Walking speed is not the best outcome to evaluate the effect of robotic assisted gait training in people with motor incomplete Spinal Cord Injury: A Systematic Review with meta-analysis

Abstract

Context: While there are previous systematic reviews on the effectiveness of the use of robotic-assisted gait training (RAGT) in people with spinal cord injuries (SCI), as this is a dynamic field, new studies have been produced that are now incorporated on this systematic review (SR) with meta-analysis, updating the available evidence on this area.

Objective: To synthesise the available evidence on the use of RAGT, to improve gait, strength and functioning.

Methods: SR and meta-analysis following the Cochrane Handbook for Systematic Reviews of Interventions were implemented. Cochrane Injuries Group Specialized Register, PubMed, MEDLINE, EMBASE, CINAHL, ISIWeb of Science (SCIEXPANDED) databases were reviewed for the period 1990 to December 2016.

Three researchers independently identified and categorized trials; 293 studies were identified, 273 eliminated; remaining 15 randomized clinical trials (RCT) and five SR. Six studies had available data for meta-analysis (222 participants).

Results: The pooled mean demonstrated a beneficial effect of RAGT for WISCI, FIM-L and LEMS (3.01, 2.74 and 1.95 respectively), and no effect for speed.

Conclusions: The results show a positive effect in the use of RAGT. However, this should be taken carefully due to heterogeneity of the studies, small samples and identified limitations of some of the included trials.

These results highlight the relevance of implementing a well-designed multicenter RCT powered enough to evaluate different RAGT approaches.

Background

Spinal Cord Injury (SCI) is a lesion that may result in sensitive, motor and autonomic impairments.¹ Damage caused to the descending and ascending tracts results in postural system impairment, which affects standing, locomotion and voluntary movements.^2–4 SCI affects the ability to walk and many individuals do not regain it, even though it is one of the goals of rehabilitation,⁵ along with upper extremity function, sexuality, bowel and bladder control.⁶ More than 50% of individuals with SCI present incomplete motor lesions and more than 75% of those individuals regain some form of ambulatory function.^1,7 However, common consequences include slow speed walking, abnormal step length, cadence, and step symmetry that negatively impact walking efficiency.⁸

There has been an evolution on gait rehabilitation programs for people with spinal cord injury in the recent decades, from manually assisted over-ground training; body-weight-supported treadmill training (BWSTT) to robotic-assisted gait training (RAGT). All these interventions have the common goal of regaining or improving locomotion by providing sensory information⁵ and interactions with supraspinal circuits (cortical and subcortical).^4,9

RAGT was introduced in the late 1990s (Lokomat, Hocoma AG, Switzerland). Today different commercial systems, such as Lokomat® (Hocoma, Volketswil, Switzerland),¹⁰ G-EO systemTM (Reha Technology AG, Switzerland), Walkbot (P&S Mechanics Co., Ltd, Korea), ReoAmbulatorTM (Motorika, USA Inc.), among others are available. The RAGT consists of a motor driven gait orthosis, controlled by a computer and secured to a patient's legs while the patient is supported by a BWS. The RAGT focuses on the correct performance of gait movements. Therapy is performed at low speed and the level of assistance by the system can be adjusted based on the patient's ability to step.

Multiple studies have demonstrated the efficacy of this orthosis to improve walking on people with incomplete motor spinal cord injury who are partially able to walk in the sub acute phase.^11–13

To practice the kinematically correct stepping is thought to enhance the afferent feedback associated with normal locomotion and, therefore, to maximize plasticity within spinal and supraspinal neural circuits.¹¹ Previous to the RAGT, BWSTT had been used, though its use is limited because of its labour-intensive requirements.¹¹

Currently, there are 5 systematic reviews related to this theme. Swinnen et al. 2010,¹⁰ Tefertiller et al. 2011,¹⁴ Mehrholz et al. 2012,¹ and Morawietz et al. 2013,¹⁵ and Karimi¹⁶et al. 2013.

Those previous studies have used gait velocity as a measure of overall motor capacity and gait recovery. Results of locomotor recovery after an intervention such as RAGT or BWSTT probably should not be assessed only by improvement on gait velocity after intervention. In fact, gait is a complex motor task produced by interaction of neurological and biomechanical systems. Additionally, the RAGT focuses on the correct performance of gait movements, and the training speed is usually slow.

