The current meta-analysis included only adults with SCI, regardless of traumatic or non-traumatic injury, the time since injury, and gender.

Functional electrical stimulation combined with active cycling training for lower limbs in the individuals with SCI was included. FES-cycling combined with other modality, such as resistance training, overground walking, and locomotor training, was excluded to decrease the heterogeneity and avoid confounding factors.

The primary outcome measure was spasticity assessed by MAS or AS for lower limbs. The secondary outcome measures were walking abilities, such as 6MWT, TUG, and LEMS. The 6MWT measured the walking distance in 6 min. TUG assessed the time that a subject took to rise from a chair, walk 3 m, turn around, walk back to the chair, and sit down. LEMS assessed motor score for lower limbs according to American Spinal Injury Association (ASIA) standard.

Because there were few randomized control trial (RCT) studies, both RCT and non-RCT studies were included in this meta-analysis.

The searching criteria were limited to the human studies and English language.

Five electronic databases [PubMed, Scopus, Medline (Proquest), Embase, and Cochrane Central Register of Controlled Trials (CENTRAL)] before September 2021 were searched. Medical Subject Heading terms were combined with keywords to search. Searching key words included cycling, functional electrical stimulation, FES, spinal cord injury and spasticity in title, abstract, and keywords. Supplementary Appendix 1 shows the combinations used. The hand searching from references lists for relevant studies was also done.

Two authors reviewed and screened the bibliographies of articles and abstracts related to the FES-cycling training in the individuals with SCI independently.

Two authors independently searched and screened the titles, abstracts, and literatures to identify the potentially relevant studies. Then, full texts of relevant studies were obtained and assessed to determine whether the articles met the inclusion criteria. Any disagreement was discussed and solved with a third author to reach consensus in every relevant detail.

Two authors extracted data independently from the included studies and filled into an extraction form. The following data were extracted: (1) authors; (2) year of publication; (3) study design; (4) inclusion/exclusion criteria; (5) subject demographics (age, gender, number of subjects, level of lesion, classification of ASIA, and time since injury); (6) intervention; and (7) outcome measures.

The data measured at the beginning (baseline) and at the end of interventions were extracted for meta-analysis. We did not analyze the measurements taken during interventions or at follow-up due to inconsistent measuring time points used across different studies. The studies were excluded if the primary outcome measures were missing or not measured. Of these included studies, two met the inclusion criteria from all searching sources but had only one-session intervention, were also included and analyzed separately.

The methodological quality of the selected RCT studies was independently assessed by the two authors using the Cochrane risk of bias assessment tool (Higgins and Altman, 2008). For the assessment of the quality of the selected clinical trials, the Newcastle Ottawa Scale (Wells et al., 2000; Stang, 2010) was employed and done by two authors independently. Any disagreement was resolved through discussion and consensus with a third author. The publication bias was analyzed by using the funnel plots.

As only one RCT was included, all the included RCT and non-RCT studies were grouped for analysis, except two studies. These two studies were analyzed separately due to only one-session performance of FES-cycling. The scores of MAS and AS of lower limbs were pooled for the analysis. Mean differences and 95% CI were calculated for each primary and secondary outcome measure. The pooled mean difference estimates were calculated according to the random effect models if heterogeneous was significant. The fixed effect models were used to calculate the pooled mean difference estimates if no heterogeneous occurred. Because the dose of number of training sessions was varied across the studies, the effectiveness of spasticity was unknown. In such condition, a subgroup analysis was performed according to the training sessions. As a result of 20 electrical stimulation training sessions decreasing spasticity in stroke (Stein et al., 2015; Nakipoğlu Yuzer et al., 2017), 20 training sessions were set as the cut-point, whereas below 20 training sessions were classified as subgroup 1, and the others were classified as subgroup 2. A meta-regression analysis was also analyzed to examine the linear relationship between the training sessions and spasticity. The random effect model was chosen in meta-regression.

Comprehensive Meta-Analysis Software (Version 3) was used to analyze the data. A p-value less than 0.05 was considered as significant.

