Robot-assisted rehabilitation for total knee or hip replacement surgery patients: A systematic review and meta-analysis.

BACKGROUND: This study was performed to update the current evidence and evaluate the effects of robot-assisted rehabilitation (RAR) in comparison with conventional rehabilitation (CR) in patients following total knee (TKR) or hip replacements (THR).
METHODS: PubMed Central, OVID Medline, Cochrane Collaboration Library, and EMBASE for a comprehensive search for all relevant studies, from database inception to July 2022. The following inclusion criteria were used to determine eligibility for studies: randomized and matched controlled trials recruiting men and women who underwent TKR and THR; and studies examining the effect of RAR on outcome measures of physical function and pain.
RESULTS: A total of 9 studies (230 patients) were included in this review and 4 were included in the meta-analysis. The meta-analysis of 2 studies showed that Hybrid Assistive Limb (HAL) training for 5 days, significantly improved pain measured on a visual analogue scale, compared to CR in patients following TKR (SMD = 1.05, 95% confidence interval [Cl] 0.39-1.71). Heterogeneity for I2 value was lower than moderate (tau^2 = 0.0121; I2 = 5%; P = .30). There were 2 studies that assessed self-selected walking speed. The meta-analysis of these studies showed that HAL training was significantly superior to CR in patients following TKR (SMD = 48.70, 95% Cl -50.53 to 147.94) at 2 months. A high heterogeneity was detected (P < .01; I2 = 97%).
CONCLUSION: The result of this systematic review and meta-analysis suggest that RAR may be an effective treatment in TKR and THR patients. However, high-quality studies are needed to verify the long-term effect on their recovery.

1. Introduction

Total knee replacement (TKR) and total hip replacements (THR) are common surgical procedures for patients suffering from end-stage arthritis, with an accelerating number of TKR and THR performed in recent years.[ These procedures are reportedly effective in alleviating pain, enabling functional recovery, and improving quality of life for patients.[ Considering the economic cost of these procedures, improving the efficiency of management models for arthroplasty patients is a matter of considerable policy interest. Rehabilitation, as a multidisciplinary approach, can perform the majority of required management for TKR and THR patients.[ However, previous studies reported various barrier to using rehabilitation services, including structural, personal, financial, and social, which limit access.[

Although traditional postoperative rehabilitation for TKR and THR patients improves patient’s gait effectively, it is time- and energy-consuming. It is also difficult to maintain consistency of training intensity among different patients. Rehabilitation using innovative technologies – such as eHealth, telemedicine, wearables, virtual reality, and online education tools – are being applied on a trial basis in clinical trials.[ Among these technologies, a type of wearable consisting of a robotic support structure is being tested. Robotic devices are suited to make intensive, task-oriented motor training for moving the patient’s limbs, supervised by physical therapists, with a promising approach to rehabilitation and reducing the burden on caregivers.[ Lower limbs robot-assisted training has become of particularly interest as it can effectively resolve the problems traditional rehabilitation currently faces. Most studies using robot-assisted training have focused on central nervous system diseases, such as stroke,[ spinal cord injury,[ and cerebral palsy.[ The study by Carda et al[ showed a progression of the improvements up to 6 months in patients with early-stage Parkinson’s disease through robotic gait training. Like this, development of robotic technologies to restore the functionality of impaired neurological circuits has extended to the neural rehabilitation field for restoration of sensorimotor control and functions.[ Recently, a growing body of literature supports the use of robotic devices for improving the pain and function of patients following TKR and THR.[ Several forms of robotic devices are commercially available at this time, and are typically classified as either exoskeletons or end-effector robots, depending on the motion they apply.[ In all of the studies in which patients undergoing TKR or THR were followed, exoskeletons not only moved joints such as knees and hips, in coordination with phases of gait but also increased the range of motion of joints. Robot-assisted rehabilitation (RAR) is used either in combination with conventional rehabilitation (CR) or alone and is believed to have satisfied the necessary external requirements for improving motor function.[ In the meantime, the quantification of gait training intensity is needed to allow the standardization of gait parameters, because the applied robots are merely tools in the hands of physical therapists and therapists themselves, ultimately provide unique benefits to the treatment protocols in patients following TKR and THR. Indeed, outcome measurements for evaluating the functional recovery of TKR and THR were all different.[ Consequently, the quantification and standardization of RAR through review of existing studies is clearly needed.