According to the International Classification of Functioning, Disability and Health, functioning involves individual body function, body structures, activities and participation, and denotes the interaction between a health condition and his/her living conditions. Two constructs, ‘performance’ and ‘capacity’, are used to operationalize domains of activities and participation.¹⁷ Performance can also be measured using alternative scales that assess walking in people with spinal cord injury related to what the individual actually does in his or her current environment.¹⁷ For performance assessments the ICF, WISCI (walking index for spinal cord injury), SCIM (spinal cord independence measure), FIM (functional independence measure), among other scales may apply.

This study contributes to the available evidence on the use of RAGT in people with spinal cord injury by incorporating the latest evidence from clinical trials as well as by widening the scope with the inclusion of additional indicators of effectiveness: improve gait, strength and functioning in people with spinal cord injury in comparison to other modalities of training.

Methods

Types of studies

RCT, systematic reviews and crossover trials (only the first period) were included, although analysis were implemented independently for each study type.

Types of participants

Individuals with any level of traumatic incomplete SCI, regardless of the time since injury, sex and age, were included.

Types of interventions

The study included all trials that addressed the effectiveness of RAGT compared with other training modalities as part of a neurorehabilitation program to improve gait, strength and functioning in comparison to an alternative intervention.

RAGT interventions were required to have a main focus on gait, strength or functioning. There were no restrictions regarding to frequency and duration of the RAGT interventions.

Outcome measures

• Gait parameters: Instrumented gait assessment, 10MWT, 6MWT, WISCI or any other available scale.

• Strength: Isokinetic, L-Force, MRC (medical research council scale), or any other available scale.

• Functioning: SCIM, FIM or any other available scale.

Searching criteria

The search was not restricted by language or publication status. The search was limited to studies published after 1990, the year when RAGT was introduced.¹⁰

Data sources

Searches were conducted in using the following databases: Cochrane Injuries Group Specialized Register (recent issue); Cochrane Central Register of Controlled Trials (CENTRAL); MEDLINE (Ovid) 1990 to december 2016; EMBASE (Ovid) 1990 to december 2016; CINAHL 1990 to december 2016; ISIWeb of Science: Science Citation Index Expanded (SCIEXPANDED) 1990 to december 2016; Pub Med (www.ncbi.nlm.nih.gov/sites/entrez/). The following terms were used for the search:

Cochrane Injuries Group Specialised Register (Search from 5 of August 2016)((spine or spinal) and (damag* or trauma* or injur* or broke* or break or fracture)) and (walk* or locomotor or rehabilitat* or robot* or orthos* or orthotic or automat* or "computer aided" or "computer assisted" or Lokomat or Locomat* or electromechanical or electro-mechanical or mechanical or mechanised or mechanized or driven) in Title, Abstract, Keywords in Cochrane Reviews

Medline (Pubmed)

1. exp Spinal Cord Injuries/

2. exp Spinal Cord Ischemia/

3. exp Central Cord Syndrome/

4. ((spine or spinal) and (fracture* or wound* or trauma* or injur* or damag*)).ab,ti.

5. (spinal cord adj3 (contusion or laceration or transaction or trauma or ischemia)).ab,ti.

6. central cord injury syndrome.ab,ti.

7. central spinal cord syndrome.ab,ti.

8. exp Cervical Vertebrae/in [Injuries]

9. exp Spinal Cord/

10. SCI.ab,ti.

11. exp Paraplegia/

12. exp Quadriplegia/

13. (paraplegia* or quadriplegia* or tetraplegia*).ab,ti.

14. or/1–13

15. exp Gait/

16. exp Walking/

17. (locomotion and walking).ti,ab.

18. *Locomotion/

19. locomotor?training.ab,ti.

20. *Dependent Ambulation/

21. (walk* or gait* or ambulat* or mobil* or locomot* or stride*).ti,ab

22. or/15–21

23. 14 and 22

24. exp Automation/

25. exp Robotics/

26. exp Orthotic Devices/

27. exp Weight-Bearing/

28. (weight?bearing or load?bearing).ab,ti.

29. (electromechanical or electro-mechanical or mechanical or mechanised or mechanized or driven or exoskeleton).ti,ab

30. (robot* or orthos* or orthotic* or automat* or computer?aided or computer?assisted or BWS or harness* or treadmill* or Lokomat or Locomat or G-EO).ab,ti

31. ((gait or walk* or ambulatory) adj3 (recover* or test* or abilit* or function or speed*)).ab,ti

32. or/24–31

33. 23 and 32

34. randomi?ed.ab,ti.

