Strategies for Rapid Muscle Fatigue Reduction during FES Exercise in Individuals with Spinal Cord Injury: A Systematic Review

Background Rapid muscle fatigue during functional electrical stimulation (FES)-evoked muscle contractions in individuals with spinal cord injury (SCI) is a significant limitation to attaining health benefits of FES-exercise. Delaying the onset of muscle fatigue is often cited as an important goal linked to FES clinical efficacy. Although the basic concept of fatigue-resistance has a long history, recent advances in biomedical engineering, physiotherapy and clinical exercise science have achieved improved clinical benefits, especially for reducing muscle fatigue during FES-exercise. This review evaluated the methodological quality of strategies underlying muscle fatigue-resistance that have been used to optimize FES therapeutic approaches. The review also sought to synthesize the effectiveness of these strategies for persons with SCI in order to establish their functional impacts and clinical relevance. Methods Published scientific literature pertaining to the reduction of FES-induced muscle fatigue was identified through searches of the following databases: Science Direct, Medline, IEEE Xplore, SpringerLink, PubMed and Nature, from the earliest returned record until June 2015. Titles and abstracts were screened to obtain 35 studies that met the inclusion criteria for this systematic review. Results Following the evaluation of methodological quality (mean (SD), 50 (6) %) of the reviewed studies using the Downs and Black scale, the largest treatment effects reported to reduce muscle fatigue mainly investigated isometric contractions of limited functional and clinical relevance (n = 28). Some investigations (n = 13) lacked randomisation, while others were characterised by small sample sizes with low statistical power. Nevertheless, the clinical significance of emerging trends to improve fatigue-resistance during FES included (i) optimizing electrode positioning, (ii) fine-tuning of stimulation patterns and other FES parameters, (iii) adjustments to the mode and frequency of exercise training, and (iv) biofeedback-assisted FES-exercise to promote selective recruitment of fatigue-resistant motor units. Conclusion Although the need for further in-depth clinical trials (especially RCTs) was clearly warranted to establish external validity of outcomes, current evidence was sufficient to support the validity of certain techniques for rapid fatigue-reduction in order to promote FES therapy as an integral part of SCI rehabilitation. It is anticipated that this information will be valuable to clinicians and other allied health professionals administering FES as a treatment option in rehabilitation and aid the development of effective rehabilitation interventions.


Introduction
Muscle atrophy, and the consequent alteration of the proportion of slow-twitch type I motor units (MU) to fast-fatigable type IIB (with low aerobic-oxidative enzymatic capacity) MU, are negative neuromuscular sequelae secondary to spinal cord injury (SCI) [1,2]. As a consequence of these morphological and histochemical adaptations, resistance to rapid fatigue is impaired in denervated skeletal muscles compromised by upper-motor neuron lesions in the spinal cord [3]. Accordingly, the power output and exercise capacity of such muscles are diminished due to inactivity and unloading concomitant with post-SCI wheelchair confinement [4]. This is clearly evident in the decline of force-generating capacity of muscle (i.e. specific tension (NÁcm -2 )) [5], due to the problem of muscle fatigue in individuals with neurological disorders [6]. Current efforts to offset the issue have limited impact, due partly to the limitation of pathophysiologic understanding of muscle fatigue and its complications in this population [6,7]. Consequently, there is on-going research interest into developing effective strategies to counteract the effects of rapid muscle fatigue, particularly during neuromuscular electrical stimulation for therapeutic and functional interventions in persons with neurological impairment, specifically SCI [8].
Neuromuscular electrical stimulation (NMES) applied over the human neuromusculature produces muscle contractions by depolarizing motor axons beneath the stimulating electrodes [9]. The generated muscle contractions result in therapeutic and/or functional gains by exploiting the adaptive potential of skeletal muscles fibres to increase the loading effect on joints. When evoking functional or performance gains, NMES has often been characterized as "functional electrical stimulation" (FES) [9]. FES has been applied to maintain, improve or restore muscle trophism, improve health and augment functional outcomes after SCI [9], in postacute care, rehabilitation settings and exercise [10][11][12]. The usefulness of FES therapy to promote the 'restoration' of purposeful function has been demonstrated in several studies [11,13,14]. Yet, the inherent non-physiological response of paretic or paralyzed muscles (due to changes in their histological composition [2] and the reversal of usual MU recruitment order [15]) to electrical stimulation often leads to non-optimal recruitment of fast fatigable muscle fibres over fatigue-resistant fibres [16,17]. Similarly, majority of evidence supports that the ordering of FES-induced MU recruitment is non-selective [15,18], the consequent of which is the exaggerated metabolic cost of an electrically-evoked contractions that lead to rapid muscle fatigue [9,18]. This limits the duration of functional tasks that FES may evoke [19]. Therefore, the need for practical solutions to the "rapid fatigue problem" is of paramount importance if FES therapy is to become more widespread in deployment for patients' rehabilitation.
The low 'take-up' of FES interventions within conventional clinical practice, should not be assumed to indicate any lack of medical benefits, since scientific understanding of electrical stimulus-induced fatigue in the clinical population is still rudimentary [6,20]. Previous studies, that have sought to improve fatigue-resistance during FES therapy, have investigated the effects of size of stimulating electrodes or their position over some specific locations (such as anatomical landmarks and motor points) [16], modulation of neuromuscular stimulation parameters [21], optimization of the mode and frequency of exercise [22] and biofeedback-controlled stimulation [23]. Although these prior studies were undertaken to improve the effectiveness of FES rehabilitation within clinical populations, none has consistently improved the fatigue resistance characteristics of paralyzed muscles [15]. Moreover, these techniques have not been fully incorporated into the routine clinical practice. Rather, the techniques have been more investigated for optimizing FES-assisted muscle contractions in able-bodied (AB) compared to SCI populations [21]. However, evidence supporting the successful transfer of the FES techniques from AB to clinical population remains poorly documented.
The current review sought to synthesize knowledge about effectiveness of fatigue-reduction strategies for persons with SCI in order to establish their functional impact and clinical relevance. Earlier, there have been some historical reviews on the implications and management of muscle fatigue in clinical populations, such as Binder-Macleod and Snyder-Mackler [24] and Maffiuletti et al., [25]. Apart from the populations considered by the authors of those study (i. e., AB and persons with SCI), only the manipulation of neuromuscular stimulus parameters was considered to improve FES contractions [24]. Subsequently, changes in the skeletal muscle characteristics following SCI have been highlighted in relation to the effects of fatigue resistance during FES therapy [26,27]. Recently, Maffiuletti and colleagues [25] suggested a new strategy for FES deployment (i.e., multi current pathway FES). Although the authors' paradigm produced more forceful muscle contractions, the rapid-fatigue cycle did not show any significant improvements compared with traditional FES strategies. Clearly, in the clinical environment, more widely accepted and 'FES-efficient' strategies need to be devised, and a comprehensive insight into an effective management of rapid muscle fatigue is needed.