A total of 763 studies were identified from the electronic databases and one additional study from hand searching from references lists of available articles. After removing 52 repeated articles, the authors screened and assessed full text for eligibility. There were 12 studies for qualitative synthesis, such as one RCT study and 11 non-RCT studies. Eight of them were quantitatively analyzed for meta-analysis (Sköld et al., 2002; Krause et al., 2008; Szecsi and Schiller, 2009; Mazzoleni et al., 2013, 2017; Ralston et al., 2013; Kuhn et al., 2014; Yas̨ar et al., 2015), and four studies were excluded due to reasons, such as case report (n = 1) (Fattal et al., 2018), no MAS result of post-test (n = 2) (Reichenfelser et al., 2012; Popovic-Maneski et al., 2018), and only LEMS data included (n = 1) (Sadowsky et al., 2013). The flowchart is shown as Figure 1. The characteristics of all the included studies are shown in Table 1.

After screening, 683 studies were eliminated and 29 full-text articles were assessed for eligibility. The reasons for exclusion were as follows: passive cycling, review articles, no full text, and no assessing primary outcome measure of the current meta-analysis.

Among 12 included studies, 2 trials were done in the United States (Sköld et al., 2002; Sadowsky et al., 2013), 2 in Italy (Mazzoleni et al., 2013, 2017), 3 in Germany (Krause et al., 2008; Szecsi and Schiller, 2009; Kuhn et al., 2014), others in Australia (Ralston et al., 2013), Austria (Reichenfelser et al., 2012), France (Fattal et al., 2018), Serbia (Popovic-Maneski et al., 2018), and Turkey (Yas̨ar et al., 2015).

The eight studies for quantitative analysis included 99 subjects with SCI (male: female = 83:16). The age of the participants ranged from 20 to 67 years. The time since injury was below 4 weeks to 17 years. ASIA level of the number of participants was 59 for ASIA A, 11 for ASIA B, 18 for ASIA C, and 11 for ASIA D. There were 43 subjects with tetraplegia and 56 subjects with paraplegia.

The four studies which were excluded for the quantitative analysis comprised 78 participants with SCI and 19 healthy controls (male: female = 80:17). The age of the participants ranged from 22 to 62 years. The time since injury was from 1 month to 21 years. ASIA level of the number of participants was 32 for ASIA A, 22 for ASIA B, 15 for ASIA C, and 9 for ASIA D. There were 43 subjects with tetraplegia and 35 subjects with paraplegia.

The frequency of FES was from 20 to 100 Hz and pulse width was from 50 to 600 μs in all the included studies. The FES-cycling training protocol was ranged from 20 min to 1 h per session, 2 to 4 sessions per week, and 2 to 118 weeks in total.

Tables 2, 3 summarize the risk of bias judgments related to one RCT and 11 non-RCT studies. The RCT study (Ralston et al., 2013) and the nine non-RCT studies had good quality, and the two non-RCT studies (Krause et al., 2008; Szecsi and Schiller, 2009) comprising one session of FES-cycling were of moderate quality.

Supplementary Appendix 2 showed the funnel plot and Egger’s test of MAS. The funnel plot seemed asymmetrical. However, the Egger’s test showed no significant publication bias (p = 0.60094).

The MAS is commonly assessed for spasticity in the clinic. The level of MAS represents the reflex and non-reflex components of spasticity (Chang et al., 2007). Abnormal spinal circuitry function is related to the reflex. H reflex reflects the excitation of the alpha motor neuron pool and is related to spasticity (Phadke et al., 2010). A previous study reported one session of sub-threshold electrical stimulation of gastrocnemius muscle (15–20 min, < 10 mA, 2,000–6,000 Hz) decreased Hmax/Mmax in healthy human adults (Yuan et al., 2010). Other study reported that one session of neuromuscular electrical stimulation of spastic quadriceps muscles (20 min, 300 μs, 25 Hz, and maximum 100 mV) reduced spasticity in people with SCI (Tancredo et al., 2013). Non-reflex part may come from the rhythmical leg movements. Repeated stretches of the ankle joint can reduce the resistance during passive stretch in the healthy adults (Avela et al., 1999) and H reflexes in individuals with SCI (Chang et al., 2007, 2013; Fang et al., 2015).