A small number of systematic reviews of rehabilitation using any technology-based interventions, have been conducted using a variety of different technologies such as education, monitoring or treatment delivering via telecommunication technologies, internet, software or virtual reality devices.[ However, currently no study has systematically analyzed the effects of RAR for TKR and THR. Therefore, the aim of the present systematic review and meta-analysis is to update the current evidence and evaluate the effectiveness of RAR in comparison with CR in patients following TKR and THR.

2. Methods

2.1. Design and registration

This systematic review and meta-analysis was done according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.[ It has been registered in the PROSPERO database, an international prospective register of systematic reviews in health and social care (National Institute for Health Research, CRD 42020218723).

2.2. Literature search

We used PubMed Central, OVID Medline, Cochrane Collaboration Library, and EMBASE for a comprehensive search for all relevant studies, from database inception to July 2022. We used the following search terms: ((rehabilitation) AND hybrid) AND total hip arthroplasty)) OR (((rehabilitation) AND total knee arthroplasty) AND hybrid)) OR (((hybrid) AND total knee arthroplasty) AND robot rehabilitation)) OR ((hybrid) AND robot rehabilitation) (Supplemental File 1, http://links.lww.com/MD/H557). We also did a manual search of possibly related references. Two researchers (CH and JI) reviewed the titles, abstracts, and full texts of all potentially relevant studies independently, as recommended by the Cochrane Collaboration.[ Any disagreement was resolved by the third reviewer. We reviewed full-texts of the screened articles and then selected eligible articles. The reviewers were not blinded to authors or institutions.

2.3. Inclusion and exclusion criteria

Studies were selected by means of the following criteria: study design: prospective comparative studies, randomized controlled studies (RCTs); study population: patients who underwent TKA or THR; intervention: RAR with CR, and outcomes: demographic factors, clinical results during postoperative follow-up. Studies, which did not meet above criteria, posters, letters, and review articles were excluded.

2.4. Assessment of risk of bias

Two authors (CH and JI) independently evaluated the risk of bias of each study. In RCTs and crossover study designs, selection, performance, detection, attrition, and reporting biases were assessed using the Cochrane Risk-of-Bias Tool.[ For assessing the risk of bias in non-RCTs, we used the Newcastle-Ottawa scale for case-control studies.[ Studies were considered to have a low risk of bias when the quality assessment was 50% of, or above the checklist criteria. A funnel plot was used for estimating and adjusting publication bias.[

2.5. Study selection and data extraction

Data extraction was independently performed by 2 investigators (C.H. and J.I.) using a structured table that included the study characteristics (author, date of publication, study design, and country, participants characteristics (sample size, mean age, and numbers of male and female participants), diagnosis and follow-up period, specific interventions; rehabilitation using robot or Hybrid Assistive Limb (HAL), and primary outcome measures. The primary outcomes pooled in this analysis included visual analogue scale (VAS) score and self-selected walking speed (SWS).

2.6. Data synthesis and statistical analysis

In the studies by Kotani et al[ and Goto et al,[ the change in VAS was evaluated using meta-analysis (Table 1). We aggregated the data obtained from 2 studies reporting the effects of HAL in comparison to CR in the form of mean (standard deviation) to produce an overall mean effect (standardized mean effect [SMD]). VAS score was summarized as mean plus/minus standard deviation, when provided or calculated. Also the change in SWS was evaluated using meta-analysis in the studies by Yoshikawa et al[ and Tanaka et al.[ These 2 studies had different units, m/s[ and rate of change,[ so we unified them to rate of change. We used a pooled meta-analysis to analyze the dislocation rate. For dichotomous results, we calculated the risk ratio (RR) and 95% confidence interval (CI). For continuous outcomes, SMD and 95% CI were calculated.[ The heterogeneity of each study was estimated using Higgins I2 statistics and the Chi square test.[ Studies with a P < .10 and I2 > 50% were considered as having heterogeneity. We employed a fixed-effect model in studies without heterogeneity and a random-effects model in studies with heterogeneity. The trim and fill method was used to estimate and adjust the number and outcomes of missing studies in the meta-analysis.[ Statistical analysis was done using R software 3.02 (R Foundation for Statistical Computing, Vienna, Austria) and the meaningful value was set to P < .05.

