinburgh Handedness Questionnaire). Subjective daytime sleepiness was assessed using the Epworth Sleepiness Scale (ESS; where a score of ≥10 indicates severe sleepiness), while physical activity was measured using a short, self-administered questionnaire (Baecke et al., 1982). Subject weight and height were also measured at the beginning of experimentation. Exclusion criteria applied to all subjects were a history of stroke, history of neurological or psychiatric disease, or currently taking psychoactive medication. Several subjects from both patient and control groups reported regular use of medications for a range of conditions. These included proton pump inhibitors (Pantoprazole, Esomeprazole), beta blockers (Metoprolol), alpha blockers (Minipress), statins (Lipitor), diuretics (Amiloride), angiotensin-1 receptor antagonists (Sartans), calcium channel blockers (Verapamil, Lercanidipine), ace inhibitors (Ramipril), bisphosphanates (Risendronate) and vitamin D supplements. However, participation was subject to medication not having neurological side-effects that may have affected TMS measurements.

A total of 14 patients with OSA and 14 control subjects were recruited for this study. However, one patient with OSA and three control subjects were excluded from the analysis (see 'Results'). Therefore, data from 13 patients with OSA (average ± SD age: 42.6 ±10.2 years, two females) and 11 age-matched, healthy control subjects (average age: 43.0 ± 10.3 years, two females) were included in the study. Each subject provided written informed consent before participating in the project. The study was approved by the University of Adelaide Human Research Ethics Committee and the Southern Adelaide Clinical Human Research Ethics Committee. All experimentation was conducted in accordance with the Declaration of Helsinki.

Overnight polysomnography

Patients with OSA and controls underwent full attended in-laboratory polysomnography to diagnose (patients with OSA) or rule out (controls) a sleep disorder. Subjects attended the Adelaide Institute for Sleep Health at approximately 21:00 h for overnight sleep assessment. On arrival, they were familiarised with the surroundings and allowed to dress comfortably for sleep, after which they were instrumented for study. Sleep studies were recorded using a Compumedics E-series system and software (Pro-Fusion; Compumedics, Melbourne, Australia). Signals included electroencephalography (EEG; from C3/M2 and C4/M1), eye movements via two-channel electrooculography (EOG), electrocardiogram (lead II), chin electromyography (EMG), leg movements, snoring and posture. EEG, EOG and EMG electrodes were checked, and electrodes reapplied to achieve impedances below 5 kΩ. Respiratory signals included a nasal pressure cannula and oronasal thermister, thoracoabdominal bands to assess chest and abdominal movement, and finger pulse oximetry to determine arterial blood oxygen (O2) saturation. All measurements were continuously recorded from lights-out (approximately 22:30 h) until the end of the study the following morning (approximately 06:00 h). Sleep and respiratory signals were analysed by an accredited sleep technician, blinded to group allocation, and according to current internationally agreed standards (Iber et al., 2007). AHI was determined using American Academy of Sleep Medicine ‘alternative’ criteria (Iber et al., 2007; Ruehland et al., 2009). Within these criteria, respiratory events are scored as an apnoea following complete cessation of airflow for ≥ 10 s, whereas hypopnoeas are scored based on a 50% reduction in airflow with an associated 3% reduction in O2-saturation, or an arousal from sleep (Iber et al., 2007). An AHI of < 10 events/h was used to rule out OSA. The arousal index (AI; number of arousals per hour of sleep) was calculated to produce an index of sleep fragmentation, and sleep efficiency was obtained by dividing the amount of time spent asleep by the total amount of time available for sleep (i.e. the lights-out duration).