Search strategy
Initially, 1933 articles within the electronic databases of Science Direct, Medline, IEEE Xplore, SpringerLink, PubMed, and Nature, and 2 additional articles identified through other sources (online request) were obtained from the earliest returned record until June 2015. The relevant search terms included 'spinal cord injury', 'paralysis', 'paraplegia', 'tetraplegia' with 'muscle fatigue', 'reduction', 'delay', 'functional electrical stimulation', 'electrical stimulation', therapy', 'contractions', 'walking', 'standing' and 'cycling'. The synonyms for the terms (e.g., 'gait' or 'stepping' for walking) were also included on the list of search terms. Further, we conducted a free search in Google Scholar using the reference listed in the primary citations in order to accommodate a wider context. Electrode databases were searched online through the University of Malaya, Malaysia library. However, the literature search was restricted to the English language only.

Eligibility criteria
The eligibility criteria were set to accommodate the focus of the review which sought to evaluate the methodological quality as well as synthesize the effectiveness of muscle fatigue-reduction strategies during FES exercise for functional and therapeutic benefits in persons with SCI. The titles and abstracts of all identified studies were screened to determine eligibility. In the event that the title or abstract did not provide adequate information, such article was retrieved for full review. In a second round of filtering, the full text of all potentially relevant studies were reviewed against the following inclusion criteria by two reviewers (MOI and NAH): (i) the study participants were humans with SCI; (ii) the study objective involved the delayed onset of muscle fatigue i.e., assessed the ability of muscle to sustain a significant longer duration of repeated contractions following the administration of a specific strategy; (iii) at least one outcome quantified the improvement/or otherwise of resistance of rapid muscle fatigue in terms of endurance, force output (or its derivatives), number of contractions and contractile speed; (iv) study outcomes were measured based on a well-defined muscle fatigue reducing strategy. Studies with unclear protocol or data presentation were excluded. Four major categories of strategies were clearly evident i.e. optimization of electrode positioning, fine-tuning of stimulation patterns and other FES parameters, adjustments to the mode and frequency of exercise training, and biofeedback-assisted FES-exercise.

Data extraction
In order to reduce the risk of bias, two reviewers (MOI and NAH) were independently involved in extraction of data for all outcomes derived from the four identified fatigue resistance strategies. Other reviewers (AKA and NH) verified the validity of the data before a final compilation, while two others (NH and GMD) checked the technical soundness, clarity and the flow of the study. Information on participants' physical characteristics, study objectives, methodologies, interventions, outcomes/results and limitations, and future research suggestions were retained.

Assessment of study quality
The quality of the studies was measured using the Downs and Black (D&B) Scale [28] relevant in assessing both randomised and non-randomised studies and adjudged a valid scale for methodological quality assessments [29]. The quality of each study was independently assessed by two reviewers using the D&B Scale [28]. A third "judge" was invited when the two reviewers disagree to settle the discrepancies. In line with other previously published reviews [30,31] some criteria on the D&B scale (i.e., 17, 22 and 27) were considered inapplicable to the objective of the current review and, therefore, not graded. Thereafter, we noted significant variations between the levels of spinal lesion of the participants. In general, subjects with complete and incomplete lesions, as well as those with acute and chronic injuries were grouped together. A meta-analysis could not be conducted due to significant variations in the methods used to evaluate the outcomes, subject heterogeneity and diverse experimental methodologies. The maximum score on the D&B scale was normalised to 100% as some criteria were relevant to some studies and irrelevant to others.

Included studies
At the outset, 1935 citations were generated and 1900 failed to fit the inclusion criteria. Thirtyfive studies met the inclusion criteria and were critically analysed for this review (Fig 1). Out of these studies, eighteen assessed the modification of neuromuscular stimulation patterns, five investigated the optimization of electrode position, ten evaluated the optimization of exercise interventions and two appraised the 'quality' of exercise when augmented by biofeedback-controlled FES muscle contractions. Outcome measures were assessed based on the type and mode of muscle contractions (i.e., isometric versus non-isometric-isotonic/isokinetic). The characteristics of the included studies are presented in Tables 1-4.