35. exp randomized controlled trial

36. controlled clinical trial.pt.

37. placebo.ab.

38. clinical trials as topic.sh.

39. randomly.ab.

40. trial.ti.

41. exp review

42. or/ 34–41

43. (animals not (humans and animals)).sh.

44. 43 not 45

45. (rat or rats or rodent* or mouse or mice or murine or dog or dogs or canine* or cat or cats or feline* or rabbit or rabbits or pig or pigs or porcine or swine or sheep or ovine* or guinea pig*).ti

46. 44 not 45

47. 33 and 46

ISI Web of Science: All databases 1990 to present

• #1 Topic=(SCI OR spinal cord injuries OR central cord syndrome) AND Topic=(walk* OR gait* OR ambulat* OR mobil* OR locomot*) AND TS=(robot* OR orthos* OR electromechanical OR orthotic* OR automat* OR lokomat OR locomat OR G-EO)

• #2 Topic=(randomi?ed OR randomized controlled trial OR controlled clinical trial OR placebo OR clinical trials OR randomly OR trial OR review) NOT Topic=(animal* OR rat OR rats OR rodent* OR mouse OR mice OR murine OR dog OR dogs OR canine* OR cat OR cats OR feline* OR rabbit OR rabbits OR pig OR pigs OR porcine OR swine OR sheep OR ovine OR guinea pig*)

• #3 #1 and #2

Searching other resources

We contacted key authors and institutions to request details of any recently published, in press, unpublished or ongoing trials, reference lists of included studies and literature reviews, searched bibliographies of relevant studies and reviews; relevant experts in the field were also contacted.

Literature screening & study selection

Three authors independently examined titles, abstracts and keywords to identify potentially relevant studies. After the initial search, full texts of identified relevant studies were obtained; the studies were assessed by two authors and included if the inclusion criteria were met according to both researchers. Disagreements were resolved by a third reviewer.

Data extraction and management

Extracted data were filled into a pretested data collection form by at least 2 reviewers. Detailed instructions and training were provided to all authors involved in data extraction.

Assessment of risk of bias in included studies

Risk of bias was assessed to evaluate trial quality by at least two reviewers independently, using the criteria described in the Cochrane Handbook for Systematic Reviews of Interventions.¹⁸ Disagreements were resolved through consultation with a third reviewer.

Statistical analysis

The strength of the study findings was discussed by level of evidence, which was based on methodological quality. Classification of outcome measures in terms of the assessed domain (gait, strength, and functioning) was done by at least 2 reviewers independently.

All analysed outcome measures are reported as continuous variables. Mean differences (MD) and 95% confidence intervals (CIs) were calculated. A random-effects model was used for all outcomes analysed. Since it was difficult to identify the subset of participants with reported baseline and final value measurements, only final value measurements were used for the analysis. The approach suggested by Wan et al. 2014,¹⁹ was used to enter missing mean and S.D. from median and interquartile ranges. Briefly, inequalities are developed for each observation using upper and lower limits from the minimum, the three quartiles, and the maximum. These are summed to give bounds for the sum and hence the mean of the observations, the average of these bounds in the estimate.

Whenever a study presented results for several periods of follow-up for the same outcome, the last assessment was included as the final value measurement.

Studies’ authors were contacted to acquire any missing data as well as information on whether or not data could be assumed to be missing at random.

The number of participants in each meta-analysis corresponds with the number of participants included in the analysis of the published trials per group of intervention. Data were analyzed using Review Manager Software 5.3.

Results

293 studies were identified on the electronic search. 47 of them were duplicated studies and were eliminated. Results of the search are shown in Fig. 1.

Figure 1.

Flow Diagram

The full text review was carried out for the 28 preliminary selected trials, 20 of them met the inclusion criteria for this review: five SR,^1,10,14–16 and 15 RCT.^20–34 Characteristics of included studies are shown in table 1.

Table 1. Characteristics of included studies.

