This review was conducted according to the Cochrane Collaboration methodological guidelines [40]. We searched 6 databases (PubMed, Web of Science, EMBASE, Cochrane Library databases, CNKI, and WANFANG) for RCTs published from January 1, 2000, to September 3, 2023. Relevant articles and reference lists were manually searched. The obtained articles were reviewed by 2 investigators (WX and HX) independently.

The overall search strategies were performed by using a combination of relevant medical subject heading terms and free-text words (telemedicine or e-health or telehealth or telerehabilitation or internet or web or online or app or wearable or sensor) and knee OA. The detailed search strategy is described in Multimedia Appendix 1.

Data extraction was completed by 2 authors (WX and HX) independently. Disagreements between the 2 investigators were resolved by a third reviewer (BJJ). The extracted data included basic information of the study, participants, type of intervention for the experimental group and the control group, and outcomes. Outcomes reported as continuous variables are presented as the mean (SD).

We extracted 8 telerehabilitation strategies according to 3 categories: technology applications (1 strategy), care objectives (3 strategies), and care support methods (4 strategies) (Table 1).

The primary outcome measures were focused on pain assessed by the Western Ontario and McMaster Universities Osteoarthritis (WOMAC) pain subscale and physical function assessed by WOMAC functional subscale.

A summary of the methodological risk of bias of the included studies was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [34] by 2 investigators (JYW, LJL) using the risk of bias tool in the Cochrane Collaboration’s review-writing software RevMan (version 5.4). The risk of bias assessment of RCTs mainly included 7 aspects: random sequence generation, allocation sequence concealment, blinding of participants and personnel, blinding of outcome assessment, completeness of outcome data, and selective outcome reporting [40]. Each item was judged as being at a high, low, or unclear risk of bias [40].

The mean (SD) of continuous outcome variables after therapy was used to calculate the total effect size via the mean difference and 95% CI. The standardized mean difference (SMD) was calculated when studies used different methods to measure the same outcome. The heterogeneity of RCTs in each group was examined by the P value and I² statistic. A random-effects model was applied when P<.05 or I²>50%; otherwise, a fixed-effects model was used [43]. The meta-analysis methods and tests were performed using RevMan 5.4.

The literature search results are presented in Figure 1. A total of 2958 articles were searched from the databases and 3 articles were searched manually, resulting in a total of 2961 articles. After removing 1190 duplicate articles, 1768 articles were obtained for screening. Of these, 1629 articles were excluded for not fulfilling the inclusion criteria, and 139 articles were obtained for a full-text assessment. Of these, 133 articles were excluded based on the inclusion and exclusion criteria. Finally, 6 RCTs were included in this review, for which pain was assessed by the WOMAC pain subscale and physical function was assessed by WOMAC functional subscale.

Among the bias risk assessment elements, the blinding of participants and personnel was the least used method in the RCTs (Figure 2). There were 2 RCTs that did not blind participants and personnel (Figure 3 [27-32]), while 2 RCTs did not report their blinding status and 2 RCTs used a blinding approach. Meanwhile, selective reporting and other bias was the least reported element, and 3 RCTs (50%) had an unclear risk of bias.

The 6 RCTs included 734 participants. The baseline descriptive characteristics (country, sample size, age, and gender) of the 6 studies included in the systematic review are summarized in Table 2. One study was from the United States [28], 2 were from Australia [27,32], 1 was from Brazil [31], and 2 were from China [29,30]. The mean age of patients with knee OA ranged from 53.1 (SD 8.5) years to 72.25 (SD 8.84) years, and all studies included both men and women.

We extracted 9 telerehabilitation strategies from the 6 RCTs, as shown in Table 3. Some strategies were commonly used, such as physiotherapy support (n=4, 67%), physician support (n=4, 67%), intervention for education (n=3, 50%), and physiotherapist support (n=3, 50%). Strategies that were not commonly used included intervention for depression and anxiety (n=2, 33%) and psychologist support (n=2, 33%). The telerehabilitation programs in the RCTs generally contained multiple strategies, with a mean of 4.33 strategies per care program.

We assessed pain and physical function in the 6 RCTs (n=624) using the WOMAC pain and function subscales, respectively. The outcomes of pain and function with 95% CIs are shown in Figure 4 and Figure 5, respectively [27-32]. Overall, telerehabilitation was found to be more effective than conventional treatment for the improvement of pain (SMD –0.21, 95% CI –0.35 to –0.07; P=.003), but not physical function (SMD –0.09, 95% CI –0.25 to 0.06; P=.24). The outcomes of both pain and physical function were heterogeneous, with a low level of heterogeneity (I²=0%) in both the pain and physical function outcomes.

Compared with conventional treatment, the group of RCTs that provided various telerehabilitation strategies was not found to be more effective for improving pain and physical function, as shown in Tables 4 and 5.