On a separate day, subjects attended the University of Adelaide for neurophysiological testing. This session took place in the afternoon or evening to avoid time of day effects (Sale et al., 2007). During testing, subjects were seated in a comfortable chair with their right forearm resting on a padded arm-rest and right hand in a pronated position. Surface EMG was recorded from the first dorsal interosseous (FDI) and abductor digiti minimi muscles of the right hand. Two Ag–AgCl electrodes arranged in a belly-tendon montage were used. EMG signals were amplified (1000 ×), filtered (20 Hz–1 kHz), digitised at 2 kHz using a CED1401 interface (Cambridge Electronic Design, Cambridge, UK) and stored offline for analysis. TMS was applied to the left primary motor cortex using a figure-of-eight coil (external wing diameter 9 cm) with two Magstim 200 magnetic stimulators connected through a Bistim unit (Magstim, Dyfed, UK). The coil was held tangentially to the scalp at an angle of 45° to the sagittal plane with the handle pointed backwards, producing a current flow in the brain with a posterior to anterior direction. The coil was positioned on the scalp over the location producing an optimum response in the relaxed FDI muscle. This location was marked on the scalp for future reference and continually checked throughout the experiment. Stimuli were delivered at a rate of 0.2 Hz, with a 10% variance to avoid stimulus anticipation.

Baseline measures

Measurements of corticospinal excitability included resting motor threshold (RMT), active motor threshold (AMT) and MEP amplitude. RMT was defined as the minimum TMS intensity producing a response amplitude ≥ 50 μV in three out of five trials in the relaxed FDI muscle. AMT was defined as the minimum TMS intensity producing a response amplitude ≥ 200 μV in three out of five trials while FDI was voluntarily activated (5–10% of maximum voluntary contraction). Force feedback was provided via an oscilloscope placed in front of the subject at eye level. The TMS intensity producing a MEP amplitude of approximately 1 mV in resting FDI (MEP1 mV) was determined before the cTBS intervention. Fifteen trials were recorded at this intensity to establish baseline corticospinal excitability.


Short- and long-interval ICI (SICI and LICI, respectively) were assessed prior to the intervention to obtain baseline ICI. SICI was measured using a paired-pulse paradigm consisting of a subthreshold conditioning stimulus followed 3 ms later by a suprathreshold test stimulus (Kujirai et al., 1993). Three conditioning stimulus intensities of 70%, 80% and 90% AMT were used, while the test stimulus was set at MEP1 mV. Ten trials of each conditioned state and 10 test-alone trials were recorded for each subject, resulting in a single block of 40 trials. LICI was measured in a separate paired-pulse paradigm consisting of suprathreshold conditioning and test stimuli (Valls-Sole et al., 1992). Two different interstimulus intervals (ISIs) of 100 and 150 ms were used. Both test and conditioning stimuli were set at MEP1 mV. Ten trials of each state (test-alone, 100 and 150 ms ISI) were applied in a randomised manner, resulting in a single block of 30 trials. For both SICI and LICI, ICI was quantified by obtaining the MEP amplitude from each individual trial, averaging the MEPs from each state, then expressing the average conditioned response as a percentage of the unconditioned response. During offline analysis, all frames were inspected for EMG activity prior to the stimulus. Trials containing activity were excluded from further analysis. Baseline SICI and LICI data were obtained by dividing each individual conditioned MEP response by the average unconditioned test-alone MEP response for each subject, in each state. Normalised values were then compared between groups.

Neuroplasticity intervention

cTBS was applied to the left primary motor cortex using a Magstim Super Rapid magnetic stimulator (Magstim, Dyfed, UK) while the subject was relaxed. This stimulation paradigm was originally described by Huang and colleagues (Huang et al., 2005), and consists of a basic unit of three stimuli applied at 50 Hz, which is then repeated at 5 Hz for 40 s (resulting in a total of 600 stimuli) at an intensity of 80% AMT. In healthy control subjects, this paradigm results in suppression of MEP amplitude that can last up to 60 min (Huang et al., 2005). Neuroplastic changes in the motor cortex induced by cTBS were assessed by recording 15 MEPs at 10, 20 and 30 min post-intervention at the MEP1 mV TMS intensity.


Differences in age, handedness, physical activity, physical measurements [height, weight, body mass index (BMI)], baseline RMT and AMT, MEP1 mV, AHI, sleep efficiency and sleep respiratory data were compared between groups (patients with OSA, controls) using unpaired Student's t-tests.