The effect of FES-cycling training on spasticity might partially come from the FES itself. Clinically, FES had been applied to improve spasticity. Sivaramakrishnan et al. (2018) reported that both 30 min, one session of FES and transcutaneous electrical nerve stimulation (TENS) decrease spasticity measured by MAS in acute to chronic SCI. The review from Naro et al. (2017) reported that FES 30 min two times a day, 5 days a week for 4 weeks can decrease spasticity in patients with SCI and stroke. FES itself has been applied to reduce spasticity not only in SCI (Bochkezanian et al., 2018) but also in stroke or cerebral palsy (Ho et al., 2014; Naro et al., 2017). A previous study also reported high-intensity knee extension NMES training two times a week for 12 weeks decreased spasticity in the patients with chronic SCI (Bochkezanian et al., 2018). However, a previous meta-analysis from Thomaz et al. (2019) reported that electrical stimulation did not improve spasticity in people with SCI. That could be due to their included studies were heterogeneous in intervention methods, such as FES walking, FES-cycling, and TENS. In our meta-analysis, we specifically selected FES-cycling training as the intervention to examine its effect on spasticity. The results showed that the effect of spasticity reducing was not simply from electrical stimulation.

The effect of reducing spasticity by FES-cycling training might also be associated with the cycling part. Cycling training was suggested to have effect on neural plasticity through the normalized spinal reflexes caudal to spinal injury level in the humans and rats (Lynskey et al., 2008). Cycling exercise could generate joint motions on the hips, knees, and ankles. The previous studies showed that passive cycling of ankle joint at a speed of 50 cycles/min decreased the reflex excitability in patients with SCI after one bout (Fang et al., 2015) and multiple sessions of training (Chang et al., 2013).

Several possible underlying mechanisms might be responsible for FES-cycling training to reduce spasticity. One is the restoration of post-activation depression. Post-activation depression is a pre-synaptic inhibition which serves to stable the joint and prevent clonus (Masakado et al., 2005; Yang et al., 2015). Post-activation depression was found to be decreased in the chronic phase after SCI that was considered as a cause of spasticity (Chang et al., 2013). A previous study showed that post activation depression could be restored after passive cycling training on ankle joints for 4 weeks with a parallel improvement of spasticity in the individuals with chronic SCI (Chang et al., 2013). Another possible mechanism for spasticity reduction might be through Renshaw cell inhibition and/or recurrent inhibition (Milosevic et al., 2019). The electrical current activated the muscles orthodromically and antidromic volley depolarized the α-motoneuron cell bodies. The antidromic impulse might also activate Renshaw cell interneurons resulting in recurrent inhibition (Mazzocchio et al., 1994). The other possible reason for reducing spasticity by FES-cycling training might be muscle fatigue induced by FES intervention (Krause et al., 2008).

The current meta-analysis showed the score of the MAS significantly reduced 0.86 after FES-cycling training in people with SCI. No previous study reported the minimal clinically important difference of MAS for people with SCI. The possible clinical impact in contrast to the minimal clinically important difference obtained from the patients with stroke, which was 0.73 (Chen et al., 2019), suggested that FES-cycling training induced spasticity decrement might be beneficial for people with SCI clinically.

