2.2.1. Populations.

Studies were eligible for inclusion in this review if they included an experimental group consisting of adults (age ≥ 18 years old) who experienced either experimental pain (tonic (long lasting) or phasic (brief noxious stimuli)) or clinical pain. Clinical pain was considered as any symptom of pain included in the IASP definition for pain [27] excluding those occurring in the presence of neurological disease or due to delayed onset of muscle soreness.

2.2.2. Intervention.

Motor skill learning interventions were selected using the definitions presented in Table 1. This encompassed the most common motor paradigms used in research including both, implicit and explicit: motor sequence learning, visuomotor learning, repeated ballistic movements, motor adaptation and ecological learning. Prism adaptation paradigms were excluded from this review in an attempt to reduce confounding variables, such as the impact of visual perception.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Classification of motor learning.

https://doi.org/10.1371/journal.pone.0274403.t001

2.2.3. Comparator.

A comparison group consisting of adults (age≥ 18 years old) with no pain was required for inclusion. All subjects were required to complete the same motor skill learning to compare learning in the presence of pain versus learning in the absence of pain.

2.2.4. Outcomes.

Primary outcome measures of interest in this review included task performance measures related to motor learning, or activity-dependent plasticity measures (See protocol for more details). Measures of task performance included both spatial and temporal measures. Activity dependent plasticity measures included, but not limited to, changes in amplitude, temporal or spatial characteristic obtained using magnetic resonance imaging (MRI), transcranial magnetic stimulation (TMS) or electroencephalogram (EEG).

2.2.5. Study design.

Randomised control studies (RCTs) and non-randomized studies were included. Single case studies, case series and review papers along with any studies not published in English were excluded. Also excluded were any study including treatments as an adjunct to motor learning, in an attempt to minimise heterogeneity of interventions subjects were exposed to and consequently minimise confounding variables that may impact on the amount of learning observed. Study duration was not limited.

3.4.1. Motor sequence learning.

Three studies [30–32] explored the impact of experimental pain on motor sequence learning. Implicit and explicit learning paradigms were explored across the studies. All three studies used a tonic experimental pain paradigm (two capsaicin and one thermal).

3.4.1.1. Task performance measure. Three outcome measures, accuracy, speed and reaction times, were employed by these studies. Accuracy, measured by number of errors, was classified as high risk of bias, and therefore excluded. No impact of experimental pain on improvements in speed or reaction time during early stage/acquisition was reported. Two of the studies [31, 32] explored the impact of experimental pain during acquisition on a one-off performance of the motor sequence learning task 24hrs or 48hrs later. No impact of experimental pain was observed. The GRADE assessment for both outcome measures was judged to be low, suggesting limited confidence that experimental pain does not impact on the speed or reaction time of a motor sequence learning task at acquisition or retention.

3.4.1.2. Activity dependent plasticity measures. Two studies [30, 32] explored the impact of experimental pain on measures of neuroplasticity during motor sequence learning. Both explored the impact of capsaicin on SEPs using EEG. Only SEPs associated with changes across learning are discussed here. For detailed information on all SEPs examined in these studies please see Table 3. Both studies reported no impact of experimental pain on N24 and N18 SEPs associated with the cerebellum. Dancey et al. [30] reported a significantly greater increase in N30 SEP, associated with somatosensory integration and the motor cortex, across learning in the presence of experimental pain. Whereas, Dancey et al. [32] showed a significantly greater increase in N20 SEP, associated with the primary somatosensory cortex, following motor learning in the absence of experimental pain. GRADE scores for all SEP outcome measures in these studies were considered ‘low’, recommending limited confidence in the above impact on SEPs associated with the cerebellum, somatosensory integration and motor and primary somatosensory cortex.

3.4.2. Visuomotor learning.

Five studies explored the impact of experimental pain on visuomotor learning. All five studies [10, 16, 33–35] used a tonic pain paradigm and explicit learning paradigms. One of the five experimental pain paradigms used intramuscular hypertonic saline injections [35] whereas the others used cutaneous capsaicin application.