Sleep architecture was compared using a two-factor repeated-measures analysis of variance (anovaRM) with a between-subject factor of group (OSA, control) and within-subject factor of sleep stage [rapid eye movement (REM) sleep, non-REM (NREM) Stages 1 and 2, and slow-wave sleep (SWS; comprised of NREM Stages 3 and 4)]. Significant main effects and interactions were further investigated using one-factor anova with Bonferroni correction for multiple contrasts. Mixed-model analysis was used to examine the fixed effects of group and time (post 10, post 20 and post 30) on the response of subjects to cTBS. Subject was included as a random effect, and data were fitted with an autoregressive (AR1) covariance structure (PASW software, version 18.0; SPSS, Chicago, IL, USA). Mixed-model analysis was also used to compare differences in SICI and LICI between groups, assessing fixed effects of subject group and conditioning intensity on SICI (70%, 80% and 90% AMT), and subject group and ISI on LICI (100 and 150 ms). Subject was again included as a random effect, and data were fitted with a diagonal covariance structure. Significant interactions were further investigated using Bonferroni corrected custom contrasts. To further investigate relationships between OSA and corticomotor excitability, linear regression of individual subject data was used to relate indices of disease severity (AHI, ESS, O2-saturation) to baseline TMS measurements (RMT and MEP1 mV). Linear regression was also used to investigate relationships between subject characteristics and responses to cTBS. Contrasted variables included measures of baseline cortical excitability and ICI, physical activity (work, sport, leisure), anthropometric (weight, BMI and age) and polysomnography data (AHI, AI, sleep efficiency, respiratory data and sleep stage). Statistical significance was set at ≤ 0.05 for all comparisons. Data are shown as mean ± SEM in figures, and mean ± SD in tables and text.


Two control subjects showed evidence of OSA on diagnostic testing (AHI = 15.8 and 20.1 events/h) and were excluded from any further analysis. One patient with OSA was unable to complete the TMS session due to a high TMS threshold that resulted in discomfort caused by facial muscle activation. Subsequently, all data from this subject were excluded from the analysis. One control subject showed a marked increase in MEPs after cTBS, with MEP amplitudes at all time points more than three SDs away from the group mean. Given an outlier response, with clear facilitation of MEPs in contrast to the well-characterised reduction in MEP response of healthy subjects to cTBS (Huang, 2010), this subject was excluded from further analysis. Therefore, all analyses were performed on a total subject cohort of 13 patients with OSA and 11 control subjects.

Subject characteristics

Table 1 shows baseline data for 13 patients with OSA and 11 healthy controls before rTMS. There were no significant differences between groups in age, height or handedness, but patients were 29% heavier and had a 26% greater BMI than controls. Subjective daytime sleepiness (as measured by the ESS) was also significantly higher in patients than controls. Assessment of physical activity showed no significant differences between groups for the index of work activity, but controls showed a 22% higher activity index during leisure time and a 31% higher index of sporting activity than patients.

Table 1. Subject demographic, baseline TMS and polysomnographic characteristics
  1. Data are presented as mean ± SD. AHI, apnoea–hypopnoea index; AI, arousal index; AMT, active motor threshold; BMI, body mass index; ESS, Epworth Sleepiness Scale; LQ, laterality quotient; MEP1 mV, stimulus intensity producing an MEP amplitude of approximately 1 mV; NREM, non-rapid eye movement sleep; OSA, obstructive sleep apnoea; REM, rapid eye movement sleep; RMT, resting motor threshold; SWS, slow-wave sleep.