We had no choice but to descriptive analysis, as most of studies had various outcomes that did not overlap. We not only analyzed the secondary outcome measures for studies following TKR and THR separately, after categorizing according to the interventions used in included studies, but also classified and re-analyzed these secondary outcome measures according to the purpose of the assessment. The secondary outcome measures could be divided into 3 groups. Firstly, outcome measures related to walking ability were as follows: maximum walking speed (MWS), step length in the SWS and MWS, cadence at SWS and MWS, cadence, single/double support time, velocity, stride length, anterior/posterior/lateral variability, and joint moment. Secondly, extension lag, quadriceps strength during knee extension, hamstring muscle strength during knee flexion, knee range of motion (ROM), Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), knee kinesthesia grade, knee proprioception grade, functional ambulation category, Berg balance score, 10-m sitting-standing time, 6-minute walking distance, modified functional reach test, manual muscle test, 10 meter walk test, timed up and go test, strength (hip flexion/extension), and hip ROM were used in the assessments related to knee or hip function. Lastly, Harris hip score, hospital for special surgery, 36-Item Short Form Health Survey score were used in the assessments for clinical evaluation.

Seven studies[ of 7 investigated rehabilitation with HAL. Another study[ showed the efficacy of the robot-assisted walking training (AVATAR-M; Shanghai Zhanghe Corporation, China) following TKR, and the remaining study[ was performed by using trunk control rehabilitation robot training (TCRRT, 3DBT-33, Man and tel, Gumi, Korea) following THR.

In studies using HAL, 6[ included patients having post-TKR rehabilitation and one[ included patients undergoing post-THR rehabilitation. First of all, among 6[ studies, the outcome measures were analyzed descriptively (with exception of VAS) to evaluate the knee function in the studies by Kotani et al,[ Goto et al.[ Improvement in active and passive knee ROM and strength of knee flexion were assessed for the evaluation of function in the study by Kotani et al.[ Goto et al[ reported the effectiveness of HAL assisted rehabilitation through the extension lag. VAS and extension lag were analyzed descriptively as the specific data of VAS was not reported in the study by Yoshioka et al.[ In the other 2 studies, Yoshikawa et al[ used the following outcome measures: muscle strength of knee extension and flexion, the knee ROM, and WOMAC to evaluate the knee function at the pre-surgery, and at weeks 1, 2, 3, 4, and 8 after training. As well, MWS, step length in the SWS and MWS, cadence at SWS and MWS were used to assess the walking ability. The quadriceps strength for knee function and NRS were analyzed prior to surgery and on postoperative weeks 1, 2, and 3 in remaining study by Tanaka colleagues.[ Fukaya et al[ performed the gait analysis to verify the effect of HAL on the kinematic and kinetic variables of lower limb joint function at 5 weeks after TKA. In studies using HAL, one[ included patients undergoing post-THR rehabilitation. Gait analysis was performed before and after surgery, and comparisons were made between the 2 groups. This study used several outcome measures. They consisted of waling ability parameters such as single support time, double support time, cadence, velocity, stride length, and anterior/posterior/lateral variability as well as hip function through hip and knee ROM and clinical assessments such as Harris hip score, hospital for special surgery, and 36-Item Short Form Health Survey score.

The study by Li et al[ were also analyzed descriptively, because the intervention program and outcome measures were completely different when compared to other studies which were conducted in the patients following TKR. Li colleagues[ used AVATAR-M that could help the patients take up early rehabilitation, while the other studies were performed in patients following TKR through HAL assisted training. In addition to the hospital for special surgery score for clinical evaluation, assessment related to knee function such as knee kinesthesia grade, knee proprioception grade, functional ambulation category, Berg balance score, 10-m sitting-standing time, and 6-minute walking distance were examined, and the effects of robot-assisted walking training were compared with those of CR.

Lastly, one study[ was performed by using TCRRT following THR. The modified functional reach test, manual muscle test of lower extremity, 10 meter walk test, timed up and go test, and strength of hip and knee as well as VAS were analyzed to evaluate of function and pain.

