Efficacy and safety of radiofrequency treatment for improving knee pain and function in knee osteoarthritis: a meta-analysis of randomized controlled trials

The meta-analysis was performed in accordance with PRISMA (Preferred Reporting Items for Systematic review and Meta-Analysis) guidelines [20]. It was registered with PROSPERO (CRD42021292558). We searched the PubMed, Web of Science, EMBASE, Cochrane Library, China National Knowledge Infrastructure, and Wanfang Data databases until August 30, 2021. The following medical subject heading terms were used: “knee,” “osteoarthritis,” “radiofrequency,” “genicular nerve,” “intra-articular,” “randomized controlled trial,” “controlled clinical trial” and “humans.” No language or country limitations were applied to our meta-analysis. All RCTs in the search results were screened by two authors according to the inclusion criteria. If the article title and abstract description were ambiguous with respect to the inclusion criteria, the full text was downloaded and reviewed carefully.

Reported outcomes in published RCTs were recorded for RF and control groups of knee OA patients. Studies were eligible for inclusion if they met all of the following criteria: (i) patients were diagnosed with knee OA; (ii) patients in the experimental group received RF therapy (RFA, PRF, CRF, and other form); (iii) the clinical trial was designed with a control group; (iv) the study included the following outcome measurements: the Visual Analogue Scale (VAS) or Numerical Rating Scale (NRS), the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Oxford Knee Score (OKS), Global Perceived Effect (GPE) scale, and adverse effects at different time points after treatment; and (v) studies were RCTs.

The exclusion criteria for the meta-analysis were as follows: (i) patients underwent knee arthroplasty or arthroscopic surgery; (ii) full text was not available; (iii) provided unextractable or insufficient data; and (iv) case reports, abstracts, conference presentations, editorials, and expert opinions.

Two experienced researchers independently assessed the quality of the included RCTs with the Cochrane Handbook for Systematic Reviews of Interventions, which includes 10 specific domains consisting of random sequence generation, allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data/intention-to-treat analysis/loss to follow-up (attrition bias), compliance, and selective reporting (selection bias). Disagreements were resolved by a third person who served as an intermediary and made the final decision. Each included study was graded as having a high risk, low risk, or unclear risk of bias.

Two reviewers independently extracted relevant data from the original studies using a standardized data extraction form and clarified discrepancies by re-evaluation and discussion with the other authors. The following data were extracted for analysis: name of the first author, year of publication, country, study design, sample size, age, sex, Kellgren–Lawrence classification, mode of RF, ultrasound transducer parameters, primary outcomes such as VAS/NRS, OKS, WOMAC, and GPE scale at baseline and at available follow-up times, and adverse effects. The corresponding authors of RCTs were contacted to request missing data or clarification regarding unclear data.

The statistical analysis was performed according to the recommendations from the Cochrane Collaboration. Weighted mean difference (WMD) and corresponding 95% confidence interval (CI) values for the difference in means were used to evaluate continuous data, and the risk difference with 95% CI was calculated for dichotomous data. Heterogeneity across studies was assessed by the Cochran Q test (significance level of P < 0.05) and the I² statistic. For the I² statistic, we considered I² < 25% as low heterogeneity and I² > 75% as high heterogeneity. Data were also analyzed with a fixed-effects model for P > 0.05 and I² < 50% or a random-effects model for P < 0.05 and I² ≥ 50%. Based on the differences in variables such as RF modes, location, intervention target, diagnosed nerve block (DNB), sex ratio among cases, and body mass index (BMI), subgroup analyses were performed. Sensitivity analysis was conducted to observe the impact of any single study on the pooled WMD. Review Manager Version 5.3 (The Nordic Cochrane Center, The Cochrane Collaboration, 2014, Copenhagen) software was used to analyze the pooled data. A two-tailed P value < 0.05 was considered statistically significant.

