Effects of balance training on functional outcomes after total knee arthroplasty: a systematic review and meta-analysis of randomized controlled studies

Background: Many previous studies have compared the effects of preoperative balance training and non-balance training on daily performance and knee functional outcomes after surgery, however, whether preoperative balance training is more beneficial to patients is still a big debate. Comparing the postoperative joint and daily function of balance training and non-balance training is the main purpose of our study. Methods: Cochrane library, Pubmed and Web of Science databases searched by us, and searched again before submitting our submission. This mete-analysis included 22 studies that directly compared postoperative performance and functional outcomes after training group (TG) and control group (CG). We used the software endnote X9 for data selection, and the software Review Manager 5.3 for data analysis to make funnel plots and forest plots. Results: The pooled data indicated that balance training significantly improved 2/6 MWT (2/6-min walk test) (weighted mean difference (WMD) = 25.17, 95% confidence interval (CI) 12.88 to 37.46, P < 0.0001), gait speed (WMD = 0.15, 95% CI 0.10 to 0.19, P < 0.00001), TUG (timed up and go) (WMD = 1.02, 95% CI 0.75 to 1.29, P < 0.00001), BBS (berg balance scale) (WMD = 1.79, 95% CI 0.50 to 3.08, P = 0.006), FHR (ratio of functional reach distance to body height) (WMD = 9.34, 95% CI 6.69 to 11.98, P < 0.00001), TCS (WMD = 1.20, 95% CI 0.86 to 1.53, P < 0.00001), early stage vitality (WMD = 14.41, 95% CI 13.53 to 15.30, P < 0.00001) and KOOS (Knee Injury and Osteoarthritis Score) sysptoms (WMD = 6.34, 95% CI 2.07 to 10.60, P = 0.004), middle stage function (WMD = 5.85, 95% CI 0.13 to


Background
The number of patients requiring total knee arthroplasty (TKA) for osteoarthritis of the knee continues to increase as the population ages. [1] This is mainly because TKA can improve joint mechanics, reduce pain, increase angle of joint motion, and improve function and balance. [2][3][4] However, studies show that about one in five patients still have chronic knee pain after surgery, and in 15% of the patients, no cause was found in the radiographic images of the joint. [5,6] Although total knee arthroplasty largely alleviates the adverse symptoms of the affected limb, patients may still have balance disorders, [7,8] long-distance walking and difficulty in climbing stairs for several months after the operation, [9,10] indicating that patients may not necessarily achieve the expected physiological function results after TKA.
Most patients were satisfied with total knee arthroplasty, [11] however, there were still some patients with persistent sensorimotor dysfunction. [12] Incomplete proprioceptive recovery of the knee joint, poor mobility and unstable dynamic balance are factors that increase the risk of falls in the elderly, and are also common factors that lead to difficulty in completing daily activities. [13,14] It has been reported that 25% of patients will fall within 2 years after total hip or total knee arthroplasty, [15] resulting in physical and psychological trauma and a high cost to the patient. [16] Therefore, limb balance is particularly important for the basic daily activities and safety of patients with total knee arthroplasty. Balance training is proprioceptive or neuromuscular training, as well as overall sensorimotor training, [17] which aims to control pain, increase range of motion, restore deteriorating muscle strength, and improve functional performance and balance.
Recent studies have found that postoperatie balance training appears to be effective in improving early stage postoperative outcomes [18] and supporting the use of balance training as a rehabilitation program during total knee arthroplasty. [18,19] However, there were few studies on preoperative intervention, and the evidence of results were uncertain, so the effectiveness of preoperative balance training remains to be determined. The purpose of this study was to evaluate the effects of preoperative balance training on early and middle stage postoperative functional outcomes in TKA patients.