Methodological quality
The quality of the included studies was moderate i.e. meeting up to an average of 50% of the criteria (range: 38-70%, Table 5) on the D&B scale. None of the studies reported the following criteria; adverse events that may be consequence of the intervention (criterion 8), concealed allocation and blinded participants, therapists and assessors (criterion 14 and 15), adjustment to the effects of confounders in the data analysis (criterion 25). Among controlled studies (studies with treatment and controlled groups [32]) only that of Downey et al., 2014 [33] partially described the distribution of principal confounders which may account for differences between the treatment and control groups. The criteria commonly unfulfilled were; the reporting of exact P values (criterion 10) and randomisation of study participants to the intervention groups (criterion 23). However, all studies specified the eligibility criteria, and measured key outcomes in up to 100% of the initial participants except Godfrey et al., 2002 [34] and Scott et al., 2005 [14] which derived the outcomes from 50% and 67% of the recruited participants, respectively.

Study characteristics
Out of the 35 included studies, thirty-one were peer-reviewed journal articles, three were case studies or pilot investigations, and one conference paper. There were no RCTs in the scope of interest, and most studies lack control conditions. Subject heterogeneity and diverse methodology partially limited a quantitative or meta-analytical comparison between studies, and amongst strategies. Most of the studies were published from 2005 to 2014 highlighting an increased interest recently in this field.

Modification of stimulation patterns
Functional electrical stimulation to evoke muscle contractions is typically effected through the modification of key stimulus parameters (i.e., pulse width, frequency, amplitude) [35], the effectiveness of which could be graded by their ability to offset rapid muscle fatigue. The effectiveness of optimal activation strategies comprising stimulation patterns and modulation of different stimulation trains for reducing muscle fatigue in persons with SCI (Table 1) were investigated in eighteen studies.  Contraction: Isometric. Objective: (i) to investigate if nonlinear summation due to closely spaced stimulation pulses applies to human skeletal muscle; (ii) to identify the N-let pulse parameter that will produce the most force per pulse; (iii) to verify the hypothesis that optimal N-let trains will enable significantly longer FES muscle contractions. Study design: The quadriceps muscle was stimulated for 1 hour each day and exercised on a FES-aided bicycle 3 times a week. The optimal pulse interval was calculated by parameters of N-let pulse trains for number of pulses (N) = 1-6 in order to obtain the most force per pulse. The duration of torque tracking was used to compare the different modes of stimulation. The distal and proximal electrodes were placed over the metacarpophalangeal joint and the base of the thumb, respectively. A ground wire was placed along the wrist crease. The thumb was extended and strapped against a force transducer that registered both abduction and flexion forces at right angles to each other. Resultant force was calculated off-line. Stimulation pattern: Doublet pattern (i.e., a brief inter-pulse interval) and twitch contractions. PW: 50 μs was used to determine the best site for stimulation. Frequency: 50 Hz.
In paralyzed and control muscles, the maximal force and force-time integral were produced when brief (5-15 ms) interpulse interval was followed by longer intervals. The increments in nPT at 15 and 20 min with the DT protocol (i.e., 23% and 29%) were greater than the increments at 5 and 10 min (i.e., 10% and 12%) P <0.05. Therefore, the DT protocol, as compared with the CT protocol, was more effective during lowfrequency fatigue as time after fatigue increased. Graham  In the non-fatigued and fatigued conditions, the VFT and DFT peak forces were greater than the CFT peak forces at 10 Hz (P = 0.001). In the fatigued condition the 20 Hz VFT peak forces were greater than the CFT peak forces (P = 0.05), and there was a trend for the DFT peak forces to be greater than the CFT peak forces (P = 0.07) but no significant effect of train-type on the FTi(s). (Continued) Fatigue Reduction during FES Exercise   Configuration: HPF-Pals Platinum square-(7 × 10 cm) as the anode and cathode. LPF-Four oval cathodes (4 × 6 cm) and one square anode (7 × 10 cm). Anatomical position: HPF-Cathode electrode at the top and the anode at the distal part of the quadriceps, LPF-Anode at the distal while the four cathodes were distributed over the quadriceps such that no current pathway between the cathode and anode intersected.
The fatigue interval was extended more than 150% in the LPF (multipleelectrode) configuration as compared with HPF (two-electrode). Configuration: SDSS-Four separate (4.5 × 2.5 cm) were used as active electrode and a single (9 × 5 cm) was used as the reference. SES-Two separate (9 × 5 cm) were used as active and reference electrodes. Anatomical position: Active electrode-Located on the calf, near the gastrocnemius muscle at the distal part of the quadriceps Reference-Located above the achilles tendon.
SDSS produced 234% higher torques (compared with SES) at the end of the 2 min stimulation relative to the maximum torque with an improvement of 171% torque time integral (P<0.001). (Continued)