The primary characteristics of the included RCTs are presented in Table 1. We included 15 studies in our meta-analysis, of which 13 were single-center studies and 2 were multi-center studies. The included RCTs were conducted in eight countries and were published between 2011 and 2021. A total of 1009 patients were enrolled from the 15 eligible trials. Of these, 503 recipients had been assigned to receive RF and were included in the RF group, and the remaining patients who were not treated with RF were assigned to the control group. The patients ranged in age from 47.8 to 70.9 years.

For RF therapy, RFA was applied in 7 trials, PRF in 4 trials, CRF in 2 trials, and CRMRF and RF thermocoagulation in 1 trial, respectively. Regarding to the intervention targets, eight studies focused on the genicular nerve, and seven studies applied an intra-articular procedure. Moreover, five studies applied DNB to obtain the source of pain and position of targets for RF therapy. In addition, VAS/NRS scores were available in 12 studies for comparing the pain improvement between the two groups at different follow-up time points. The OKS and WOMAC scores were available in 4 and 5 studies, respectively, for evaluation of knee functional improvement. GPE scale data were available in 3 studies for assessing the patients’ degree of satisfaction regarding treatment effectiveness. The detailed intervention procedural parameters, results, and adverse effects are presented in Table 2.

Two authors independently assessed the quality of each RCT. Among the 15 RCTs, 12 trials had relatively high methodological quality and met allocation concealment criteria. The overall details of quality assessment are shown in Figs. 2 and 3.

Twelve trials with 706 patients reported the outcome of pain score on VAS or NRS, with 10 RCTs reporting scores at 1–2 weeks, 9 RCTs reporting scores at 4 weeks, 11 RCTs reporting scores at 12 weeks, and 6 RCTs reporting scores at 24 weeks (Fig. 4). Because the pooled results from data at four time points exhibited significant heterogeneity, the random-effects model was used to obtain WMDs and the corresponding 95% CIs (all I² > 50%, P < 0.05). Meta-analysis indicated that significant pain relief was achieved at four follow-up time points by the application of RF compared with that of patients in the control group (1–2 weeks, WMD = − 1.72, 95% CI − 3.96 to − 0.1.44, P < 0.001; 4 weeks, WMD = − 1.49, 95% CI − 1.76 to − 1.21, P < 0.001; 12 weeks, WMD = − 1.83, 95% CI − 2.39 to − 1.26, P < 0.001; 24 weeks, WMD = − 1.96, 95% CI − 2.89 to − 1.04, P < 0.001).

Due to significant between-study heterogeneity for OKS, the random-effects model was applied, except that the fixed-effects model was used at 1–2 weeks after treatment (Additional file 1: Fig. S1). Notably, the majority of results for OKS indicated insignificant knee function improvement with RF treatment (1–2 weeks, WMD = − 2.73, 95% CI − 5.10 to − 0.37, P = 0.02; 4 weeks, WMD = − 2.15, 95% CI − 9.51 to 5.21, P = 0.57; 12 weeks, WMD = − 0.23, 95% CI − 11.24 to 10.77, P = 0.97). However, the pooled results for WOMAC score at three time points exhibited no significant heterogeneity, and the fixed-effects model was used to obtain WMDs and the corresponding 95% CIs (all I² < 50%, P > 0.05; Fig. 5). The pooled results showed that RF treatment significantly improved knee function (4 weeks, WMD = − 10.64, 95% CI − 13.11 to − 8.17, P < 0.001; 12 weeks, WMD = − 6.12, 95% CI − 7.67 to − 4.57, P < 0.001; 24 weeks, WMD = − 10.89, 95% CI − 12.28 to − 9.51, P < 0.001, respectively).