Study	Design	Participants	Intervention	Outcome measures
Alcobendas-Maestro 2012	RCT	n= 80 Time since injury (months)= <6 AIS= C or D Level of injury= C2 to T12	Group 1: Patients received 40 walking reeducation sessions of equal time using a Lokomat program with overground practice. Group 2: Overground mobility therapy alone.*training in both groups consisted of 1 hour of training, the Lokomat group used the system for 30 minutes, the other 30 minutes were completed with conventional therapy.	WISCI 6MWT FIM-L LEMS MAS VAS
Duffel 2014	RCT	n= 78 patients with incomplete SCI	Group 1: Lok group, locomotor training was provided using a robot-assisted locomotor training device Training was provided three times per week; each session lasted <1 hour, with 30–45 minutes of training. (4 w). Group 2: Tiz group,.03 mg/kg of Tizanidine was administered four times a day for four weeks. (4 w). Group 3: Control subjects received no intervention.	Walking speed Endurance and mobility TUG 10MWT 6MWT
Duschau 2010	RCT	n= 15 Participants with chronic incomplete SCI	Group 1: POS: Position control with the stiffness of the Lokomat controller set to Khip = 1200 Nm/rad, Kknee = 900 Nm/rad2. Group 2: SOFT: Impedance control with the stiffness set to Khip = 192 Nm/rad, Kknee = 144 Nm/rad. Group 3: COOP: Path control with window set to 20% of the gait cycle and the support gain ks adjusted individually for each patient3 Group 4: COOP+: Path control with window set to 20% of the gait cycle and the support gain ks increased to 130% of the value used in the previous condition Single session	Spatio-temporal characteristics Peek knee extension Maximal hip flexion during swing phase
Esclarín-Ruz 2014	RCT	n= 88 Time since injury (months)= <6 months AIS= C or D Level of injury= C2 to T11	Group 1: Subgroups A1 and B1 (LKOGT) were imparted 30 minutes of conventional mobility training plus 30 minutes of robotic-assisted mobility training. Group 2: Subgroups A2 and B2 (OGT) were imparted 60 minutes of conventional mobility training. **	10MWT 6MWT WISCI LEMS FIM-L
Field-Fote 2005	RCT ***	n= 27 Time since injury (months)= >12 months Level of injury= at or above T10	Group 1: Treadmill training with manual assistance (TM). Group 2: Treadmill training with stimulation (TS). Group 3: Over- ground training with stimulation (OG). Group 4: Treadmill training with robotic assistance (LR). ****	Walking speed Training speed Step length Step symmetry
Field-Fote 2011	RCT *****	n= 74 AIS= C or D Level of injury= at or above T10	Group 1: Treadmill training with manual assistance (TM). Group 2: Treadmill training with stimulation (TS). Group 3: Over- ground training with stimulation (OG). Group 4: Treadmill training with robotic assistance (LR). ****	Walking speed Walking distance LEMS (right/left)
Kressler 2013	RCT	n= 62 AIS= C or D Level of injury= at or above T10	Group 1: Treadmill training with manual assistance (TM). Group 2: Treadmill training with stimulation (TS). Group 3: Over- ground training with stimulation (OG). Group 4: Treadmill training with robotic assistance (LR). ****	VO2 Walking velocity Walking economy Substrate utilization: slow, moderate and maximal walking speeds.
Labruyère 2014	Randomized cross-over	n= 9 Chronic incomplete SCI Time since injury (months)= >12 AIS= C or D	Group 1: received 16 sessions of RAGT (45 min each) within 4 weeks followed by 16 sessions of strength training (45 min each) within 4 weeks. Group 2: received the same interventions in reversed order.	10 MWT (preferred and max) Waking speed Balance Strength Risk of falling and pain ******
Lam 2015	RCT	n= 15 Motor incomplete SCI Time since injury (months)= >12 AIS= C or D	Group 1: BWSTT with Lokomat-applied resistance (Loko-R). Group 2: conventional Lokomat-assisted BWSTT (Control). Training sessions were 45 min, 3 times/week for 3 months.	Walking capacity OG (SCI-FAP) Walking speed Walking distance
Niu 2014	RCT	n= 40 Incomplete SCI Spasticity at lower extremities	Group 1:1-hour Lokomat trainings over one month. Group 2: control subjects received no interventions.	10MWT TUG 6MWT MVC MAS
Nooijen 2009	RCT	n= 51 Chronic incomplete SCI Time since injury (months)= >12 Level of injury= at or above T12	Group 1: BWSLT on the treadmill with manual assistance for stepping (TM). Group 2: BWSLT on the treadmill with peroneal nerve stimulation to assist stepping (TS). Group 3: BWSLT overground with peroneal nerve stimulation (WalkAide2TM, Hanger Orthopedic Group, Inc., Bethesda, MD; OG). Group 4: BWSLT on the treadmill with assistance of a locomotor robot (Lokomat, Hocoma AG, Zurich, Switzerland; LR). 12 weeks of training	10MWT 6 mts of the walkway
Quinzaños 2014	RCT	n= 31 Time since injury (months)= >6 AIS= C or D	Group 1: 12 training sessions, 20 min., 4 times per week (3 weeks). body weight support was 50%, every week a decrease of 10% was made. Guidance was determined depending on lovet scale; 4 and 5, 20%, for a 3, the 40% was used, for a 2, 60% was used, and for 1 and 0 a guidance of 80% was assigned. Group 2: The same parameters of the Lokomat were used + an auditive feedback.	Spatio-temporal variables Cadence Range of movement Spasticity SCI-FAP
Shin 2014	RCT	n= 60 Time since injury (months)= <6 AIS= D Non progressive SCI	Group 1: RAGT three sessions per week at duration of 40 minutes with regular physiotherapy in 4 weeks. Group 2: The conventional group underwent regular physiotherapy twice a day, 5 times a week.	LEMS AMI SCIM WISCI
Tang 2014	RCT	n= 30 Incomplete SCI AIS= D Level of injury= T8-L3	Group 1: The total set-up and treatment time for the Lokomat never exceeded 1 hour. The initial training speed was 1.5 km/h and it was progressively raised to 1.8 km/h. The body weight system was initiated at 35%, and 70% guidance force. Group 2: The Ergo_bike group subjects were instructed to pedal at a pedaling rate of 45 rpm with a work load of 60 W. 1 training session of 40 minutes.	P-RT 10 m máximum walking speed
Wu 2012	Cross-over	n= 10 AIS= D Level of injury= C2 to T10	Group 1: One group received 4 weeks of assistance training followed by 4 weeks of resistance training. Group 2: The other group received 4 weeks of resistance training followed by 4 weeks of assistance training.	Walking velocity (self-selected and fast) 6MWT balance Muscle tone Strength