This systematic review and meta-analysis investigated whether pain and physical function in patients with knee OA could be improved by telerehabilitation programs and different telerehabilitation strategies. The results showed that the pain, but not the physical function, of patients with knee OA could be significantly improved by telerehabilitation compared with traditional therapy or usual care. Subgroup analyses revealed that the pain and physical function in patients with knee OA could not be further improved by combining different telerehabilitation strategies. This finding adds evidence to support telerehabilitation interventions for patients with knee OA.

Pain is the primary symptom in patients with knee OA; it occurs gradually, worsens with time, can lead to many problems, and is the number 1 reason most patients seek medical attention. Consistent with previous systematic reviews and meta-analyses [44-48], our findings indicated that the pain of knee OA could be relieved by telerehabilitation after patients are discharged from the hospital. The positive effects may benefit from telehealth intervention features, which enable patients living in remote or medically resource-poor areas to receive professional medical help [26]. Programs such as IBET were shown to be effective for pain reduction [22], which may be attributed to personalized exercise plans to reduce pain [22]. Meanwhile, telerehabilitation strategies, including educational lectures, medical suggestions, and psychotherapy were effective for the reduction of pain [23]. Furthermore, Bennell et al [49] suggested that telehealth-delivered exercise and diet programs improved pain in people with knee OA and overweight or obesity, which indicated that diet also plays an important role in alleviating pain in patients with knee OA. These telerehabilitation programs could combine various interventions to ease pain. For patients, the place of rehabilitation exercise is more convenient.

As for the effect of telerehabilitation on physical function, this systematic review and meta-analysis suggested that the physical function of patients with knee OA could not be improved by telerehabilitation. Previous systematic reviews have reported inconsistent results [22,44-48,50]. Some studies [22,44-46,50] showed no significant improvement in physical function in patients with knee OA. For example, Allen et al [22] found that there was no significant difference in the effect of network sports training on improving physical function in patients with knee OA compared to the conventional physical therapy group, which may be related to the emphasis on exercise training guidance and low patient participation. However, Safari et al [47] found that a digital-based, structured self-management program improved physical function in patients with knee OA, and similar results were reported by Schäfer et al [48]. There was no significant improvement in physical function following internet-based exercise training compared with face-to-face supervised exercise [22]. Hinman et al [26] showed that physical function could be modestly improved by telephone-delivered physiotherapist-led exercise advice and support at 6 months, but functional benefits were not sustained at 12 months. The reasons for this result might be due to the fact that the recruited participants often had better baseline physical function and functional improvement required a longer-term intervention and more intervention forms.

In addition to performing a meta-analysis of overall effectiveness, we used a subgroup comparison method to analyze the effects of different telerehabilitation strategies on the improvement of pain and physical function. The results indicated that there is no correlation between different telerehabilitation strategies and the improvement of pain and physical function in patients with knee OA. Anwer et al [51] found that home exercise programs with and without supervised clinic-based exercises were beneficial in the management of knee OA, which is consistent with the results of this review. However, Sinatti et al [52] showed that education seems to be effective in reducing pain and improving function in patients with knee OA. We speculate that this may be related to the sample size included in the various intervention strategies and the length of intervention time. However, these strategies should not be ignored, and further investigation of their contribution to knee OA treatment remains important to continuously improve telerehabilitation outcomes in future studies.

There are several limitations in this systematic review and meta-analysis. First, fewer studies were included in this review. Second, the objective of our systematic review was to evaluate different telerehabilitation strategies, and our meta-analysis did not rigorously exclude RCTs with risk of bias. Third, the outcome measures to assess pain and physical function in patients with knee OA were subjective. Fourth, moderator variables (eg, age, gender, and sample size) for telerehabilitation effects were not analyzed. Finally, considering the diversity of outcome indicators, only studies using the WOMAC scale were included in the analysis to ensure the reliability of the study.

Internet-based rehabilitation is a promising strategy for patients with knee OA. Compared with conventional rehabilitation, the results of this meta-analysis suggest that telerehabilitation programs could improve the pain but not the physical function of patients with knee OA. Meanwhile, there was no significant correlation between different telerehabilitation strategies and the pain and physical function of patients with knee OA. These results indicate that telerehabilitation is beneficial for the implementation of home rehabilitation exercises for patients with knee OA, thereby reducing the economic burden of health. However, there is currently relatively little research on the effects of telerehabilitation on knee OA. In the future, more high-quality studies with large samples are needed to focus on the long-term outcomes of telerehabilitation for patients with knee OA and the effect of different telerehabilitation strategies. The completion of high-quality trials will ultimately advance our knowledge about optimal telerehabilitation strategies for patients with knee OA.