Anthropometric data
Age (years ± SD)42.6 ± 10.243.0 ± 10.30.93
n (male : female)11 : 29 : 2 
Height (cm)176.9 ± 6.8172.7 ± 8.30.18
Weight (kg)110.0 ± 21.078.4 ± 11.6< 0.001
BMI (kg/m2)35.3 ± 7.826.2 ± 3.30.002
Handedness (LQ)0.91 ± 0.10.88 ± 0.10.52
ESS9.2 ± 3.02.9 ± 1.7< 0.001
Physical activity
Work index2.5 ± 0.72.4 ± 0.40.61
Leisure index2.5 ± 0.83.2 ± 0.70.03
Sport index2.0 ± 0.72.9 ± 0.90.01
Baseline cortical excitability data
RMT (% maximal stimulator output)53.0 ± 11.944.4 ± 7.40.05
AMT (% maximal stimulator output)41.9 ± 8.835.7 ± 6.20.06
MEP1 mV (% maximal stimulator output)63.0 ± 13.052.7 ± 10.60.05
Polysomnographic data
AHI (events/h)45.5 ± 19.32.9 ± 2.0< 0.001
AI (events/h)
Total27.3 ± 16.011.5 ± 5.10.005
Respiratory17.0 ± 15.90.9 ± 1.00.003
Limb movement4.1 ± 2.85.0 ± 5.00.60
Spontaneous6.3 ± 6.35.6 ± 3.00.77
Sleep efficiency (%)69.8 ± 15.678.6 ± 7.20.10
NREM (%)
Mean91.2 ± 3.994.5 ± 1.00.01
Minimum84.2 ± 4.190.1 ± 3.0< 0.001
REM (%)
Mean90.2 ± 6.694.5 ± 1.40.04
Minimum82.5 ± 10.991.0 ± 2.80.02
< 90% (% sleep time)7.3 ± 9.60.2 ± 0.60.02
Sleep structure (% sleep time)
NREM 120.8 ± 14.36.8 ± 3.9< 0.001
NREM 251.2 ± 13.755.3 ± 10.90.28
SWS15.6 ± 9.720.8 ± 9.70.18
REM12.4 ± 10.417.2 ± 4.20.21

Polysomnography results

Patients with OSA showed severe OSA (i.e. AHI > 30 events/h), with significantly higher AHI and significantly lower average and minimum O2-saturation during both NREM and REM sleep (Table 1). Patients also demonstrated a significantly higher proportion of sleep time spent with O2-saturation below 90%, and significantly elevated total and respiratory-related AIs. Although sleep efficiency was not significantly different between groups, there was a significant main effect of sleep stage (F3,22 = 58.27, < 0.001), and a significant sleep stage × group interaction effect (F3,66 = 3.58, P = 0.02) in percent time within each sleep stage. A subsequent one-way anova showed that patients with OSA spent significantly more time in NREM Stage 1 than controls. There were no other significant group differences in other sleep stages (Table 1).

Baseline cortical excitability

RMT and the TMS intensity producing MEP1 mV were both significantly higher in patients, whereas AMT just failed to reach statistical significance between groups (Table 1). Figure 1A and B shows the average responses for SICI and LICI compared between each group in each stimulus condition. A significant main effect of conditioning intensity was found for SICI, with higher intensity conditioning stimuli resulting in increased inhibition in FDI (F2,314 = 23.27, < 0.001). However, there was no difference between groups (F1,23 = 0.98, P = 0.33) or group × conditioning intensity interaction effect (F2,314 = 0.31, P = 0.74). A significant main effect of ISI was also found for LICI, with increased inhibition at the shorter ISI (F1,236 = 36.51, < 0.001). This analysis also showed no difference between groups (F1,27 = 0.56, P = 0.46) and no group × ISI interaction (F1,236 = 0.32, P = 0.57).

Figure 1.

ICI in 13 patients with obstructive sleep apnoea (OSA) and 11 control subjects at baseline. (A) SICI in patients with OSA (white circles) and control subjects (black triangles) for three conditioning stimulus intensities. There were no differences between groups (= 0.3). (B) LICI in patients with OSA (white bars) and healthy control subjects (black bars) at two interstimulus intervals (ISIs). There were no differences between groups (= 0.4). Data represent the average motor-evoked potential (MEP) amplitude produced by a conditioned stimulus expressed as a percentage of average MEP amplitude produced by an unconditioned stimulus, with no inhibition representing 100%. AMT, active motor threshold; FDI, first dorsal interosseous.

Assessment of neuroplasticity

An example of mean MEPs obtained before and after rTMS is shown for one patient with OSA and one control subject in Fig. 2A. Representative subjects are matched for age (control, 51 years; patient, 49 years), height (control, 175 cm; patient, 173 cm) and weight (control, 91 kg; patient, 85 kg), whereas patient AHI was 22.4 events/h compared with the control value of 4.3 events/h. In this example, the control subject showed response amplitudes of 50%, 40% and 90% of baseline at 10, 20 and 30 min post-intervention, respectively. In contrast, the OSA patient response showed MEP amplitudes of 124%, 152% and 159% of baseline at 10, 20 and 30 min post-intervention, respectively.