3.4.2.1. Task performance measure. The impact of experimental pain was explored using four outcome measures: accuracy, number of errors, movement time and an accuracy speed trade off measure. Four out of the five studies [16, 33–35] demonstrated no impact of experimental pain on changes in accuracy following early-stage learning, although three studies [16, 33, 34] reported overall accuracy in the pain groups was higher at all data collection points. One study [10] reported experimental pain resulted in an impairment of learning across a single training session. This study utilised a tongue protrusion task whereas all the others utilised a task involving the upper limb. Experimental pain did not impact on any other outcome measure regardless of the method of application of the pain. Two studies [16, 33] explored the impact of experimental pain during acquisition on a one-off performance of the visuomotor learning task 24hrs to 48hrs later. Dancey et al. [33] reported a significant decrease in accuracy in the pain group. In contrast, Dancey et al. [16] reported no differences in accuracy between groups. GRADE scores for all outcome measures were considered low. Therefore, based on the five studies there is limited confidence in the conflicting evidence of the impact of experimental pain on accuracy measures, and the lack of impact of experimental pain on temporal measures during a visuomotor learning task.

3.4.2.2. Activity dependent plasticity measures. Four studies [10, 16, 33, 34] explored the impact of experimental pain on activity dependent plasticity measures. One study investigated SEPs, [33] three studies [16, 33, 34] collected data on single pulse TMS measures, and one study explored paired pulse TMS measure. The single study exploring SEPs reported a significant difference in evoked potentials associated with the cerebellum (N18 and N24) between groups. A decrease in N18 SEP in the pain group compared to the control group and a significant decrease in N24 SEP in the control group compared to the pain group was reported. Three studies reported a significant difference in single-pulse measures between groups. One study [10] reported a decrease in cortical excitability in the control group but not the pain group across training. In contrast, two studies reported a significant increase in the control group, halfway through training [34] and immediately post training [16]. Mavromatis et al. [34] reported the increase in excitability observed halfway through training, in the control group, returned to baseline by the end of training where it was not significantly different from the pain group. The same research group reported no significant difference in short intracortical inhibition following training. GRADE scores were low for all activity dependent plasticity outcome measures implying low confidence in the single study’s findings on the impact of experimental pain on SEPs associated with the cerebellum and the conflicting results of three studies on the impact of experimental pain on single pulse TMS measures.

3.4.3. Repeated ballistic movements.

One study [13] investigated the impact of experimental pain on learning using explicit repeated ballistic movement training of the upper limb digits. Ingham et al. [13] employed hypertonic saline injections to induce pain.

3.4.3.1. Task performance measure. The experimental pain study by Ingham et al. [13] reported no impact of either local or remote pain on improvement in finger acceleration during early-stage learning. GRADE scores for this outcome measure were considered low and therefore there is limited confidence in the absence of impact of experimental pain on learning employing repeated ballistic movement tasks.

3.4.3.2. Activity dependent plasticity measures. Ingham et al. [13] investigated two single pulse TMS activity dependent measures, TMS evoked peak finger acceleration and flexor digitorum indicis MEP. No significant difference was observed between local pain and control groups for either measure. Only TMS evoked finger acceleration improved across training. In contrast, when experimental pain was applied remotely (patella fat pad) there was no change in TMS evoked finger acceleration following repeated ballistic finger training. Low GRADE scores suggest limited confidence in the absence of effect of pain on MEPs, and differing effect of local pain and remote pain on TMS finger induced acceleration.

3.4.4. Motor adaptation.

Motor adaptation paradigms employed by the five studies exploring the impact of experimental pain on learning included: reaching [36, 37] and locomotion tasks [38–40]. All studies required participants to adapt to the application of an external force. Three studies [36, 38, 39] used capsaicin applied to cutaneous areas of relevant limb and two studies [37, 40] injected hypertonic saline into the muscle.

3.4.4.1. Task performance measure. Outcome measures assessed included movement error and timing of errors. All five studies reported no impact of experimental pain on movement errors during early stages of learning. One study [39] reported when pain was present during initial acquisition there was an impairment in subsequent learning during a subsequent training session (retention) 24hrs later. Four follow up studies [36–38, 40] from the same research group demonstrated no impact on retention. The same four studies [36–38, 40] reported a significant difference in movement strategies across learning. Bouffard et al. [38, 40] reported a difference in timing of errors suggesting reduced anticipatory strategies. This was supported by Salomoni et al. [37] who reported reduced muscle activity in the presence of experimental pain during first exposure to the force field. In contrast, Lamothe et al. [36] reported larger feedforward preparatory movements in the pain group. GRADE scores for all outcome measures were classified as low. Overall, there is limited confidence in the absence of impact of experimental pain on task performance measures during acquisition and retention or altered movement strategies in the presence of experimental pain during motor adaptation learning.