3. Results

3.1. Identification of eligible studies

The initial search identified 372 articles from the selected databases. After screening titles and abstracts for duplicates, unrelated articles, and case reports, 338 articles were excluded. The remaining 34 studies underwent full-text review and 25 studies among them were excluded. This left 9 studies, which were included in the final analysis (Fig. 1). There were 4 RCTs and 5 prospective comparative studies.

Figure 1.

PRISMA flow of information through the different phases of meta-analysis. PRISMA = preferred reporting items for systematic reviews and meta-analyses.

3.2. Study characteristics

Table 1 summarizes the characteristics of study participants. These 9 studies included 230 patients (116 in the experimental group and 114 in the control group). Among them, 4 studies[ were RCTs with 133 patients and the other 5 were prospective comparative studies with 97 patients.

Table 1

Characteristics of the included studies.

Study	Participant characteristics				Intervention program
Reference/country	N (e.g./CG)	Female, n (%) (e.g./CG)	Disease	Mean age  ± SD (yr) (e.g./CG)	EG	CG	Length of intervention	Outcomes	Time point
Total knee replacement
Yoshikawa[14] JapanKotani[15] JapanGoto[16] JapanLi[18] ChinaTanaka[19] JapanYoshioka[21] JapanFukaya[22] Japan	9/10	8 (88.9)/8 (80)	OA and RA	61.2 ± 8.9/59.1 ± 9.8	Training with HAL + CR	CR	e.g.:10 to 12 times HAL training (14.4 min/time) over 4 weeks + 60 to 80 min a day of CR.CG: 60 to 120 min a day of CRTotal times were 26.5 (e.g.) hours and 28.2 hours (CG) respectively.	SWS, MWS, SL-SWS, SL-MWS, cadence at SWS, cadence at MWS, strength of knee extension and flexion, knee ROM, WOMAC-P, WOMAC-F	Pre-surgery, 1, 2, 3, 4, 8 weeks
	12/10	10 (83.3)/8 (80)	OA	77.3 ± 3.8/75.0 ± 5.1	Use of HAL-SJ in knee flexion + CR	CR	e.g.: 50 active knee flexion at each session every other day using HAL-SJ (starting on POD 5) + CR (starting on POD 1). 5 to 10 min of average time over 5 days trainingCG: 50 active knee flexion at each session every other day (starting on POD 5). + CR (starting on POD 1). 10 to 15 min of average time over 5 days training	VAS, Strength of knee flexion, knee ROM	Pre-intervention, 5 days
	10/10	10 (100)/8 (80)	NR	74.2 ± 7.5/73.2 ± 8.0	Use of HAL-SJ in active assistive knee exercise + CR	CR	e.g.: 50 active assistive knee exercise at each session (HAL-SJ applied every other day, starting on POD 5) + CR (starting on POD 1) over 5 days trainingCG: CR (starting on POD 1) over 5 days training	VAS, extension lag	Post-surgery 5 and 10 days
	30/30	41 (68.3)	OA	NR	Robot-assisted walking training + CR	CR	e.g.: 5 times a week, 10 times in total (30 min each and twice a day, starting POD 7) over 2-weeks trainingCG: CPM training, peri-knee neuromuscular, electrical stimulation, isometric contraction of peri-knee muscle (starting POD 7) over 2-weeks training	HSS, knee kinesthesia grade, knee proprioception grade, FAC, BBS, 10-m sitting-standing time, 6m walking distance	Pre-intervention, 1, 2 weeks, 1, 3, 6, 12 months
	13/1312/129/9	10 (76.9)/10 (76.9)8 (66.7)/11 (91.7)8 (88.9)/8 (88.9)	OAOAOA and RA	75.5 ± 5.4/75.6 ± 4.571.3 ± 6.2/74.9 ± 8.774.1 ± 5.7/76.4 ± 7.6	Training with HALUse of HAL-SJ in knee exercise and flexion + CRTraining with HAL	CRCRCR	e.g.: 5 times a week, 10 times in total (HAL training for 40 min) over 2-weeks trainingCG: 5 times a week, 10 times in total (CR for 40 min once a day) over 2-weeks traininge.g.: knee exercise with HAL-SJ (10 extensions per set, 5 sets, twice per week, starting on POD 8) + CR (starting on POD 1) over 5 days per week training until dischargeCG: CR (starting on POD 1) over 5 days per week training until dischargee.g.: 10 to 12 times in total over 4-weeks training (starting on POD 1 to 5 weeks)CG: no significant difference in the total physical therapy time compared to the e.g.	SWS, strength of knee extension, NRSVAS, extension lagStep length, hip ROM, knee ROM, ankle ROM, joint moment	Pre-surgery,1, 2, 3 weeksPre-intervention, 8,10, 15 days,Post-surgery 5 weeks
Total hip replacemet
Yang[17] South korea	13/12	11 (84.6)/10 (83.3)	Femur neck and introchanteric fracture	79.3 ± 6.7/78.5 ± 5.4	TCRRT  + CR	CR	e.g.: 1 time a day (TCRRT for 15 min + CR for 20 min, total 35 min, starting POD 4) over 15 daysCG: 1 time a day (CR for 35 min, starting POD 4) over 15 days	MFRT, 10MWT, TUG, VAS, Strength (hip flexion/extension, knee flexion/extension)	Pre-intervention, 5, 10, 15 days
Setoguchi[20] Japan	8/8	6 (75)/6 (75)	OA	74 ± 4.4/63.8 ± 7.7	Gait training with HAL	CR	e.g.: gait training with HAL (40 min/time, 3 times a week, 6 times in total) (On the day without HAL, 40 min/time, CR once/day) over 2 weeksCG: CR1 time a day, 40 min/1 time over 2 weeks	HHS, SF-36 score, single support time, double support time, cadence, velocity, stride length, anterior/posterior/lateral variability, hip ROM, knee ROM	Pre-surgery,1, 3 weeks