Search Strategy
We conducted this study based on the reporting project of the systematic review and meta-analysis (PRISMA) [20] . The research plan of this systematic review and meta-analysis was determined by all co-author before the start of the document retrieval, research solutions have been posted online in PROSPERO International Prospective Register of Systematic Reviews.
Two authors searched for studies in MEDLINE (through PubMed), SCI (through Web of Science) and the CENTRAL (Cochrane Central Register of Controlled Trials, through the Cochrane-Library), without language restrictions. The time was set at nearly twenty years.
Before submitting the manuscript, we searched again and found no new literature.

Inclusion/Exclusion Criteria
The search results in this systematic review were assessed independently by scanning the title/abstract or the full text. Any dispute between the two authors was settled by the third author to reach the final consensus. 16. ROM (range of motion)).

Statistical analysis
Since there are no binary variables in the included literatures, odds ratio (OR) and 95% confidence interval (CI) are not required to estimate each study. Continuous variables were assessed using mean difference (MD) or standard mean difference (SMD) to assess the data included in the literature. For studies that present continuous data, such as mean and range values, statistical algorithms were used to estimate standard deviations(SD). [21] Only studies involving standard deviation and mean values can be included in the analysis. Heterogeneity was represented by P value and I 2 . When P ≤ 0.1 or I 2 ≥ 50%, the random effect model was used instead of the fixed effect model for heterogeneity test.

Study Selection
We extracted 794 potentially relevant articles from 3 electronic databases (Pubmed, Cochrane library and Web of Science). By browsing the references in the literatures, we included 5 additional RCTs. After deleting the duplication and reviewing the titles and abstracts, 335 irrelevant references were excluded. We perused the remaining 43 articles, 4 of which were systematic reviews. We excluded another 17 articles for the following reasons: Not RCT (n = 9); No comparison (n = 5); No data required (n = 2); Others (n = 1). Of the 22 studies, 1000 patients treated with total knee arthroplasty were in the training group and 1000 were in the non-training group. The longest follow-up time was 1 year and the shortest was 3 weeks. The study characteristics, patient demographics, and clinical results of each study were shown in Table 1. Outcomes 2/6 MWT The data was divided into two subgroups according to time after training: < 6 months (early stage) and > 6 months (middle stage). 5 studies (566 participants) were contained in this study. The results indicated that early stage MWT of balance training was well than non-balance training, with a statistically significant difference (WMD = 25.93, 95% CI 8.36 to 43.51, P = 0.004, Fig 4). The results indicated that there was statistically significant difference between the middle stage MWT of balance training and that of the non-balance training (WMD = 24.44, 95% CI 7.25 to 41.63, P = 0.005, Fig 4).

Gait speed
The results of 4 studies (435 participants) indicated that early stage gait speed of balance training was well than non-balance training, with a statistically significant difference (WMD = 0.12, 95% CI 0.07 to 0.17, P < 0.00001, Fig 5). The results indicated that there was statistically significant difference between the gait speed of the middle stage balance training and that of the non-balance training (WMD = 0.20, 95% CI 0.13 to 0.28, P < 0.00001, Fig 5).
TUG 9 studies (669 participants) were contained. The results indicated that early stage TUG of balance training was well than non-balance training, with a statistically significant difference (WMD = 1.43, 95% CI 0.99 to 1.88, P < 0.00001, Fig 6). The results indicated that there was statistically significant difference between the TUG of the middle stage balance training and that of the non-balance training (WMD = 0.79, 95% CI 0.45 to 1.13, P < 0.00001, Fig 6). Funnel plot showed that the data included in the meta-analysis were basically symmetrical, which suggested that bias was minimal.  FHR 3 studies (539 participants) were contained. The results indicated that early stage FHR of balance training was well than non-balance training, with a statistically significant difference (WMD = 8.27, 95% CI 4.37 to 12.17, P < 0.0001, Fig 11). The results indicated that there was statistically significant difference between the FHR of the middle stage balance training and that of the non-balance training (WMD = 11.23, 95% CI 6.65 to 15.82, P < 0.00001, Fig 11).