3.4.4.2. Activity dependent plasticity measures. No studies have explored the impact on experimental pain on activity dependent plasticity measures during motor adaptation learning tasks.

3.4.5. Ecological learning.

A single study [41] utilised an ecological learning paradigm, dart throwing, to explore the impact of experimental pain on accuracy and movement variability.

3.4.5.1. Task performance measure. This study reported no impact of experimental pain on improvements in accuracy of performance, but increased movement variability at upper limb joints during the performance. Due to low sample size, GRADE score were considered low, resulting in limited confidence in findings that experimental pain does not impact on the accuracy of performance of ecological learning task.

3.4.5.2. Activity dependent plasticity measures. No studies have explored the impact of experimental pain on activity dependent plasticity measures during ecological learning tasks.

3.5.2. Motor adaptation.

A single study [45] explored the impact of pain from isolated ankle injuries on a motor adaptation locomotive task.

3.5.2.1. Task performance measure. Outcome measures assessed included movement error and timing of errors. This study found no impact of clinical pain on movement error or timing of errors at either acquisition or retention. Similar to findings for experimental pain, clinical pain was found to alter movement strategies associated with anticipation of movement during the motor learning. A reduction in TA muscle activity prior to the perturbation was observed, across the two days in the no pain group but not the pain group. GRADE scores for all outcome measures were classified as low. Overall, there is limited confidence in the absence of impact of clinical pain on task performance measures during acquisition and retention or altered movement strategies in the presence of clinical pain during motor adaptation learning.

3.5.2.2. Activity dependent plasticity measures. No studies have explored the impact on clinical pain on activity dependent plasticity measures during motor adaptation learning tasks.

4.1.1. Pain interference with task performance measures.

All fifteen studies used a tonic experimental pain paradigm. Nine out of 10 task performance measures explored during acquisition, across all five motor learning paradigms, demonstrated consistent results. No impact of pain on motor learning was found for all measures except for ‘timing of errors’ measured during motor adaptation. A smaller value for ‘timing of errors’ is associated with the participant switching to a more anticipatory strategy whereas a larger value suggests a more reactive strategy [38, 40]. The two studies exploring ‘timing of errors’ in motor adaptation reported that the presence of pain significantly reduced the transition to smaller values compared to the control group. This continued use of reactive strategies is supported by findings that EMG activity was lower in the pain group compared to the control group on first exposure to a force field [37]. This finding is consistent with previous reports of pain delaying preparatory muscle activity during one-off tasks [48–50] but also extends the research to show an inability for participants to adapt this strategy despite practice.

Studies exploring the impact of pain on accuracy measures during acquisition of a visuomotor task have reported contrasting results. One study demonstrated a reduction in motor learning across acquisition in the presence of pain compared to a control group [10]. In contrast, the other three studies [16, 33, 35] demonstrated no difference between groups across training, but two studies [16, 33] showed a significantly greater accuracy in the pain group at all points during the training.

Boudreau et al. [10] applied pain to the tongue during a tongue protrusion task compared to elbow pain during a motor tracing task involving the hand/wrist [16, 33]. It could therefore be argued that the interference of pain may vary depending on the location of pain. Rohel et al. [19] reported provisional evidence of a reduction in cortical excitability in response to hand and face pain but no reduction in forearm pain. The authors suggest that this may be due to greater sensory acuity and fine motor control of the hand and face associated with larger sensory and motor cortical representations.

A further difference between the studies is the type of task utilised. Complex or precision tasks such as the tongue protrusion task used by Boudreau et al. [10] may depend more on somatosensory feedback than tracing tasks. Pain has been shown to cause non-noxious sensory disturbances [51] and impaired processing of sensory information such as proprioception especially during complex tasks [52]. Researchers have theorised that somatosensory feedback plays a vital role in skill acquisition. Evidence from animal studies have demonstrated impaired learning of novel movements in the presence of lesions of the primary somatosensory cortex [53]. Similar findings have been demonstrated in humans using a repetitive TMS virtual somatosensory cortex lesion model [54]. In addition, peripheral manipulation of sensory information using lidocaine has been shown to decrease overall performance, increase errors and reduce reaction times during the same tongue protrusion task used by Boudreau and colleagues [14]. Both Vidoni et al. [54] and Boudreau et al. [14] utilised complex tasks requiring both visual information and graded movements during performance. Caution should be taken when comparing research using central and peripheral somatosensory disruptions, as mechanisms underlying central and peripheral sensorimotor interactions may differ [54].

