Identification of studies
A PRISMA flowchart of the search strategy, study selection, and exclusions by stage of the systematic review is shown in Fig. 1. Searches of the databases identified 853 studies. The manual search screening through the reference lists identified 3 studies. After duplicates removal, 634 potentially relevant studies were screened by title and abstract. Subsequently, the full texts of the remaining 93 studies were assessed. A total of 14 studies met our inclusion criteria and were described in qualitative analysis (Fig. 1). All studies were published in English. Four trials were conducted in Spain, 4 in the U.S.A, 1 in the UK, 1 in Finland, 1 in Denmark, and 2 in Austria.
Study selection and characteristics
Characteristics of studies are summarized in Table 1, and Table2 and summarized below. All of the included controlled studies were RCTs. Eight trials compared parallel intervention groups [10-13, 15, 22-24] and 6 studies [25-30] used a crossover design. The test groups all received NIBS, and patients in the control group received the sham stimulation. In addition, 2 studies combined physical therapy [10, 27], 5 combined robot training [11, 12, 15, 23, 24], 1 combined peripheral nerve stimulation[22], 1 combined massed practice[13],1 combined antispastic medication [30]. In terms of the stimulation pattern of TMS, all but one [22] of the remaining studies used excitability stimulation patterns. All articles involving tDCS used anodal stimulation. The frequency of treatment ranged from three [24, 27] to four[22], five[10-13, 15, 23, 26, 28-30] or seven[25] times per week. The duration of treatment ranges from 3[25] to 36 [24] sessions. The treatment intensity (in terms of session duration) ranged from 200s [26, 29] to 30 min[13] and the treatment time did not differ between the control and experimental groups. Due to considerable heterogeneity in study designs about follow-up, the first assessment available post-intervention was chosen as a follow-up. The data were sorted and analyzed based on outcome to provide an overview of the effects of NIBS on each of the categories: 5 studies used results from UEMS[10, 12, 23, 25, 26] and 2 used JTHFF [11, 27] on the upper extremity strength, 5 used LEMS[10, 12, 23, 25, 26] on the lower extremity strength, 2 used UMAS[22, 26], 7 used LMAS [15, 22, 23, 25, 28-30] and 3 used H/M ratio[28-30] on spasticity, 3 used BBT[15, 24, 25] on body balance, 4 used 10MWT [10, 15, 24, 25], 3 used 6MWT[10, 15, 24] and 4 used TUG [15, 24] on mobility. Thus, the results of these clinical trials were pooled in different meta-analyses.
A total of 225 participants were pooled in the meta-analyses and the demographic characteristics are shown in Table1. The number of participants in each study ranged from 5[22] to 43[15]. The pooled sample was predominantly males (73.78%) with a mean (SD) of 44.31 (15.08) years of age and 2.77 (5.32) years of the duration of post-injury. All included studies provided information on the level of spinal cord injury and baseline severity according to the ASIA. The number of patients with the injury level cervical was 128 (56.89%) of the sample, whereas thoracic was 92 (40.89%) and lumbar was 5 (2.22%). Complete SCI at A level of impairment in ASIA was present only in 1 patient, and incomplete SCI at B level was in 5(2.22%), level C in 115 (51.11%), and level D in 104 (46.22%) of these patients.
Adverse effects
Among 14 included studies, 8 reported no obvious adverse effects [12, 13, 22, 24-26, 28, 29]. One study has reported that 1 patient experienced a seizure during TMS stimulation [10]. Five studies have reported minor adverse effects [11, 15, 23, 27, 30], such as tingling, itching, skin redness, sleepiness, facial muscle contraction, or headache, which were observed also in the sham group.
Quality
Figure 2 presents the review authors judgments about each risk of bias domain and percentages of risks across all included studies. 8 studies (57.14%)[12, 13, 22, 23, 25, 28-30] reported adequate random sequence generation and 3 (21.43%) [10, 11, 15] hid the allocation scheme, all presented blinding of participants and personnel, 13 (92.86%) presented blinding of outcome assessment and all described a low risk for attrition, showing a low risk of bias. Therefore, all of the included studies presented moderate to high methodological quality.
Effects of interventions
Extremity strength
In the NIBS group, lower extremity strength measured by LEMS was greater than that in the sham groups (SMD=0.58, 95% CI=0.02 to 1.14, P =0.04, I2=52%) (Fig. 3a). No greater improvements in the upper extremity strength were observed in the NIBS groups compared to the sham groups (SMD =0.4, 95% CI= -0.03 to 0.83; P= 0.07; I2 =0%) (Fig. 3a).
Balance
Significantly greater improvements by BBT were observed in the NIBS groups compared to the sham groups (SMD =0.64, 95% CI= 0.05 to 1.24, P =0.03, I2 =0%) (Fig. 3b).
Spasticity
The changes of spasticity in upper limbs assessed by UMAS were not more significant in the NIBS groups than in sham groups (SMD= -1.25, 95% CI=-2.83 to 0.34, P=0.12, I2 =62%) (Fig. 4a). In the NIBS groups, significant changes in lower limbs assessed by LMAS were observed than in the sham groups (SMD=-0.61, 95% CI=-1.20 to -0.03, P=0.04, I2=64%) (Fig. 4a). However, the overall changes in spasticity of lower limber measured by H/M ratio were similar in the sham groups and sham groups (SMD= -0.95, 95% CI=-2.64 to 0.73, P=0.27, I2 = 86%) (Fig. 4b).
Mobility
Mobility was similar in the NIBS groups and sham groups, which was evaluated by the gait distance of 6MWT (SMD=-0.17, 95% CI =-0.68 to 0.34; P=0.51; I2 =63%) (Fig. 5a), the speed of 10MWT (SMD=0.85, 95% CI= -0.07 to 1.76; P=0.07; I2 =14%) (Fig. 5b), time-to-complete the 10MWT (SMD=-0.35, 95% CI= -0.88 to 0.18; P=0.19; I2 =0%) (Fig. 5c),and TUG (SMD=0.01, 95% CI= -0.51 to 0.52, P =0.98, I2=16%) (Fig. 5d).
Sensitivity analysis results
Sensitivity analysis revealed that the heterogeneity across subgroups did not change after excluding any one study, suggesting the source of the heterogeneity was multifaceted. The results are shown in Additional file 2.