Fatigue Reduction during FES Exercise
Karu and colleagues [35], originally demonstrated that a strategy of N-let pulse trains (i.e., a set of closely spaced doublet stimulation pulses) showed a significantly longer torque time integral (TTi) when compared to the traditional single-pulsed stimulation. This stimulation pattern enabled the optimization of force production per stimulation pulse. However, the small sample size of this study and a marked between-subject variability led to the heterogeneity of responses amongst individuals. Two studies, one by Griffin et al., [36] with matched sample sizes and the other by Chang and Shields [37] with wider sample size demonstrated that the doublets pulse pattern was more effective for producing and sustaining force in paralyzed thenar and soleus muscles, respectively. Another four studies considered the effects of modulation of different pulse trains on the force augmentation. Switching frequency trains (i.e., from a constant frequency trains (CFT, where six 200 s square wave pulses separated by 70 ms were used) to a double frequency train (DFT, where a train of pairs of closely spaced pulses (doublets, 5ms) separated by longer intervals were used)) demonstrated improvements in fatigue Configuration: SDSS-Four separate (3 of (4.5 × 2.5 cm) and 1 of (2.5 × 2.5 cm; for dorsiflexors)) were used as active electrode. SES-A pair of (9 × 5cm) were used as active and reference electrodes except for dorsiflexors on which (5 × 5cm) electrodes were used.    [14], whereas administration of CFT and DFT separately were less optimal [21]. In addition, during paralyzed quadriceps' strength training, Deley et al., [20] verified that variable frequency train (VFT, where a train identical to the CFT were used, except that the first 2 pulses were separated by 5 ms) stimulation pattern and VFT followed by CFT generated less fatiguing contractions. Taken together, these findings suggest that the frequency modification of different stimulation trains offsets rapid muscle fatigue.   Table 5. In contrast, Thomas et al., [38] evaluated the fatigue-reducing ability of variable versus constant frequency pulse trains in paralyzed thenar muscles and observed that both patterns did not significantly reduce muscle fatigue of functional relevance unless complemented with other fatigue reducing strategies. Bickel and co-workers [39] reported that the use of VFT may not be efficacious to enhance torque during fatiguing contractions. The authors identified that the fast contraction rise time that was evident by inability of the inter-pulse interval to augment the peak torque might be responsible. They observed that the mechanism of fatigue and its management may be dissimilar in able-bodied and SCI populations. However, pre-conditioning of the SCI muscles was recommended for future applications of pulse train modulation [39]. Recently, Gorgey et al. [40], examined the effect of modulating the stimulation pulse width (while the frequency value remained constant (CFT)) on the fatigability of paralyzed muscle following cycling exercise. The investigators reported that modulating pulse width could not influence the muscle fatigability significantly. Based on the available evidence [21,40], there is no significant effect of only CFT on the reduction of rapid fatigue in paralyzed muscle. Thus, the traditional method of keeping the stimulation frequency constant might not be an appropriate strategy to reduce muscle fatigability.
During repetitive electrical stimulation, progressive increases of stimulation frequency and intensity (PIFI) has been shown to generate extended contractions [41]. Kebaetse and colleagues [12] in their study, started with a low frequency and switched to a higher frequency stimulation. They were able to produce a repetitive, non-isometric contractions in persons with SCI by minimizing the rate of fatigue during a low-frequency stimulation phase. Eser et al., [42] and del-Ama and co-workers [43] observed that frequency modulation could promote extended contractions during cycling and isometric exercise and should be further investigated as a strategy to regulate power output and subvert muscle fatigue. In addition to using isometric contraction to assess fatigue characteristics (i.e., muscle force and work output) following FES cycling [44], isometric contraction exercise promotes prolonged contraction necessary to augment power output and extend cycling time.
A case study by Graupe and co-workers [45], demonstrated an improvement of FES standing and walking by random stimulation of inter-pulse interval (as against the constant frequency stimulation) in a range of ±12% between FES pulses. The random stimulation of interpulse interval was meant to excite action potential in the peripheral nerves resulting in nerve membrane depolarization to produce extended muscle contractions [45]. Although the study could not be generalised to a wider population due to a single case study design, it served as a platform for further systematic investigations to extend the standing and walking duration in a larger test group.
In contrast, Graham et al., [46] and Thrasher et al., [8] independently evaluated four modes of stimulus parameter modulation and observed that the random modulation of stimulus parameters may not be a viable strategy for fatigue resistance during FES-induced contractions. Although the authors suggested that the 10 min rest between trials might be insufficient for full restoration of muscle strength, subjects' heterogeneity might also have been responsible for their observations. Based on these evidence, the modulation of FES parameters including pulse width, frequency, and amplitude has therefore, been identified as an essential strategy used to mimic the natural MU recruitment patterns and promotes reduced fatigability [41].
A study by Godfrey and colleagues, [34] advocated for submaximal stimulation to delay muscle fatigue. However, the study was marked with inconsistences in the level of muscle fibre recruited across the subject to the extent that the investigators had to adopt the data of the half of the subjects for the analysis (i.e., those that were near to each other). This limitation may be responsible for the lack of popularity of this strategy.
A study by Decker and colleagues [47] investigated the validity of altering stimulation strategy over the traditional 'co-activation' strategy in delaying muscle fatigue during cycling. Authors identified that the alternation of stimulation among muscles offers an enhanced benefit to FES therapy because the strategy distributes activation among a set of synergistic muscles and delay fatigue by allowing individual muscle to rest during the deactivation periods [47].
Apart from a preliminary study by Graupe and colleagues [45] which reported a delay of rapid fatigue during standing and walking, that by Decker and co-workers [47], Eser et al., [42], and Gorgey et al., [40] which investigated the possibility of rapid fatigue offset during FES cycling, other investigations have only considered either improvement in the muscle force/torque or endurance during FES-evoked isometric contractions. Interestingly, there is evidence of similar cardiorespiratory responses following both isometric and cycling contractions [48]. Thus, other than promoting reduced fatigability, intermittent isometric contraction might be an equally viable alternative as cycling for improving the whole body metabolic rate during FES exercise in persons with SCI [48].
Based on the available evidence, the effectiveness of modification of stimulation pattern to prevent rapid muscle fatigue is typically assessed through comparison of the fatigue index (i.e., final torque normalized to maximum torque), fatigue time (i.e., time for torque to drop by 3 dB) [46], and torque-time integral (i.e., over the entire trial). The administration of the optimal parameters of stimulation appears to be essential for the sustenance of FES-induced functional activity. In summary, the potential for minimizing rapid muscle fatigue during FES-evoked functional activities through the modulation of neuromuscular stimulus parameters is quite evident, and this remains an important research domain for optimization of FES therapy [40,43].