10MWT = 10 meter walk test, BBS = Berg balance score, CG = control group, CR = conventional rehabilitation, FAC = functional ambulation category, HAL = hybrid assistive limb, HAL-SJ = single-joint hybrid assistive limb, HHS = Harris hip score, HSS = hospital for special surgery, e.g. = experimental group, MFRT = modified functional reach test, MWS = maximum walking speed, NR = not reported, NRS = numerical rating scale, OA = osteoarthritis, POD = postoperative day, RA = rheumatoid arthritis, ROM = range of motion, SF-36 = score 36-Item Short Form Health Survey score, SL-MWS = step length-maximum walking speed, SL-SWS = step length- self-selected walking speed, SWS = self-selected walking speed, TCRRT = trunk control rehabilitation robot training, TUG = timed up and go test, WOMAC-F/P = Western Ontario and McMaster Universities Osteoarthritis Index-function/pain, VAS = visual analogue scale.

According to the type of robots used in 9 studies, 7 studies[ investigated rehabilitation with HAL. In these 7 studies, 6 studies[ included patients having post-TKR rehabilitation and one study[ included patients undergoing post-THR rehabilitation. Another study[ reported the efficacy of AVATAR-M following TKR and the remaining one study[ was conducted by using TCRRT following THR.

3.3. Meta-analysis of primary outcome measured

Our meta-analysis of 2 studies[ showed that HAL training significantly improved pain measured on an 0 to 10-point VAS, compared to CR in patients who underwent TKR (SMD = 1.05, 95% Cl 0.39–1.71) after 5 days of training. Heterogeneity for I2 value was lower than moderate (tau^2 = 0.0121; I2 = 5%; P = .30) (Fig. 2). There were 2 studies[ that assessed SWS. These studies showed that HAL training was significantly superior to CR in patients following TKR (SMD = 48.70, 95% Cl -50.53 to 147.94) at 8 weeks. A high heterogeneity was detected (P < .01; I2 = 97%) (Fig. 3).

Figure 2.

Forest plot of overall effects of robot-assisted rehabilitation on TKR based on VAS score for standardized mean difference. TKR = total knee replacement, VAS = visual analog scale.

Figure 3.

Forest plot of overall effects of robot-assisted rehabilitation on TKR based on SWS for standardized mean difference. SWS = self selected walking speed, TKR = total knee replacement.

3.4. Descriptive analysis of secondary outcome measured

3.4.1. Type of intervention technologies with their purpose of assessment

Hal

A total of 7 studies[ investigated rehabilitation with HAL. Among them, 6[ included patients undergoing post-TKR rehabilitation. The other[ included patients having post-THR rehabilitation. In patients following TKR, HAL combined with CR was conducted in the experimental group and only CR was applied in the control group in 3 of the studies, while in the studies conducted by Tanaka et al[ and Fukaya et al,[ training with HAL and CR were provided to experimental and control groups, respectively.