A previous systematic review performed by Bekhet et al. (2019) suggested the frequency, pulse duration, and current amplitude of ES to be 20–30 Hz, 300–350 μs, and > 100 mA for reducing lower limb spasticity in people with subacute or chronic SCI. In our current meta-analysis, 6 of 12 included studies used the frequency within the range of 20–30 Hz (Sköld et al., 2002; Krause et al., 2008; Szecsi and Schiller, 2009; Kuhn et al., 2014; Yas̨ar et al., 2015; Fattal et al., 2018). One study used a higher frequency (100 Hz) (Sadowsky et al., 2013). For the pulse duration, five included studies used 250–350 μs (Sköld et al., 2002; Ralston et al., 2013; Kuhn et al., 2014; Yas̨ar et al., 2015; Fattal et al., 2018), whereas the other six studies used 500–600 μs (Krause et al., 2008; Szecsi and Schiller, 2009; Reichenfelser et al., 2012; Mazzoleni et al., 2013, 2017; Sadowsky et al., 2013), and two studies did not reported the frequency (Mazzoleni et al., 2013; Popovic-Maneski et al., 2018) or pulse duration (Popovic-Maneski et al., 2018). In general, the included studies used a stimulation frequency agree with Bekhet’s suggestion, but the pulse duration was varied. This might suggest that the frequency requirement is more restricted than pulse duration while performing FES-cycling training. According to the review of Marquez-Chin and Popovic, 20 Hz is the minimum frequency to produce tetanic contraction (Marquez-Chin and Popovic, 2020). Since FES-cycling needs the muscle to produce tetanic contraction, this could explain why the frequency used in all studies was above 20 Hz. Therefore, our current meta-analysis suggests FES at a frequency range from 20 to 30 Hz is suitable for FES-cycling training to improve spasticity. One of the selected study by Sadowsky et al. (2013) used a high frequency and longer pulse duration (100 Hz and 500 μs). This might induce high frequency fatigue (Behringer et al., 2016), but the muscle fatigue was not reported. It may be that NMES/FES can be delivered using longer pulse duration and higher frequency (1 ms of pulse widths, frequency ∼100 Hz) that are used conventionally to reduce contraction fatigability and enhance the benefits of NMES/FES-based programs for the neuromuscular and cardiovascular systems by generating contractions through spinal and possibly transcortical “central pathways” (Collins, 2007; Bergquist et al., 2011; Bastos et al., 2021).

A previous meta-analysis on the patients with stroke showed that 15–20 sessions of NMES significantly decreased spasticity (Stein et al., 2015). Prior study reported more than 10-week FES altered the neural circuit on the patients with SCI (Aziz et al., 2009). Our study is the first meta-analysis to investigate and reveal that 20 FES-cycling training sessions are required to reduce spasticity in the patients with SCI. Although the disease courses and rationale of SCI and stroke are different, the sessions required for electrical stimulation on decreasing spasticity are similar. Neural circuit adaptations, such as normalization of H reflex and post-activation depression, are suggested to be an underlying mechanism to reduce spasticity (Chang et al., 2013). Our meta-analysis also found that the number of training sessions and the effects on reducing spasticity are not linearly related. The training effect may approach a plateau as training sessions increase.

The ASIA level might be the other factor to influence the effect of FES-cycling training on spasticity. Most included studies recruited subjects with SCI with mixed ASIA levels. Hence, it was subgrouped to complete (ASIA A) and incomplete (ASIA B, C, and D) injury. The results were that spasticity significantly reduced in complete and incomplete SCI in favor of FES-cycling training. The pooled mean differences in complete and incomplete subgroup were −1.08 (95% CI = −1.281 to −0.878, p < 0.001) and −0.731 (95% CI = −1.393 to −0.070, p = 0.030), respectively. The forest plot was shown in Supplementary Appendix 3. Since no previous study reported the minimal clinically important difference of MAS for people with SCI, the possible clinical impact in contrast to the minimal clinically important difference obtained from the patients with stroke, which was 0.73 (Chen et al., 2019) might demonstrate beneficial effect of FES-cycling training for people with SCI. The results of both complete and incomplete SCI exceeded clinically important difference of spasticity after FES-cycling training, and the reduction of spasticity was more in complete SCI than in incomplete SCI. Although people with incomplete SCI can perform voluntary exercise without the aid of FES-cycling, they still are benefited from FES-cycling training on the reduction of spasticity.

The included studies did not analyze the results by the level of injury. Not every included study reported the definite injury level, such as Sköld et al. (2002), which showed C1 to T1 injury level, and Szecsi and Schiller (2009) which reported C7 to T12. For the time since injury, some included studies showed a wide range of time since injury (less than 4 weeks to 17 years), and some did not report this data (Mazzoleni et al., 2013, 2017). The results of MAS in every included study were not reported by each level of injury or time since injury. This makes the subgrouping analysis not possible.

For the session duration, all the included studies had a similar session duration which was from 20 to 60 min per session. This might be restricted by tolerance of the subjects. The frequency of FES of included studies was from 2 to 4 times a week in all the included studies. This follows the general training guidelines. Therefore, subgroup analyzation of the influence of session duration and/or frequency could not be done with the available studies.