Optimization of electrode positioning
Five studies investigated the importance of electrode position to reduce the fatigue characteristics of FES-induced muscle contractions in SCI population (Table 2). Popovic and Malesevic [49] demonstrated an increased mean fatigue interval by 153% in a multi-electrode array as compared with a single cathode electrode. Malesevic and colleagues [50] confirmed that a low pulse frequency (four smaller cathodes at 16 Hz each) was better suited for a prolonged stimulation with a stronger, and fused tetanic contractions when compared to a high pulse frequency (single cathode at 30 Hz), although at the expense of force production [50]. Equally, the spatially distributed sequential stimulation configuration has demonstrated a greater fatigue reducing ability compared to a single active electrode stimulation configuration in the upper limbs [51] and lower limbs [52]. This is because different sets of muscle fibres are activated alternatively by different electrodes, enabling activation of sub-compartment muscles [52]. Similarly, Downey and co-workers [33] demonstrated a significant reduction of FES-induced fatigue with an asynchronous stimulation protocol. Although stimulation electrode placement over the peripheral nerve trunk and motor points are frequently cited locations, a general consideration for electrode placement appears to be a function of the intended purposeful activity. Moreover, available evidence have identified the possibility of significant inter-individual variability in the location of motor points and that an overlap of the electrode must be avoided [51].
In general, the proper placement of stimulation electrodes over motor points of the target muscle is necessary for optimal muscle fibre recruitment. Additionally, the use of multi-electrode cathode has been recommended over a single cathode electrode. Clear understanding of the anatomical landmarks where sufficient motor units maybe recruited has been identified as an important precursor for selective recruitment of fatigue-resistant motor units. Thus, the application of multi-electrode stimulation arrays has not only been applied to selectively recruit the motor unit for efficient motor control, but has equally been validated to be more effective to offset rapid muscle fatigue.