An important consideration when interpreting the results from Boudreau et al. [10] is that application of the capsaicin depended on the participants engagement in the task. No data was collected on how the task was completed and therefore the impairment in learning could have resulted from reduced engagement in the task in attempt to reduce the painful stimuli. A challenge of future research in this area is to establish whether evidence of pain interference reflects pain attenuation of learning or pain disruption in performance during the learning task leading to reduced motor skill learning.

The findings by Dancey et al. [16, 33] that the pain group was more accurate during a visuomotor tracing task of the wrist and hand at all time points compared to the control groups, suggests an interesting interaction between pain and task performance. It is possible that a pain stimulus delivered to the limb performing the motor task during motor learning increased attention to the limb, facilitating performance. Previous research, utilising other sensory inputs such as non-noxious electrical stimuli to the limb completing the learning task, reported increased attention facilitates neuroplasticity [55], measured using TMS. Whereas impairment in behavioural performance following motor learning occurs when instructions are given to intentionally divert attention away [56]. The latter is consistent with findings by Ingham et al. [13]. The authors reported decreased training induced changes in excitability related to hand motor learning in the presence of experimentally induced knee pain.

Attentional bias observed in people with chronic pain [57] may play a role in ‘learning in pain’, whereby, whether an individual attends to, or away from pain during motor learning, may determine whether improvements in performance are facilitated or impaired. The interaction between pain, attention and motor learning interference may also be influenced by cognitive constructs, such as fear and motivation [11]. The balance of ‘motivation to protect’ versus ‘motivation to complete the task’ may influence the resultant interference. An important question for future research design is ‘what is the value of the pain experience to the participant?’ and ‘what factors influence the participants assessment of value, such as fear and catastrophizing?’.

Task performance measures recorded more than 24hrs after acquisition demonstrated conflicting results across studies. Definitions and approaches to measuring retention varies across the studies. Dancey et al. [32, 33] retested task performance more than 24hrs post acquisition, focusing mainly on improvements in the absence of further training (offline learning) [6]. In contrast, Bouffard et al. [38–40] Salomoni et al. [37] and Lamothe et al. [36] observed the changes in task performance across a further training session (online learning–long term retention) [58]. Physiological processes underlying the above two processes may vary and therefore comparison of findings is limited.

4.1.2. Pain interference with activity-dependent plasticity measures.

There is a lack of consensus between studies on the impact of experimental tonic pain on activity dependent plasticity measures of motor learning, such as TMS single pulse measures and SEP peaks. Paired pulse TMS measures were limited to single studies and therefore any conclusions would be premature. No studies have explored activity dependent plasticity using functional MRI. The findings in this review are consistent with a systematic review by Stanisic et al. [23] who reported acute experimental pain may influence training induced neuroplasticity defined by TMS. The authors suggested the impact of pain on training induced neuroplasticity may be dependent on the motor learning paradigms utilised. The observed variation in the findings in the present review could be in part due to the respective tasks.

Paparella et al. [59] demonstrated selectivity of cortical, corticospinal and intra-cortical changes in response to different learning tasks. Ingham et al. [13] employed a simple ballistic thumb movement task compared to the precision tasks employed in the other two studies. It could be argued that the tasks used in these studies were of contrasting difficulty, and therefore with the addition of pain, the cognitive load may differ [52], which may influence the change in TMS acquired corticomotor excitability [34].

The demands of the task may explain some of the differences seen in the effects of pain on SEPs. Dancey et al. [32, 33] reported a significant increase in N20 SEP in the control group compared to the pain group. In contrast, Dancey et al. [30] reported no changes in the control group or between the two groups. This study from this research group utilised a pure motor typing task. SEPs reflect precognitive sensory processing and are markers of somatosensory integration. N20 SEP represents the earliest cortical processing in the somatosensory area [32]. The findings discussed above may suggest a role for the somatosensory area in processing of relevant sensory information associated with learning of a complex but not a simple motor task, and that this processing can be disrupted by sensory stimuli such as experimental pain. Further evidence of the importance of adapting to sensory stimuli was noted by Lamothe et al. [36] during motor adaptation task. They suggested the pain group utilised strategies to minimise the need for online adjustments in task performance.