Walking ability

Yoshikawa et al[ reported that the treatment given to the experimental group was more effective than that given to control group in patients following TKR. The patients in the HAL group underwent HAL training (average 14.4 ± 5.9 minute a session) and CR (60–80 minute a day). Total number of HAL interventions ranged from 10 to 12 during the 4-week period. The patients in the control group underwent only CR (60–120 minute a day) after TKR. In addition to the SWS used in the meta-analysis, various outcome measurements were obtained. The SWS, step length in the SWS, and the MWS were greater in the HAL group than in the control group at 4 and 8 weeks (P < .05). The step length in the MWS was greater in the HAL group than in the control group at 2, 4, and 8 weeks (P < .05). HAL training was initiated 1 to 5 weeks after TKA in the HAL group and CR was performed in the control group in the study by Fukaya et al.[ In the HAL group, the odds ratio of hip extension was as large as 1.741, while that of knee swing was as large as 1.501 at post-surgery 5 weeks. These 2 variables were significant between the 2 groups. They suggest that HAL training increased the mobility of the knee and hip joints and that walking ability was improved by increasing the step length because knee swing and varus significantly affected step length.[

Knee function

In the study by Yoshikawa et al,[ the muscle strength of knee extension in the HAL group was greater than in the control group at 8 weeks (P < .05). Also the extension lag and knee pain (WOMAC-P) were lower in the HAL group than in the control group at 2 weeks (P < .05). In the study conducted by Tanaka et al,[ the outcome measures were evaluated prior to surgery and at weeks 1, 2, and 3 postoperatively. Rehabilitation using HAL resulted in a significantly better improvement in quadriceps strength compared to CR (weeks 1 and 2; weeks 1 and 3: P < .001). In the studies by Kotani et al[ and Goto et al,[ other outcome measures except VAS were analyzed descriptively to evaluate the function. RAR through HAL was performed to assist knee ROM exercise every other day in the experimental group in the single knee joint. Significant improvement occurred in both groups between postoperative days 5 and 10. Improvement in active knee ROM, passive knee ROM, and strength of knee flexion were significantly greater in the HAL assisted rehabilitation group than in the control group (P < .01).[ Goto et al[ reported that the HAL assisted rehabilitation group showed 89.4%±15.7% improvement (P < .01), while the control group showed 34.8% ± 32.1% improvement (P = .016) in the extension lag. Yoshioka et al[ also assessed extension lag to report the effectiveness of HAL. The extension lag significantly improved in the HAL group after the 2nd and 3rd sessions.

Hip function

Lastly, in the study by Setoguchi et al,[ HAL was used for training over 2 weeks (3 times a week) after THA. Gait analysis was performed before and after surgery, and comparisons were made between the 2 groups. Several outcome measures were used in this study, the extension angle in the sagittal full range of hip motion were better in the HAL group than control group at 3 weeks postoperatively.

AVATAR-M Knee function

The study by Li et al[ was also analyzed descriptively, because the intervention program and outcome measures were completely different from other studies and was conducted in patients following TKR. In it the authors used AVATAR-M over 2 weeks (5 times a week) to help the patients start rehabilitation early, while the other studies were performed in patients following TKR through HAL assisted training. Berg balance score, 10-m sitting-standing time, and 6-minute walking distance of RAR group were significantly higher than the control group (P < .05). Also, knee kinesthesia grade, knee proprioception grade, and functional ambulation category of experimental group were better than the control group although there was not significant (P > .05).

Clinical evaluation

The HSS score of RAR group were significantly higher than the control group (P < .05) in the study by Li et al.[

TCRRT Knee function

There was one study by Yang et al[ which was conducted in patients who had THR. The experimental group received the TCRRT for 15 minutes and CR for 20 minutes, while the control group received only the CR for 20 minutes once a day over 15 days. The experimental group had significantly more improvement in the modified functional reach test, manual muscle test of lower extremity, 10 meter walk test, and timed up and go test compared to the control group (P < .05).