Optimization of exercise training
While some previous studies have sought to improve fatigue-resistance via technical or technological approaches to neuromuscular stimulation, such as modulating neuromuscular pulse trains or deploying multi-electrode arrays, less research has addressed whether better fatigueresistance might be achieved by improved muscle training paradigms for patients. Rehabilitation of persons with SCI involves continuous training and exercise in order to achieve the desired outcomes via task specific training [22]. Ten studies assessed the effect of exercise on the fatigue characteristics during FES-induced contractions in SCI population. A study by Peckham and colleagues [53] investigated whether FES-induced contractions of finger flexors of persons with SCI could elicit usable force and endurance necessary for functional hand grasp. They reported an increase of force development and fatigue resistance following FES training oriented on functional activity. Although there were significant variations in the participants' responses in terms of force production and fatigability, the study generally demonstrated the possibility of modification of contractile properties of paralyzed human skeletal muscles through FES exercise [53]. Two independent studies by Shields and Dudley-Javoroski [22] and Shields and co-workers [54] showed an increase in endurance time and fatigue resistance following two years of isometric contractions of the ankle plantar flexors [22] and prevention of post fatigue potentiation in soleus muscles [54] respectively. Authors further suggested that electrical stimulation may be used to predict contractile characteristics in paralyzed muscles. Another study by Gerrits et al., 2002 [55] compared the effects of leg isomeric training using two patterns of FES (i.e., repetitive high-frequency stimulation (HFS) and more continuous low-frequency stimulation (LFS)) on the strength, contractile properties, and fatigability of paralyzed quadriceps muscle. Authors showed that fatigue resistance increased significantly (P<0.05) from 2 weeks of LFS training (by 43%) but not (P>0.05) with HFS training even after 12 weeks of training.
Butler et al., [56] evaluated the effect of blood pressure control during exercise and found that improved blood pressure in thenar muscles was moderately correlated to fatigue resistance. However, investigators mentioned certain difficulties and safety associated with raising the mean arterial pressure in persons with SCI as the major concern with this strategy. Similarly, Hartkopp and colleagues, [57] compared the fatigability of trained wrist extensors to high frequency stimulation (30 Hz) and low frequency stimulation (15 Hz). Although the two protocol significantly offset the rapid fatigue, only high frequency protocol considerably augmented muscle force and aerobic metabolism after training [57]. Thus, strategies that promote improved aerobic metabolism could lengthen contraction time during FES exercise.
Owing to the difficulty of translating the neurological changes into functional outcomes, rehabilitation intervention in persons with SCI has been focused extensively on functional training. Improved adaptable response and fatigue resistance ability of paralyzed muscles to functional training have been demonstrated in FES induced cycling. Gerrits et al., [44] evaluated the contractile speed and fatigability of paralyzed quadriceps muscle to the FES leg cycle ergometry (FES-LCE) exercise. Increased fatigue resistance was reported as indicated by the higher force maintained. However, authors reported large variability in fatigue resistance between participants that was also negatively correlated with the time since injury (TSI) of the participants [44]. This suggests the need for a consideration of the injury levels and TSI while selecting FES patterns for therapy. Similarly, Fornusek and Davis [58] investigated the effect of pedalling cadence upon torque production and muscle fatigue during FES cycling. Although the investigators identified significantly better delay of leg muscle fatigue during low cadence pedalling, high cadence produced better power output [58]. The authors expressed this as a trade-off in a muscle's force-velocity profile-an electrically stimulated muscle group may produce a higher power output at a higher cadence but with rapid fatigue, or lower power output at a slower cadence with reduced fatigability. The need to further investigate the relative contribution of stimulation period and leg angular velocity to different fatigue rates was also suggested.
Two independent studies [59,60] reported significant effects of FES resistance training (RT) on paralyzed muscle fatigability. Specifically, Gorgey et al. [59], in a case study, showed the possibility of once a week FES RT to offset rapid quadriceps muscle fatigue. Although being a single case report, the result could not be generalized, but a proof of concept on a simple method to sustain regional muscle adaptations to training could be inferred. Similarly, Sabatier and colleagues [60] reported 60% reduction in quadriceps muscle fatigue following 12 weeks of FES RT in five persons with SCI. These authors proposed that FES resistance training could increase muscle size and promote reduced denervated muscle fatigability.
In the interim, FES exercise therapy in persons with SCI is of considerable clinical value, especially as it increases contraction time and fatigue resistance due to the high percentage of fatigue-resistance fibres associated with the loading effect following exercise [22]. FES exercise has also been shown to promote plasticity, hypertrophy and endurance due to the considerable neural adaptation [22,27]. Although it has sometimes been claimed that FES exercise training transcends just fatigue resistance but also may promote neuroplasticity, there are few studies that have substantiated this in the SCI population (Table 3). There may be differences between fatigue responses in chronic and acute SCIs due to the variations in the duration of inactivity associated with the changes in muscle metabolism, blood flow, and fibre composition [53]. The marked differences between the exercise responses in chronic and acute SCI seemed to have been overlooked in all the studies under this category. In summary, timely and appropriate FES-evoked exercise training in persons with SCI has been frequently cited to significantly impact fatigue resistance [22]. Future research must focus on which components of the FESexercise prescription (intensity, duration or weekly frequency) are most 'dose-potent' for functional and health outcomes to patients.

Biofeedback-controlled stimulation
The effectiveness of recent application of biopotential-controlled FES exercise in reducing rapid fatigue during FES training was assessed by two studies [23,61]. Shields and colleagues [23] compared the effectiveness of online modulation (closed loop) and traditional modality (open loop) and found that online modulation of stimulation parameters (i.e., through feedback controller to regulate muscle torque) significantly enhanced the generated gastrocnemius and soleus muscles' mean and peak torques even when the pattern of stimulation was randomised [23]. Similarly, Dudley-Javoroski and co-workers, [61] reported significant improvement in the quadriceps peak force and fatigue index with the application of biofeedback controlled FES as compared to the traditional open loop FES. In summary, incorporation of biofeedback to control FES has been shown to impact fatigue resistance positively and enhanced physiologic performance of paralyzed muscles [61]. Integration of biofeedback into FES technique is promising and sensor technology for measuring biopotentials may revolutionize the scope of the FES technology if adequately substantiated in clinical populations.