Figure 2.

Single subjects and mean motor-evoked potentials (MEPs) from all subjects following cTBS. (A) Representative data from one control subject and one patient with obstructive sleep apnoea (OSA) showing changes in mean MEP amplitude 10 min (post 10), 20 min (post 20) and 30 min (post 30) after rTMS. Patient and control subjects had similar physical characteristics (see text). Arrows represent TMS pulse. (B) Test MEP amplitude at 10, 20 and 30 min after rTMS in 13 patients with OSA (white circles) and 11 healthy control subjects (black triangles). The average MEP amplitude at each time point is expressed as a percentage of average pre-intervention MEP amplitude. *P ≤ 0.05 between groups at the 20 min time point, and between the 10 and 30 min time point in the control group. FDI, first dorsal interosseous.

Group data from 13 patients with OSA and 11 control subjects are shown in Fig. 2B. When normalised to before cTBS, the MEP amplitude showed a significant main effect of time (F2,315 = 5.49, P = 0.005) and a significant group × time interaction (F2,315 = 3.93, P = 0.02), although there was no main effect of group (F1,22 = 1.78, P = 0.20). Subsequent post hoc tests showed that the MEP amplitude in control subjects at the 10-min time point was significantly lower than at the 30-min time point (P = 0.002). Furthermore, there was a significant difference in MEP amplitude between the patients with OSA and control subjects 20 min after cTBS (P = 0.05). Inclusion of the one control subject identified as an outlier in the preliminary analysis (13 patients with OSA and 12 control subjects) did not alter the main findings, with a significant main effect of time (F2,323 = 4.96, P = 0.008) and a significant group × time interaction (F2,323 = 4.71, P = 0.01), indicating that the main outcomes were not sensitive to exclusion of this subject.

Associations between OSA, corticomotor excitability and neuroplasticity

Regression plots for comparisons between AHI, ESS, RMT and MEP1 mV are shown in Fig. 3. For all subjects, AHI demonstrated significant positive relationships with both RMT (r2 = 0.19, P = 0.03) and MEP1 mV (r2 = 0.22, P = 0.02). ESS also demonstrated similar significant relationships with these measurements (RMT: r2 = 0.19, P = 0.03; MEP1 mV: r2 = 0.19, P = 0.03). Furthermore, minimum O2-saturation during NREM sleep showed significant negative relationships to RMT (r2 = 0.20, P = 0.03) and MEP1 mV (r2 = 0.23, P = 0.02; data not shown).

Figure 3.

Regression plots showing the relationships between indices of OSA severity and corticomotor excitability. Unfilled circles represent control subjects, filled circles represent patients with OSA. Epworth Sleepiness Scale (ESS) measurements explained 24% of the variation in resting motor threshold (RMT; A) and 19% of the variation in motor-evoked potential (MEP)1 mV (B), while the apnoea–hypopnoea index (AHI) explained 21% of the variation in RMT (C) and 22% of the variation in MEP1 mV (D).

Leisure time activity showed a significant relationship with the change in MEP amplitude at 10 min (r2 = 0.19, P = 0.03) and 20 min (r2 = 0.29, P = 0.006) post-intervention, with a trend towards a relationship at 30 min post-intervention (P = 0.06). The magnitude of inhibition measured during LICI with a 150-ms ISI also showed a trend towards a relationship at 30 min post-intervention (P = 0.06). No further relationships approached statistical significance.


This study is the first to use TMS to investigate neuroplasticity in patients with OSA. The main findings were that patients with moderate-to-severe OSA show an abnormal response to cTBS, indicating altered motor cortex plasticity. Furthermore, differences in ICI are unlikely to contribute to this effect. The abnormal response to cTBS suggests that changes in cortical plasticity may be a consequence of OSA pathophysiology.