* This information was provided via email from the authors.

** No subjects were familiar with the Lokomat robotic-assisted mobility training system before participating in the study.

***Preliminary study of the 2011 study.

**** Subjects in all groups were allotted a 60-minute intervention for each training day, with setup and take-down consuming an average of 10 to 15 minutes of this allotted time. Subjects were scheduled to train 5 days/week for 12 weeks.

***** Subjects in all groups were allotted a 60-minute intervention for each training day, with setup and take-down consuming an average of 10 to 15 minutes of this allotted time. Subjects were scheduled to train 5 days/week for 12 weeks.

******Data from the first period of this trial were provided via email by the authors.

WISCI, walking index for spinal cord injury. 6MWT, 6 minute walk test. 10MWT, 10 minute walk test. FIM-L, functional Independence measure-Locomotion. LEMS, lower estremity motor score. MAS, modified ashworth scale. VAS, visual analog scale for pain. TUG, timed up and go. OG, overground. SCI-FAP, spinal cord injury-functional ambulation profile. MVC, máximum voluntary contraction. AMI, ambulatory motor index.

Excluded studies

In the full text review eight studies were eliminated because they did not meet the inclusion criteria. From these potentially eligible trials, one was excluded because it is a protocol¹⁷ one is about a hybrid exoskeleton to restore gait,³⁵ another uses an electromechanical gait trainer and includes only one person with SCI per group,³⁶ and the rest of the trials^12,37–40 are not RCT.

Study location

From the 15 RCT, seven trials were done in the United States^{20,23,26,28,30,31,34} two in Spain,^25,27 two were from Switzerland,^24,32 one was done in Korea,²² one in Canada,²⁹ one in China³³ and one in Mexico.²¹

Study participants

From the 15 RCT that met the inclusion criteria, a total of 499 participants were registered. From the five SR a total of 1,227 participants were included. The number of participants ranged between 9²⁴ to 88.²⁵ The range of age varied from 16²⁵ up to 70 years,^24,25,27 though age is not reported in all included studies.