TCS
The results of 6 studies (521 participants) obtained can prove that early stage TCS of balance training was well than non-balance training, with a statistically significant difference (WMD = 0.65, 95% CI 0.01 to 1.28, P = 0.05, Fig 12). The results indicated that there was statistically significant difference between the TCS of the middle stage balance training and that of the non-balance training (WMD = 1.41, 95% CI 1.02 to 1.80, P < 0.00001, Fig 12).
Vitality 3 studies (331 participants) were contained. The results indicated that early stage TCS of balance training was well than non-balance training, with a statistically significant difference (WMD = 14.41, 95% CI 13.53 to 15.30, P < 0.00001, Fig 13). The results indicated that there was statistically significant difference between the vitality of the middle stage balance training and that of the non-balance training (WMD = 10.96, 95% CI -5.16 to 27.08, P < 0.0001, Fig 13).
Stair up/down 3 studies (231 participants) were contained. The results indicated that early stage stair up/down of balance training was well than non-balance training, with a statistically significant difference (WMD = 0.20, 95% CI -2.56 to 2.96, P = 0.89, Fig 14), and indicated that there was no statistically significant difference between the stair up/down of the middle stage balance training and that of the non-balance training (WMD = 0.21, 95% CI -1.70 to 2.13, P = 0.83, Fig 14).

KOOS sysptoms
The results of 4 studies (470 participants) indicated that early stage KOOS sysptoms of balance training was well than non-balance training, with a statistically significant difference (WMD = 6.34, 95% CI 2.07 to 10.60, P = 0.004, Fig 15), and indicated that there was statistically significant difference between the TUG of the middle stage balance training and that of the non-balance training (WMD = 3.07, 95% CI -1.07 to 7.20, P = 0.15, Fig 15).
KOOS ADL 5 studies (363 participants) were contained. The results indicated that early stage KOOS sysptoms of balance training was well than non-balance training, with a statistically significant difference (WMD = 4.91, 95% CI -0.26 to 10.08, P = 0.06, Fig 16).
There was statistically significant difference between the KOOS ADL of the middle stage balance training and that of the non-balance training (WMD = 3.65, 95% CI -0.14 to 7.44, P = 0.06, Fig 16).
KOOS QOL 4 studies (467 participants) were contained in this study. The results indicated that early stage KOOS QOL of balance training was well than non-balance training, with a statistically significant difference (WMD = -0.28, 95% CI -9.52 to 8.95, P = 0.95, Fig 17), and the results indicated that there was statistically significant difference between the KOOS QOL of the middle stage balance training and that of the non-balance training (WMD = 4.43, 95% CI -1.63 to 10.48, P = 0.15, Fig 17).

Function
The results of 6 studies (467 participants) obtained can prove that early stage function of balance training was well than non-balance training, with a statistically significant difference (WMD = 3.20, 95% CI -0.18 to 6.59, P = 0.06, Fig 18). The results indicated that there was statistically significant difference between the function of the middle stage balance training and that of the non-balance training (WMD = 5.85, 95% CI 0.13 to 11.56, P = 0.04, Fig 18).
Pain 13 studies (1098 participants) were contained. The results indicated that early stage pain of balance training was well than non-balance training, with a statistically significant difference (WMD = 0.29, 95% CI -0.10 to 0.67, P = 0.15, Fig 19) and there was statistically significant difference between the pain of the middle stage balance training and that of the non-balance training (WMD = 0.12, 95% CI -0.20 to 0.45, P = 0.46, Fig 19).

ROM
The results of 9 studies (1225 participants) obtained can prove that early stage ROM of balance training was well than non-balance training, with a statistically significant difference (WMD = 0.36, 95% CI -0.38 to 1.10, P = 0.34, Fig 20). The results indicated that there was statistically significant difference between the ROM of the middle stage balance training and that of the non-balance training (WMD = 0.46, 95% CI -0.07 to 0.99, P = 0.09, Fig 20).