In the present study, excitability of cortical areas innervating a hand muscle was used as an index of global alterations in brain function in patients with OSA, as hand muscles have strong corticospinal projections to motor neurons and are easily activated by TMS (Petersen et al., 2003). We considered this a reasonable strategy to target the neurophysiological effects of this respiratory condition, because sleep fragmentation and chronic hypoxia associated with OSA could have widespread effects on corticospinal fibre integrity (Macey et al., 2008) not specifically restricted to brain areas controlling upper airway muscles. Evidence for the non-specific effects of OSA on brain function include widespread changes in grey matter (Joo et al., 2010b; Morrell et al., 2010; Torelli et al., 2011) and deficits in cognitive function (Campana et al., 2010). Furthermore, assessing the neural control of hand muscles has been a common strategy in other conditions that produce cognitive effects, such as Alzheimer's disease (Liepert et al., 2001; Battaglia et al., 2007), mild traumatic brain injury (De Beaumont et al., 2012) and autism spectrum disorders (Oberman et al., 2010).

Previous TMS studies demonstrating abnormal corticospinal excitability to hand muscles (Civardi et al., 2004; Grippo et al., 2005; Joo et al., 2010a) also demonstrate the non-specific effects of OSA on brain function. The observation in this study of increased RMT and MEP1 mV intensities in OSA supports these previous findings, and most likely reflects a structural change in intracortical networks that are activated by TMS (Rothwell et al., 1991), or cellular factors that contribute to a reduced membrane excitability of cortical neurons (Ziemann et al., 1996b; Chen et al., 1997). However, in contrast to previous studies (Joo et al., 2010a), we were able to show significant linear relationships between indices of OSA severity (AHI and ESS), RMT and MEP1 mV intensities. These relationships explained 20–25% of the variation between subjects and support cortical hypoexcitability in patients with severe OSA. Although the mechanism underlying these changes remains unclear, significant relationships between minimum O2-saturation during NREM sleep, RMT and MEP1 mV suggest that recurrent overnight hypoxaemia may play a role. Although these associations were relatively weak, the TMS measurements were not performed on the same day as the overnight polysomnography for logistical reasons, so this may have reduced the strength of correlations between sleep architecture and TMS measurements.

Neuroplasticity and OSA

cTBS was used to induce plasticity in the present study as it has several advantages over other plasticity-inducing protocols. First, it uses a subthreshold TMS intensity that does not produce a MEP, so the effects are likely to be mediated at a cortical level (Di Lazzaro et al., 2005). Second, the cTBS paradigm is short (40 s), thereby minimising effects of attention or drowsiness on the plasticity response that can be observed in longer protocols (Stefan et al., 2004), an important consideration in patients with OSA. Third, there is a well-characterised decrease in MEP amplitude in healthy subjects, with maximum reductions of about 40% occurring between 15 and 30 min post-intervention and a return to pre-intervention levels within 30–60 min (Huang et al., 2005). Fourth, there is strong evidence that the suppression of MEP amplitudes reflects LTD-like changes occurring in the motor cortex (Huang et al., 2007). These findings suggest that cTBS represents an effective tool to examine plasticity at the systems level of the human motor cortex, and has important implications for understanding the neurophysiological consequences of OSA.

As individuals with OSA are known to have cognitive deficits (Campana et al., 2010), and hippocampal long-term potentiation (LTP) is impaired in a mouse model of OSA (Xie et al., 2010), we expected that the capacity for neuroplastic modulation would be decreased in patients with OSA. In healthy control subjects there was a suppression of MEP amplitudes following cTBS, consistent with that reported by other groups (Huang et al., 2005). However, the response in patients with OSA was markedly different, with no suppression of MEPs occurring after cTBS. Furthermore, differences in MEP amplitudes between patients and controls were most evident 20 min after the intervention. These findings were largely independent of differences in sleep architecture between patients with OSA and controls, with no significant correlations between time spent in each sleep stage and post-cTBS MEP response, although patients with OSA showed significantly more time spent in NREM Stage 1 than controls.

Previous studies have suggested altered brain function in OSA as a result of chronic intermittent hypoxia (Xie et al., 2010) and hypercapnia (Grippo et al., 2005). The present study showed no significant correlations between AHI or reductions in arterial blood O2-saturation (i.e. desaturation) and post-intervention changes in MEP amplitude, arguing against a significant role of disrupted oxygenation in mediating this response. Furthermore, carbon dioxide changes during sleep were not measured in the present study. As differences in carbon dioxide levels have previously been implicated in altered cortical excitability (Grippo et al., 2005), the role of overnight hypercapnia on neuroplasticity in OSA may warrant future investigation.