Six Minute Walk Test and TUG are related to the walking abilities. Our meta-analysis revealed that FES-cycling training improved the walking abilities after SCI. Spasticity reduction could be a factor causing this improvement as it plays a role in abnormal gait patterns after SCI (Krawetz and Nance, 1996). According to a study, the spasticity presented by thoracic SCI is the most obvious, and their knee excursion and knee angular velocity showed the greatest deviation among all the SCI groups (Krawetz and Nance, 1996). Guimaraes et al. (2016) suggested that subjects with traumatic spinal cord injury above T12 level were the best potential candidates for FES-cycling training. The subjects in our current meta-analysis were mostly thoracic and cervical injury level. Meanwhile, reduction of spasticity after FES-cycling training showed beneficial effect on the walking abilities measured by 6MWT and TUG in SCI.

Six Minute Walk Test is also an indicator of endurance (Bennell et al., 2011). Our meta-analysis supported that the individuals with SCI could be benefited from FES-cycling to increase endurance. TUG is a common test of functional mobility. It was shown to be correlated with muscle strength of the lower extremities, balance, and gait speed in elderly adults (Podsiadlo and Richardson, 1991; Chen and Chou, 2017). Our results revealed that LEMS improved in favor of FES-cycling training. The enhancement of LEMS after FES-cycling training can promote TUG performance.

Our current meta-analysis showed that the pooled LEMS improved significantly after FES-cycling training although the included studies were few. To compare the two included studies, more improvement of LEMS was found in Sadowsky et al. (2013) study than in Yas̨ar et al. (2015) study. The impairment level of the subjects was different between Sadowsky et al. (2013) and Yas̨ar et al. (2015) studies. Most subjects from Sadowsky et al. (2013) study were ASIA A level, and all the subjects but one from Yas̨ar et al. (2015) study were ASIA D. Both studies showed LEMS increased significantly after FES-cycling training. It is possible that FES-cycling is effective on improving the strength of lower limbs in the subjects with different ASIA classifications, but the improvement is more in subjects with ASIA A than with ASIA D.

A previous narrative review proposed FES-cycling training of 45–60 min, three times a week for at least 4 weeks to improve muscle strength in people with SCI (Rosley et al., 2019). It has also been shown resistance training of 18–27 weekly sets, completed over 2 weekly sessions for 6 weeks (total 12 sessions) increased muscle strength in trained people (Heaselgrave et al., 2019). The FES-cycling training protocols of two included studies (Sadowsky et al., 2013; Yas̨ar et al., 2015) showed muscle strength increase with at least 4 weeks training. The time since injury in these two included studies was 27–96 months. Motor recovery had been evidenced within 18 months after SCI (Krawetz and Nance, 1996; Guimaraes et al., 2016; Luo et al., 2020). Therefore, we can rule out spontaneous recovery of muscle strength. The effect of LEMS improvement arises from FES-cycling training. NMES and FES training have been reported to increase muscle mass, mitochondrial oxidative enzyme activities, plasma glucose level, and circulating insulin (de Freitas et al., 2018; Gorgey et al., 2019). However, detailed discussion of underlying mechanism is beyond the scope of this study.

This meta-analysis has some limitations. First, most included studies were not RCT studies. Future studies with high quality RCT would be suggested to provide stronger evidence. Second, MAS of different lower extremities joints were pooled for analysis. Future studies differentiating joint MAS are suggested for clinical management. Third, due to the design of included studies, factors, such as level of injury, time since injury, frequency of FES, or session duration could not be analyzed by subgrouping. Future studies are suggested to clarify the differential effects on level of injury, frequency of FES, and treatment session duration. Fourth, the total number of included studies was small. A meta-analysis is a statistical combination of results from two or more separate studies. Variation across studies (heterogeneity) is considered. The prediction intervals from the random-effect models are a useful device for presenting the extent of between-study variation (Deeks et al., 2021). Although the total number of studies was small, we chose the random-effect models to estimate the effect and made careful interpretation.