Only one trial did not report the proportion between men and women,³⁴ and one study included only men.³³ The total was 344 men and 132 women; the relation was 2.6:1.

60% of the trials included patients with AIS (American Spinal Injury Association Impairment Scale) C or D^{20,21,24–29,31} only one trial reported patients with AIS A, B, C or D.32 Three trials included only AIS D participants^22,30,33 and two trials did not report the AIS.^23,34 The total reported was: 140 AIS D, 61 C, two B and two A.

The aetiology is reported only in five trials.^{22,24,25,27,29} A total of 142 participants were reported with traumatic SCI and 93 participants were non-traumatic.

The most prevalent level of injury was cervical with 155 participants^{22,24,25,27,29–32}; 37 for the group of T1-T6^{24,25,29–31}; 44 participants with a T7-T11 level^{24,25,27,29–32}; and 53 below T12, 53.^25,31,32 22 participants were reported including thoracic and lumbar levels without separating groups.²²

Dose of intervention

The period of treatment was one day^32,33; three weeks²¹; four weeks^22,23,34; eight weeks^24,25,27,30; and 12 weeks.^{20,26,28,29,31} The frequency was reported from three times per week during four weeks,^22,34 up to five times per week for 12 weeks.^20,26,28 The RAGT setup was initially prescribed for the amount of body weight supported at 60%^25,27 and never less than 25%.^25,27 The guidance force was set from 100%^22,26,28 to 20%.^21,23 The lowest initial speed was reported at 1.0 Km/h²⁹ and in one trial²³ the participants accomplished 3.4 Km/h. The length of the RAGT therapy varied from 20 minutes,²¹ to 45 minutes.^28–30

Outcomes measures for analysis

The many diverse outcome measures recorded in included studies made it impossible for authors to analyse all of the documented data. Based on the pre stated relevant outcomes and the availability of data from specific measures in included trials, the analysis focused on speed (m/s), WISCI, strength (LEMS) and FIM-L.

Of the 15 RCT, 10 studies included outcome measures suitable for inclusion in the analysis. As measures were introduced, it was decided to use only 6 of the studies^{22,24,25,27,28,33} due to the different reasons listed below.

Studies included in meta-analysis comparisons

All trials involved a comparison between a RAGT and different therapeutic interventions: conventional therapy,^22,25,27 no intervention,²³ strength training.²⁴ Two trials compared resisted RAGT vs. conventional RAGT,^29,30 and one trial compared RAGT with acoustic feedback vs. conventional RAGT.²¹ Four of the studies divided the intervention in four groups; treadmill-based training with manual assistance, treadmill-based training with electrical stimulation, overground training with electrical stimulation, treadmill-based training with robotic assistance.^20,26,28,31 One study compared two groups: RAGT and Ergo_bike.³³ Another trial made 3 comparison groups: RAGT, tizanidine and no intervention.³⁴ Ultimately a trial did only one training session and the participants were randomized to four different therapeutic modalities.³²

Risk of bias

Risk of bias is summarized in (Figs. 2 and 3). The kappa values for inter-rater variability are >0.6 for most of the items.

Figure 2.

Risk of bias graph: review authors' judgments about each risk of bias item presented as percentages across all included studies

Figure 3.

Risk of bias summary: review authors' judgments about each risk of bias for each included study

Five SR were assessed using AMSTAR. All the RCT of the five SR,^1,10,14–16 were already included in our review, the studies that were not RCT or were protocols were excluded from our review.

The conclusion for the five SR is that there are not enough available studies and that the quality of existing studies is not ideal to determine the superiority of a therapeutic modality. Our review assesses 11 more trials in comparison with the latest one from Mehrholz et al. 2012, and a total of six RCT were used for the meta-analysis, of other outcomes rather than speed.

Effects of interventions

Results are described below under the comparisons carried out for each of the explored outcomes (1. Speed, m/s; 2. Strength (LEMS); 3. WISCI; 4. FIM-L).

Speed

Nine studies were included in this analysis but data for only five studies were pooled. Four studies were eliminated: two trials compared two different modalities of RAGT ²¹ and,²⁹ other trial had subgroups and it was not possible to define the number of participants for each intervention group in the subgroup of maximal speed,²⁶ and final mean was not reported in.²³ The remaining five studies showed no effect compared with control groups, showing a MD of -0.00 (95% CI -0.05 to 0.04, P = 0.95) (Fig. 4).