Main findings
The meta-analysis included 22

Comparison with other meta-analyses
Compared to previous systematic reviews, [43] we included a larger number of high-quality, up-to-date randomized controlled studies with a larger number of patients (1333 patients (TG = 675, CG = 658). The research project content were also more comprehensive. As can be seen from the tree diagram, the heterogeneity of the two groups of data in this study were very low, and the information extraction were relatively rigorous, so the results were more reliable.

Implications for clinical practice
The study showed that the TG was superior to the CG after total knee arthroplasty in regards to 2/6 MWT, gait speed, TUG, BBS, FHR, TCS, early stage vitality and KOOS sysptoms, middle stage function. Therefore, we believe that postoperative training may be a better approach, and the effect appears to be positive in patients undergoing total knee arthroplasty. In contrast, comprehensive studies have shown that balance and proprioceptive training have no additional benefits in 10-m walk, SLST, stair up/down, KOOS ADL, KOOS QOL, pain, ROM, early stage function, middle stage vitality and KOOS sysptoms. The degree of pain affects the patient's function and quality of life. One of the greatest benefits of total knee arthroplasty is the ability to reduce or even eliminate pain.
Kosek's study found that balance training and neuromuscular intervention did not reduce pain intensity, function, and quality of life in patients with osteoarthritis of knee, consistent with our findings.

Innovation and uniqueness
The studies included in this study were all randomized controlled studies, so the literature quality were high. We excluded studies that resulted in higher heterogeneity to reduce heterogeneity between data, thus making the results more reliable. The evaluation of the included literature was relatively strict and the possibility of bias was small. This study has a large amount of data and a large number of participants, so it was highly reliable.

Limitations
In this study, since there was no blind method of participants, there was a great risk of bias in most of the literature, so the subjective impression will affect the result.There was publication bias in this study, but the degree of bias was acceptable. And the number of patients included in the RCT literatures were less than 50. In addition, study follow-up was short and data collection may be incomplete, which may have an impact on the results.

Conclusions
The pooled data indicated that balance training significantly improved 2/6 MWT, gait speed, TUG, BBS, FHR, TCS, early stage vitality and KOOS sysptoms, middle stage function, so balance training has many positive effects in total knee arthroplasty. There was no significant difference in the middle stage vitality and KOOS sysptoms between the balanced training group and the non-balanced training group, while the training group had better early stage vitality and KOOS sysptoms than the non-training group, showed that the training group had a good effect on early stage vitality and KOOS sysptoms. For the function, the early results of the balance training did not show a benefit to function, but by the middle of the postoperative period, the training group had better function results than the non-training group. These findings may help guide clinical decision-making on recovery after joint arthroplasty.

Ethics approval and consent to participate
All findings in the study are based on published research and therefore do not require moral approval.

Consent for publication
The co-authors agreed to publish.

Availability of data and material
The text and charts in the article contain all the data.

Funding
None.     Figure 1 The flow diagram of study selection Comparison of 2/6 MWT between TG and CG.

Figure 5
Comparison of gait speed between TG and CG.

Figure 6
Comparison of TUG between TG and CG.

Figure 7
Funnel plot assessing publication bias.

Figure 8
Comparison of 10-m walk between TG and CG. Comparison of BBS between TG and CG.

Figure 10
Comparison of SLST between TG and CG.

Figure 11
Comparison of FHR between TG and CG.

Figure 12
Comparison of TCS between TG and CG.

Figure 13
Comparison of vitality between TG and CG.

Figure 14
Comparison of stair up/down between TG and CG.

Figure 15
Comparison of KOOS sysptoms between TG and CG.

Figure 16
Comparison of KOOS ADL between TG and CG.

Figure 17
Comparison of KOOS QOL between TG and CG.

Figure 18
Comparison of function between TG and CG.

Figure 19
Comparison of pain between TG and CG.

Figure 20
Comparison of ROM between TG and CG.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.