It is well known that sleep is important for memory consolidation and brain plasticity (Walker & Stickgold, 2006; Diekelmann et al., 2009), and increasing evidence suggests that SWS (NREM Stages 3 and 4) is associated with synaptic plasticity and learning (Huber et al., 2004; De Gennaro et al., 2008). However, there was no difference in the proportion of time spent in SWS between patients with OSA and controls in this study, although there was a tendency for a reduced proportion of time spent in NREM Stage 3 in patients with OSA. Furthermore, the possibility exists that the impaired plasticity in patients with OSA is due to sleep fragmentation. Animal studies have shown that sleep fragmentation impairs hippocampal LTP (Tartar et al., 2006), and studies in healthy subjects have shown that sleep disruption can influence motor consolidation (Walker et al., 2002, 2005), which is thought to involve sleep-related changes in cortical connectivity and plasticity (Maquet et al., 2003). However, it is not clear whether the effect of acute sleep disruption in healthy subjects is equivalent to the chronic sleep fragmentation that is typically seen in patients with OSA. Nonetheless, a recent study has shown that reduced motor consolidation in patients with mild OSA was associated with increased arousals during sleep rather than the total amount of time spent sleeping, sleep efficiency or sleep architecture (Djonlagic et al., 2012). This, combined with our findings of an increased AI in patients with OSA, suggests that a lack of sleep continuity may contribute to impaired cortical plasticity in patients with OSA.

Although application of cTBS produces important new information about the neurophysiological consequences of OSA, these results represent an investigation into LTD-like effects only. The lack of LTD-like synaptic plasticity in OSA could represent an overall reduction in cellular mechanisms of synaptic plasticity, or a shift in the threshold for induction of LTP-like plasticity in accordance with the rules of metaplasticity (Abraham, 2008). However, this latter possibility seems unlikely, as it would contradict findings in animal models of OSA pathology (Xie et al., 2010). Future studies will need to further investigate this prospect by applying intermittent TBS, or other brain stimulation paradigms thought to induce LTP-like plasticity.

Finally, due to its cross-sectional design, it is possible that several confounding factors may have contributed to the results observed in our assessment of plasticity. Many factors are known to influence the response to rTMS (Ridding & Ziemann, 2010). Some of these, such as time of day, age and gender, were well matched between subject groups in the present study. Significant positive correlations between post-intervention MEPs at the 10 and 20 min time point and indices of physical activity during leisure time suggest that reduced physical activity may have contributed to the response of patients with OSA. This is consistent with a previous study using paired-associative stimulation, which demonstrated reduced neuroplastic modulation in sedentary compared with highly active individuals (Cirillo et al., 2009). However, the strength of associations observed in the present study were relatively weak, suggesting that the extent of physical activity is unlikely to play a large role in the impaired neuroplasticity in patients with OSA. Genetic factors are also known to influence plasticity (Missitzi et al., 2011), for example, a common polymorphism of the brain-derived neurotrophic factor (BDNF) gene can influence the response to rTMS (Cheeran et al., 2008). The prevalence of this BDNF polymorphism may have been different between subject groups. Furthermore, the impact of OSA sequelae, such as comorbidities, altered cortisol levels and increased daytime sleepiness, could play a role. The contribution of these confounding factors to the impaired response to rTMS in patients with OSA remains to be determined.


Inhibitory neurons using γ-aminobutyric acid (GABA) as their transmitter constitute 25–30% of neurons in the primate neocortex (Jones, 1993), and play an important role in the reorganisation of neural connections that underlie motor learning and recovery from injury (Sanes & Donoghue, 2000). We used paired-pulse TMS to examine these GABAergic inhibitory systems in patients with OSA. SICI is thought to be mediated by GABAA receptors (Ziemann et al., 1996a,b), whereas LICI is likely to involve GABAB receptors (Werhahn et al., 1999). SICI and LICI have been shown to be abnormal in some neurological conditions (Berardelli et al., 2008), and we wanted to determine whether these measures of ICI were influenced by OSA. We found no difference in SICI or LICI in patients with OSA compared with controls, suggesting that ICI is not responsible for the observed reduction in plasticity response following cTBS.