Figure 4.

Effect on Speed

A total of 169 participants were included in the analysis of this outcome,²⁵ and²⁷ had the greater number of participants (54.4%), both of them with an effect size that favours RAGT therapy. The only study that reported an effect that favours the control group is.²⁸

Strength (LEMS)

Five studies were pooled in this analysis (217 participants). The MD is 1.95 (95% CI -1.58 to 5.48, P = 0.28) in favour of the RAGT (Fig. 5). The first 2 listed studies²⁵ and²⁷ correspond to 40.3% of the total number of participants, both trials show a clear positive effect towards the RAGT. There is only one study that favours the control group over the use of RAGT.²⁸

Figure 5.

Effect on LEMS

WISCI

Data from four studies,^22,24,25,27 (188 participants) were pooled in this analysis. The pooled MD is 3.01 (95% CI -0.54 to 6.55, P = 0.10) in favour of the RAGT (Fig. 6).

Figure 6.

Effect on WISCI

FIM-L

Data from two studies,²⁵ and²⁷ were pooled (117 participants) in the analysis. The MD 2.74 (95% CI -1.83 to 3.66, P = 0.00001), with a clear effect in favour of the RAGT (Fig. 7).

Figure 7.

Effect on FIM-L

Discussion

This review included 15 RCT and five SR that explored the effects of RAGT compared with different physical rehabilitation approaches. Data from six studies were available for meta-analysis.

Very good–quality evidence for three of the RCT analysed in the meta-analysis^24,25,27 and moderate-quality evidence on the rest of studies analysed, showed a moderate effect of the use of RAGT for strength (LEMS), and a large effect for gait performance (WISCI) and functioning (FIM-L).

Improvement in gait was evaluated through the scores in speed and WISCI. The pooled results of 169 individuals studied for speed and 188 individuals studied for WISCI, showed no benefit in terms of speed with a MD of -0.00 (95% CI -0.05 to 0.04), but a large effect for the WISCI measurement showing a MD of 3.01 (95% CI -0.54 to 6.55). In accordance to the beneficial effect in the WISCI, the pooled results showed a large effect in terms of functioning measured by the FIM-L.

Walking speed is a very specific measure of walking capacity, while WISCI assesses physical limitation for walking secondary to impairment based on the use of assistive devices and physical assistance. Even when WISCI does not evaluate gait velocity, a medium correlation with walking speed has been reported.⁸ Walking speed has shown a medium correlation with lower limb’s muscle force, individual’s global independence, use of walking aids and gait performance.⁸ There are slight differences between both measures. Walking speed has shown higher correlation with spatial measures of gait such as step length, while WISCI-II has shown higher correlation with measures of gait symmetry such as difference between step duration of both legs.⁸ It is well known that step length and cadence are accommodated in response to gait velocity.⁴¹

Consequently, the different findings in terms of speed and WISCI may reflect the nature of RAGT. RAGT effect seems to be more general, addressing gait components related to use of assistive devices and gait symmetry. This needs confirmation by inclusion of specific and sensitive instrumented evaluations within clinical trials.

On the other hand, in order to increase walking speed, subjects use different muscular coordination patterns which are believed to be composed of combinations of simple neural control strategies for the co-excitation of multiple muscles, which are called motor modules.^42,43 Therefore, the variations on the motor modules used by the subject will result on variations on kinetics, kinematics and spatiotemporal parameters of walking.⁴⁴ In addition, biomechanical characteristics of human body determine feasibility of movements. A limitation on body biomechanics, such as an exacerbated increment on joint stiffness, could make impossible to take advantage of some motor modules, limiting their usefulness. Motor modules for walking at different speeds have been reported elsewhere.⁴⁵ There are reports about reduced number and composition of motor modules used by incomplete SCI while walking.^46,47 Therefore walking at faster velocities could be necessary to acquire motor modules necessary to improve walking speed.⁴⁸ Moreover, results should be analysed considering biomechanical and neurological characteristics of SCI subjects.