Discussion
This systematic review has established that previous studies demonstrated only partial success for ameliorating rapid muscle fatigue during FES muscle contractions in persons with SCI. Muscle force and its derivatives were identified as the key indices of fatigue. The available evidence revealed a number of important strategies often cited to delay the onset of muscle fatigue including the optimization of electrode positioning, modification of patterns of stimulation and its parameters, optimization of the mode and frequency of exercise training, and biofeedback-controlled FES exercise. These modalities generally resulted in selective recruitment of fatigue resistant motor units [34,62]. The literature also supported the use of increase in endurance time during repetitive contractions as indication of muscle fatigue resistance in persons with SCI. However, limited evidence was available concerning the direct investigation of the delay of muscle fatigue during FES-assisted functional activities such as cycling [40,42,44,47,58], or standing and walking [45] in persons with SCI. Apparently, overcoming the limitations imposed by this gap of knowledge may be an essential element in promoting FES assisted activities of daily living.
The long-time goal of FES technology has been to restore functional tasks and form an integral part of SCI management. A major step in this direction is to identify fatigue resistance strategies in order to achieve an effective therapy. Modification of stimulation parameters has been mostly explored among the strategies identified. Out of eighteen studies that sought to titrate neuromuscular stimulus parameters, Graupe and colleagues [45] was a preliminary investigation, while Karu et al., [35] reported a significant within-subject variation. del-Ama and co-workers [43] clarified the limitation of their experimental findings only to the quadriceps isometric contractions during stance phase of walking. Inability of the SCI patients to develop adequate force required for swing phase was identified as a hindrance to isokinetic test in their approach [43]. Caution should be exercised in the future application of these methods due to their low methodological quality. Nevertheless, all the studies except two [8,46] reported modification of stimulus parameters as a significant fatigue-reducing strategy. The two studies with dissimilarity identified that their results might be due to the insufficient rest period between trials [8,46]. This suggests significant implication of adequate resting period in muscle fatigue assessments and reduction studies as well as the validity of modification of stimulation parameters, particularly stimulation frequency for rapid muscle fatigue reduction in FES-based rehabilitation interventions.
Peckham et al., [53] originally proposed exercise regimen induced by chronic electrical stimulation for fatigue resistance based on the inability of FES to develop sufficient force to sustain muscle contractions in paralyzed muscles. However, exercise interventions can only be effective if the electrical stimulation is robust and applied in such a way that it is substantially resistance to rapid muscle fatigue. In addition to other physiological benefits of FES exercise during rehabilitation, evidence suggests that exercise offsets rapid muscle fatigue and promote a better rehabilitation outcome in the clinical populations [11]. However, a large number of relevant studies [22,44,53,56,58] under this category failed to report any of the characteristics of participants lost to follow-up. Thus, there is a need to intensify efforts on home-based FES systems and the prescription of training 'dose' to facilitate self-care post SCI [22], since there may be significant time constraints for patients to attend clinics and gymnasiums.
Although the physical deconditioning secondary to SCI can be favourably altered by exercise interventions, some authors have argued that improvement of physical functionality does not entirely rely on exercise since the restoration of functional capacity depend on many other factors [63]. Among the factors that alter skeletal muscle morphology and function are severity of loss of voluntary motor control [63] and the level of injury [64]. It is well acknowledged that the ambulation and independence prognosis in the activities of daily living is primarily predicted by the level of injury and the completeness of the SCI [65]. Thus, the consequent variation in the motor and sensory functions in different SCI populations alters muscle contractile properties and significantly affect force-generating and fatigue-resistance capacity [65]. Therefore, investigation of the response of available strategies (with consideration to the effect of level and completeness of injury), on muscle fatigability needed to be understood in order to adequately position the clinical implication of the strategies.
Accordingly, an extensive muscle training to promote conditioning should be complemented with other regimens with consideration to specific level and completeness of injury in order to gain useful insight into key physiological functions necessary to improve muscle fatigue resistance. In all, reduction of rapid muscle fatigue during FES therapy in persons with SCI, being a complex task, has been managed using multiple, often concurrent, strategies. However, these are still mostly confined to the research laboratories based on the settings of most studies we reviewed. An integrated approach towards fatigue reduction may allow a successful deployment of the strategies for therapeutic trials in clinical settings and outdoor where the full potential of FES therapy can be derived. Currently, comparisons between the effectiveness of each of these strategies are difficult because of the lack of standardisation of the protocols, training procedures and stimulation parameters administered in different studies even within the same category of strategy. Accordingly, the immense potentials in the generation of fatigue resistant contractions bespeak the importance of randomised controlled trials (RCTs) to document the efficacy of available strategies. While further research is apparently required, preliminary data on the available strategies are promising. Although there is significant impact of each strategy which could serve as basis for further investigations, a combination of the strategies appears to be more effective.
Furthermore, it is axiomatic that the earlier the FES therapy commences following the injury the better will be the outcome, and the associated inactivity-atrophic changes in paralyzed muscles are reversible even after 20 years [10]. Apart from the study by Bickel et al., [39] and that of Chang and Shields [37], subjects with complete and incomplete lesions [49,50], as well as those with acute and chronic injuries [54] were grouped together during fatigue reduction studies. This might prevent the understanding of specific influence of fatigue reduction strategies on the time since injury-since skeletal muscles have different fatigue resistance capacity during post-SCI stages [51]. For example, although denervated muscles generally show disuse atrophy, in chronic SCI such atrophy is concomitant with an increase in the interstitial endomysial connective tissues and perifascicular fatty infiltration [66], attributable to muscle inactivity [67]. These profound morphological changes impact motor unit orientation and muscle fatigability significantly. Moreover, unlike during chronic SCI, an acutely denervated skeletal muscle might be characterized by an unusual muscle fibre composition (as indicated by the relative proportion of slow and fast myosin heavy chain isoform expression) [68], particularly during alteration of fibre type morphology and histochemistry after SCI [40,68]. These findings suggest different force-fatigue temporal responses (between acutely-denervated versus chronically-denervated muscles) to different fatigue reduction strategies. Therefore, integration of all the strategies in optimizing muscle performance should be further assessed in clinical populations with due consideration to "time since injury" in order to fully appreciate the functional benefits associated with the strategies.
Evidence indicates that most fatigue reduction investigations were conducted during isometric contractions of skeletal muscles. However, validation of those strategies during non-isometric conditions which are observed within functional tasks will be more clinically relevant. Equally, few muscle groups such as quadriceps and thenar have been evaluated, the test on the other functionally relevant muscle groups will determine how broadly the current strategies can be generalized. Therefore, further research trials are required before a definitive conclusion can be drawn on the effectiveness of FES fatigue reducing strategies in clinical population during functional activities. In all, the relationship among the strategies seems to be complementary [51], since there is not one without some limitations [8]. The lack of sufficient data to advance an external validity on a generalised mode of stimulation is conceivable as each FES candidate often present a unique clinical picture and different responses to a specific stimulation pattern. Lack of obvious methods of data extraction, different experimental settings and subject heterogeneity might have informed the considerable variation of the choice of fatigue measurement used in the studies reviewed.