Only one previous study has compared SICI between patients with OSA and healthy control subjects, showing no difference between groups (Joo et al., 2010a). However, only a single conditioning TMS intensity of 80% RMT (equivalent to ~100% AMT in our study) was used, which may be influenced by intracortical facilitatory circuits (Ortu et al., 2008). In the present study, we used three different conditioning TMS intensities (70%, 80% and 90% AMT), which allowed us to compare the recruitment of inhibitory interneurons between groups, and included a conditioning intensity of 70% AMT, which is unlikely to be influenced by intracortical facilitation (Ortu et al., 2008). Although our assessment of SICI failed to show significant differences at any of the three conditioning TMS intensities, the largest difference between groups was observed at 70% AMT. This result warrants further investigation of SICI in patients with OSA, potentially by optimising the assessment of SICI by altering the TMS current direction to preferentially generate late indirect waves in the descending corticospinal volley, which are known to be more sensitive to SICI (Zoghi et al., 2003).

Perhaps the most robust change in motor cortex function in patients with OSA is a prolonged CSP (Civardi et al., 2004; Grippo et al., 2005; Joo et al., 2010a). This measurement applies a single TMS pulse to the cortex while the target muscle is voluntarily activated and is seen as a suppression of EMG activity directly after the MEP. At intervals > 50 ms, EMG suppression is thought to represent GABAB-mediated inhibition that is cortical in origin (Siebner et al., 1998). To extend these findings, the current study assessed LICI as an alternative measure of GABAB-mediated ICI in patients with OSA. In contrast to previous studies using the CSP (Civardi et al., 2004; Grippo et al., 2005; Joo et al., 2010a), we found no difference in LICI between patients with OSA and controls. This lack of effect may reflect methodological differences between the assessment of LICI and CSP. For example, assessment of the CSP requires voluntary activation of the muscle, whereas LICI was assessed at rest, suggesting there may be differences in LICI due to muscle activation under some conditions (Clark et al., 2008; McGinley et al., 2010). Further clarification of cortical inhibition in patients with OSA would require assessment of LICI in an active target muscle, as well as additional paradigms measuring GABAB cortical inhibition, such as interhemispheric inhibition.

In conclusion, we used cTBS to show that cortical plasticity was reduced in patients with OSA, possibly due to altered sleep fragmentation or chronic hypoxia/hypercapnia. We showed no difference in SICI or LICI in patients with OSA compared with controls, suggesting that altered ICI was not responsible for the reduced response to cTBS in these patients. These differences in plasticity within the motor system may contribute to impairments in motor learning and consolidation that have been observed in patients with OSA (Djonlagic et al., 2012), and reflect more global changes in neuroplasticity that may contribute to known cognitive deficits in patients with OSA (Campana et al., 2010). Whether impaired neuroplasticity in OSA can be restored with common treatments for the disorder (e.g. CPAP) remains to be determined.


We gratefully acknowledge the Adelaide Institute for Sleep Health clinical and laboratory personnel for their support in conducting sleep studies. These studies were performed with support from the NHMRC (project grant 480438) and Adelaide Centre for Neuroscience Research. M.C.R. holds a Senior Research Fellowship from the National Health and Medical Research Council of Australia. The authors have no conflicts of interest to declare.


apnoea–hypopnoea index


arousal index


active motor threshold


brain-derived neurotrophic factor


body mass index


continuous positive airway pressure


cortical silent period


continuous theta burst stimulation








Epworth Sleepiness Scale


first dorsal interosseous


γ-aminobutyric acid


intracortical inhibition


interstimulus interval


long-interval intracortical inhibition


long-term depression


long-term potentiation


motor-evoked potential


stimulator intensity producing an MEP 1 mV in peak-to-peak amplitude


non-rapid eye movement


obstructive sleep apnoea


rapid eye movement


resting motor threshold


repetitive transcranial magnetic stimulation


short-interval intracortical inhibition


slow-wave sleep


transcranial magnetic stimulation