Finally, results showed a moderate benefit in the improvement of strength, favouring the RAGT intervention. Reduction of guiding force as training progresses could be responsible for this result. As the guidance force diminishes, the subject must improve its lower limbs’ strength in order to perform the right movements. However, this improvement of strength does not transfer to gait speed because no RAGT trains ankle plantar flexors. Biomechanically, gait velocity is determined by the conversion of potential energy of the centre of mass (CoM) into kinetic energy and calf muscles determine this amount of energy. Ankle joint is also critical for propulsion, shock absorption, and balance during walking.⁴⁹ Also, an improvement exclusively on body biomechanics does not result automatically on an improvement in motor performance if the neurological system cannot take advantage of it.⁵⁰

RAGT devices were designed to provide appropriate sensory information to the spinal cord by imposing a normal walking pattern in order to evoke locomotor activity. This relies on the hypothesis of the existence of central pattern generators (CPG) in humans which are neural networks within the spinal cord that generate basic rhythmical motor patterns involved in walking.^51–55 However it has also been shown that supraspinal pathway plays an important role in inter-limb coordination.⁴⁹ Miyai and colleagues⁵⁶ showed that medial sensorimotor cortices and the supplementary motor cortical areas were involved in the control of walking. Until now three types of robotic-assisted device have been developed: exoskeleton, end-effector and portable robotic exoskeletons. Up to date only RCT involving Lokomat have been performed, and no RCT of the portable robotic exoskeletons or the end-effector type. Even though we cannot generalize our conclusions to all the RAGT systems, we believe they could be applicable due to the fact that all of them have similar design, performance and therapeutic philosophy. All RAGTs control lower limb motion according to predefined joint trajectories and perform therapy at low speed (< 3.6 km/h) allowing adjustment on assistance level based on the patient's ability to step. G-EO system is an end effector robot, that can simulate other movements such as stair climbing at the expense of losing control over every lower limb joints trajectory.

Improving ankle plantar flexor recruitment is a critical component for rehabilitation. A RAGT should target plantar flexor strength and motor modules used for walking propulsion. For the design of training protocols for RAGT, it is crucial to understand how training affects patterns of muscle activity and the influence of assistance and training speed.⁵⁷ Generalization of motor learning can be sensitive to speed.⁵⁸ Fast walking can help improve motor function. It can be a promoter of motor plasticity⁴⁸ and encourage motor exploration by requiring participants to walk at more challenging speeds allowing greater practice (more steps). It also emphasizes subcomponents of walking such as propulsion in order to allow reconstruction of motor modules.⁴⁵

In addition to possible intrinsic limitations of the study methods, it is important to consider that the trials included in the review had considerable heterogeneity in terms of trial design, characteristics of the interventions and participant’s characteristics. Similarly, there are differences in treatment dosage and training parameters for each study.

There is an unclear risk of bias for allocation concealment and blinding in the majority of the studies, however, as blinding in such intervention studies is difficult, it was considered as low risk if the author mentioned that this did not influence on the obtained results.

Limitations of the review

The risk of publication bias exists in all SR. In an attempt to minimize this situation, an extensive search was made in some of the most important databases for the theme, the authors of the studies were contacted in aim to complete the missing data, and a search was made on the references of the studies, to find possible titles with our inclusion criteria. However, it remains possible that “grey literature” may have not been identified; nevertheless it is considered that this would not have a significant impact on the results.

Our study results may be limited due to the heterogeneity of the people with spinal cord injury studied, as pooling all the populations together may lead to missing the subgroup with the major benefit from this type of intervention, but there were insufficient trials and participants to conduct a subgroup analyses.

Conclusions

Results show that gait training in a robotic orthosis have positive effects in terms of improvements in performance of gait, strength and functioning, but no effect on speed, which is expected for all the reasons listed above. However, these results must be considered in light of the SR limitations.

In terms of the availability of studies that deal with the measurement of RAGT effect in people with spinal cord injury, it was found that studies are limited in number and heterogeneous in terms of treatment dosage and training parameters, with small sample sizes and lack of quality in their methodological designs. For these reasons, there is a great need to carry out larger sample multicentre randomized controlled trials that evaluate different locomotor training approaches, specify different subgroups and include specific and objective outcomes that assess functioning and performance rather than speed.

Acknowledgements

We thank the entire team at the Neurological Rehabilitation Department from the National Institute of Rehabilitation (INR) and the entire health personnel at this National Institute for their support.

Disclaimer statements

Contributors None.

Funding None.

Conflict of Interest Statement The Authors declare that there is no conflict of interest.

Ethics approval None.

ORCID

Aberto Isaac Pérez-Sanpablohttp://orcid.org/0000-0003-0550-928X