Future directions
Despite an extensive evidence-based knowledge establishing the decreased fatigue resistance following SCI, little research has been performed to subvert this trend. Therefore, the prevalent incidence of SCI could not be matched with the studies in this field. Available efforts to automatically minimize this phenomenon remained limited, creating a wide knowledge gaps for the development of an ideal FES device. There may be device dependent limitations in FES therapy due to the manifestation of innate nature of the FES device and its effect on muscle physiology (i.e., unnatural recruitment patterns). The emerging trend is tending toward a systematic approach to the administration of the optimal stimulation parameters [37] and identification of exercise dose potency necessary to augment the outcomes [22]. These strategies may require sophisticated FES devices in terms of number of channels, signals controls and portability.
Another concern in muscle fatigue management beyond FES device is the interfacial relationship between the device and the muscle of interest. Important advancements in the area of rapid muscle fatigue reduction includes the stimulation at the motor point innervating the target muscles [69]. At the moment, the studies under this category [33,[49][50][51][52] have limited functional applications as they were conducted during isometric contractions. The electrophysiological procedure recommended by Gobbo and colleagues [69] on able-bodied should be further validated in SCI population during functional tasks. This is because the result of the investigations on able-bodied volunteers may not be simply or directly applied to SCI cohorts [21].
As identified from the literature, studies on the application of FES are marked with heterogeneity of stimulation patterns that may be responsible for the wide variation of outcomes. This problem may be partly ameliorated by an automated biofeedback-controlled FES which could be used to control the stimulation parameters to suit the fatigue state of different muscles. Development and validation of sensors to modulate and automate the stimulation pattern in FES muscle contractions in order to mimic the physiological coordination of muscular activities remain a wide knowledge gap. Indirect measures of neural activities including electromyogram (EMG) [70], electroencephalogram (EEG), electroneuragram (ENG) [71] and mechanomyogram (MMG) [47,72] (i.e., indices of neural information to decode functional intentions) have been validated as physiological sensing modalities. These sensors are promising in the design of biofeedback systems to achieve a 'closed looped' FES control in clinical populations [73]. They may be deployed to serve as proxy of muscle contractions (i.e., generated muscle force during fresh and fatiguing contractions), and the knee-joint dynamics [74] for the development of biopotential feedback controllers during FES-induced contractions.
There are indications of improved FES exercise outcome in SCI populations if combined with dietary supplementations. Nutrition such as whey protein and carbohydrate supplement [75], and other supplementations including creatine [76] have been proposed to augment functional responses (i.e., by boosting the lactate threshold during high intensity exercise) and promote endurance, time to fatigue and calorie expenditure during FES exercise. However, the available studies are marked with limitations including investigation on small case series with no control groups and were conducted without consideration to other circulatory dysfunction following SCI (i.e., which may have contributed to the fatigue patterns observed) [75]. Apart from these limitations, there is inadequate awareness on the precise mechanism of actions and side effects of these dietary intakes [77]. Nevertheless, there is an encouraging trend that these supplementations may enhance exercise capacity in SCI populations. Further investigations to quantify its short and long-time effects in promoting endurance, and reducing time to fatigue will be enlightening.

Study limitations
The results of this review are limited to studies written in English language and included in the databases used to identify the articles. Other studies may exist outside these domains which we were unable to retrieve. Selection bias is another potential limitation of systematic reviews. However, we attempted to minimize the bias by searching for both randomized and non-randomized studies and stating clearly the quality of each study. Incidentally, there was no randomized controlled trial in the scope of interest. This was assessed independently by two reviewers. There may be other limitations beyond our control such as publication bias. Frequently, only studies with positive results are submitted to peer reviewed journals for publication. This situation may bias results toward positive treatment effects. We endeavour to widen the search scope from the earliest time to the most recent, June 2015 to limit such effect. Nevertheless, an interesting trend in rapid fatigue management could still be objectively inferred.

Conclusion
The strategies for the rapid muscle fatigue reduction in SCI population have been highlighted in order to advance the clinical knowledge and practices, particularly in the management of persons with SCI. Inference from current state of evidence warrants the assessment of each method during functional task essential for activity of daily living as there is large treatment effect only on isometric contractions. We highlighted the emerging evidence on modulation of FES parameters, electrode positioning to enable asynchronous stimulation, and application of biopotential controlled FES exercising and muscle conditioning through exercise as the major fatigue reducing strategies. Despite the huge opportunities in the substantial independence leading to an improved quality of life for the FES candidates, the level of current evidence shows that the effective fatigue management in FES-induced contractions is still rudimentary. Most studies were conducted within short duration and were too modest, therefore, of limited clinical relevance. Many aspects of the studies were insufficiently researched to draw definitive conclusions. Moreover, several reviewed studies were marked with inadequate randomization, therefore caution should be exercised in interpreting authors' conclusions. However, the highlighted proof of concepts provide basis for more systematic studies in a wider test population. Although the recent improvement in the commonly used strategies and current interest are of importance, evaluation of the strategies while employing more vigorous methodological design, particularly, in non-isometric contraction settings will be more clinically relevant.