3.5. Risk of bias within individual studies

Study quality assessment of RCTs by the Cochrane Risk-of-Bias Tool is summarized in Figure 4. Newcastle-Ottawa scale was used to assess the quality of the selected studies. All included studies scored 5–8 points, indicating relatively high quality.[

Figure 4.

Risk of bias assessment in the included studies.

3.6. Publication bias

Since the number of included studies were less than 10, funnel plot analysis was not performed.

4. Discussion

Our review focused on the effects of RAR following TKR or THR in patients with osteoarthritis, rheumatoid arthritis, or hip fracture. This is a unique review that considered the implication of RAR on several outcome measures.

Of the 9 studies included, only 5[ (4[ for TKR and 1[ for THR) analyzed the change of pain and the VAS and NRS were used to those. NRS is the simplest and most commonly used scale. The numerical scale is most commonly 0 to 10, with 0 being: no pain: and 10 being “the worst pain imaginable.”[ NRS could have been used as a substitute for VAS in the study by Tanaka colleagues.[ However, the NRS score was not summarized as mean plus/minus standard deviation, as instead, the change in the number of participants with high NRS scores (6 points or higher) was identified. Thus, the NRS score was not able to be included in the meta-analysis. Although the details were not described separately in the results, we were able to identify through descriptive analysis that rehabilitation using HAL resulted in a significantly greater improvement in NRS scores (week 1: P = .03) when compared to CR.[ In the study by Yoshioka et al,[ there was a reduction in the VAS for pain after training, which was not significant. Two[ of the remaining 3 studies evaluated pain through meta-analysis and the other studies[ using TCRRT was analyzed through descriptive analysis as it was performed in patients following THR. The TCRRT group had significantly more improvement in the VAS compared to that before intervention (P < .05) at the 5, 10, and 15 days.[ In sum, all 4 studies[ using HAL showed superior effect of RAR over those of the control group. The superior therapeutic effects of RAR in the reduction of pain can be attributed to the following reasons: in TKR patients, training with HAL improved active and passive ROMs. This is partly because knee movements are less painful with the HAL system than CR, as knee flexion is possible without excessive effort. Also, pain from the surgical incision and damage to the soft tissues inhibit normal hip joint movement in THR patients.[ These can induce emotional and psychological pain as well as chronic pain.[ Also, it is believed that TCRRT’s consistent repetitive muscle strengthening exercises in the lower extremities may have had a positive effect on pain reduction by increasing the ROM, reducing adhesion, and promoting muscle strengthening.

SWS, one of the principle findings of our meta-analysis, were improved more by RAR than CR at 8 weeks. The change in walking speed after TKA is also considered an indicator of the effectiveness of rehabilitation.[ In the study by Tanaka et al,[ the difference in change of walking speed between the 2 groups was significant during the early part of the postoperative recovery phase and this difference diminished over time. On the other hand, the walking speed in the RAR group was better than in the CR group at weeks 4 and 8 in the study by Yoshikawa et al.[ The difference in timing of walking improvements was likely due to differences in the experimental and conventional groups in terms of the length of time that training was applied. In the study by Tanaka et al,[ the experimental group using HAL performed one 40-minutes session once a day and also undertook a single 20-minutes rehabilitation session consisting of ROM exercises and walking. In the control group, CR was done 40 minutes once a day, and the same 20 minutes session of ROM exercises and walking as the experimental group was performed. The training time was 6 hours because both groups received 10 times in total. Meanwhile, total physical therapy time using HAL was 26.5 ± 4.2 and 28.2 ± 5.2 hours in the experimental and control group, respectively, in the study by Yoshikawa et al.[

For pain after total joint arthroplasty, Jonathan et al[ investigated the minimal clinically important difference (MCID) for VAS in patients following TKA. The authors reported that the MCID for pain improvement in TKA was 2.26 cm. The VAS is a horizontal line, 10cm in length, with 0 cm labeled “no pain” and 10 cm labeled “worst pain I have ever had.” Patients mark the point on the line based on their perception of their current state.[ In our meta-analysis, we found that the average change on the VAS was lower than MCID value for TKA in the experimental groups. When considering control groups, all control groups in the studies underwent CR. Therefore, the SMDs in 2 studies were small. The average change of the VAS was lower than MCID, may have been caused by effects of CR in 2 studies. The difference of SWS did not exceed the MCID in our review,[ which is also thought to be the result of CR used in all control groups, as was the case with VAS.

We showed that various outcomes that did not overlap were significantly improved in the experimental groups compared to the control groups through descriptive analysis. This review found sufficient evidence to support the conclusion that RAR had significant clinical effects for TKR and THA surgery patients.

Among the 9 studies of our review, HAL was used in the 7 studies.[ Six studies[ included patients having post-TKR rehabilitation. The other study[ included patients following THR. HAL is a wearable robot designed to facilitate movement and interactively provides motion according to the wearer’s voluntary drive.[ Still, the mechanism through which the use of HAL improves gait function after TKA and THA remains unclear. Movement of HAL is triggered by bioelectric signals from the muscles accompanying the exercise of intention, correctly supporting spontaneous movement of impaired limb and generating sensory feedback.[ In patients following TKA, quadriceps arthrogenic muscle inhibition may be one reason for the effect of rehabilitation and is the phenomenon of inhibition of the quadriceps femoris muscle after surgery.[ This inhibition is believed to result from pain or swelling of the knee, or damage to pressure receptors.[ The suggested mechanisms for quadriceps arthrogenic muscle inhibition include possible inhibition of α motor neurons via the spinal reflex[ and involvement of pathways engaging upper motor neurons.[ Using the HAL for post-THA therapy would improve hip extension and lead to a correct gait. This improvement could be attributed to assistance provided by HAL during the extension of the hip. The motion assist feature has the effect of motor learning, by feedback from HAL to the muscle, nerve, spinal cord, and brain.[ Several studies have shown the effect and feasibility of HAL-assisted rehabilitation and single joint HAL-assisted rehabilitation for stroke, spinal cord injury, and post-TKA knee extension failure.[

In 2 studies[ not using HAL, it was shown that AVATAR after TKR can provide patients with a safe and comfortable driving force to assist them in initiating walking training as early as possible and thus speed their walking recovery in a manner maximally close to a physiological condition as shown in the study by Li colleagues.[ As they reported, the periodical movement of the hip, knee, and ankle during a normal walking process was simulated in a partial weight-supported condition, and a tolerable and comfortable walking speed was adjusted. TCRRT was used in the experimental group following THR in the study by Yang and colleagues.[ TCRRT was feasible and effective for improving dynamic balance, lower extremity strength, gait ability, and pain after THR. It was believed that TCRRT’s consistent repetitive muscle strengthening exercise of lower extremities might have had a positive effect on pain reduction by increasing ROM, decreasing adhesion, and enhancement of muscle strength.

This review had some limitations. First, only 2 studies for each outcome measure were included in meta-analysis. We examined VAS and SWS to determine the effect of RAR in our meta-analysis. However, all of the included studies were prospective comparative studies, and we performed the quality assessment of the risk of bias to overcome this limitation. Second, most trials used various outcome measures which limited the pooling of results. Future RCTs should demonstrate the effect of RAR on consensus on a set of suitable outcome measures. Third, most studies (except the study by Tanaka et al[) did not perform an a priori sample size calculation, which can increase the risk of underpowered results. This has been caused by a lack of reports on the use of robot in rehabilitation after TKR and THR. Fourth, it could not be described the standardization of antalgic therapy because that there was no description of antalgic therapy in the papers included in this study. Lastly, our review cannot provide the long-term effects (more than 8 weeks) of RAR.

5. Conclusion

The result of this meta-analysis of primary outcome and descriptive analysis of secondary outcome suggest that RAR may be an effective treatment for each TKR and THR patients. However, high-quality studies are needed to verify the long-term effects of RAR for TKR and THR patients. Further, it is important to consider a careful stratification of patients’ characteristics such as cognitive function and sarcopenia.

Acknowledgments

We give our sincere thanks to Harrisco (translation company) for providing language help.

Authors’ contributions

Conceptualization: Yoo JI, Lee CH.

Data curation: Yoo JI, Oh MK.

Formal analysis: Yoo JI.

Investigation: Oh MK, Lee SU.

Methodology: Yoo JI.

Supervision: Oh MK.

Validation: Oh MK.

Writing-original draft: Yoo JI.

Writing-review & editing: Lee CH.