Three studies reported the outcome of GPE scale, which also showed significant heterogeneity (4 weeks, I² = 93%, P < 0.001; 12 weeks, I² = 78%, P = 0.01, respectively). Therefore, the random-effects model was used (Fig. 6). The pooled results indicated no significant difference at 4 weeks after treatment and a significant difference at 12 weeks between the two groups (4 weeks, WMD = − 0.63, 95% CI − 0.15 to 1.42, P = 0.12; 12 weeks, WMD = 1.12, 95% CI 0.61 to 1.63, P < 0.001).

Adverse events induced by RF were reported in 91 patients in two RCTs and were not serious (Fig. 7). The majority of these adverse events were deemed unrelated to the study intervention. Davis et al. reported that three patients in the CRF group experienced four severe adverse events (SAEs), whereas seven patients in the control group experienced eight SAEs. However, they illustrated that none of the SAEs were related to the study treatments. In addition, no adverse events were reported in 9 studies. There was no significant heterogeneity among the 12 studies (I ² = 17%, P = 0.28), and a fixed-effects model was used. Based on the available data, the use of RF treatment did not significantly increase adverse effects (risk difference 0.03, 95% CI − 0.01 to 0.06, P = 0.14).

Subgroup analysis was conducted to find the sources of heterogeneity for pain score associated with the pooled results. The following subgroup analyses were performed: 1) with or without RFA treatment; 2) location (Asia vs. others); 3) application of DNB before treatment or not; 4) treatment target of genicular nerve or other; 5) ratio of females/males ≥ 2 or not; and 6) patient BMI ≥ 30 or < 30 kg/m². Summarized quantitative data for these subgroups at 1–2 weeks, 4 weeks, and 12 weeks after treatment are presented in Table 3 and Additional file 1: Tables S1 and S2. The data at 1–2 weeks after treatment showed that RF mode (RFA, WMD = − 1.76, 95% CI − 2.30 to − 1.22, P < 0.001), location (Asia, WMD = − 1.63, 95% CI − 2.07 to − 1.20, P < 0.001), site of radiofrequency (genicular nerve, WMD = − 1.64, 95% CI − 2.19 to − 1.09, P < 0.001), DNB (no, WMD = − 1.81, 95% CI − 2.27 to − 1.35, P < 0.001), sex ratio (≥ 2, WMD = − 1.59, 95% CI − 2.15 to − 1.02, P < 0.001), and BMI (< 30 kg/m², WMD = − 1.80, 95% CI − 3.29 to − 0.31, P = 0.02) were potential sources of heterogeneity (all P < 0.05).

Due to the significant between-study heterogeneity and results from subgroup analysis, we performed a sensitivity analysis to assess the stability of the pooled WMDs regarding pain score. After excluding each individual study separately, the WMDs were recalculated to identify any significant change in our results. The results of sensitivity analysis showed that the elimination of any single study was unlikely to overturn our findings (Additional file 1: Fig. S2).

Osteoarthritis is a considerable cause of disability due to the increasing prevalence of obesity and current aging of the global population [36, 37]. As a result, the development of innovative therapeutic strategies for knee OA to relieve the persistent pain and improve knee function is an important challenge. RF therapy has emerged as one of the most investigated and effective approaches in modern knee OA treatment [38]. In the present study, a meta-analysis was performed to comprehensively assess the efficacy and safety of RF in patients with knee OA. Our results suggest that the use of RF correlated with improvements in pain relief (VAS/NRS score) and knee function (WOMAC score) at four follow-up time points after treatment, but did not lead to significant improvement in the OKS. Moreover, adverse effects showed no statistically significant difference between the RF and control groups.

In 2018, Hong et al. [39] performed a systematic review and meta-analysis that included 12 RCTs with 841 patients and suggested that the use of RF could decrease the pain scores (VAS) of patients at 1 week, 1 month, and 3 months after treatment, but revealed no significant improvement in knee function, which is inconsistent with the results of our meta-analysis and those of Zhang et al. [40]. Differently from our study, the negative effect on knee function in the previous systematic review by Hong et al. was evaluated by OKS, while the WOMAC score was not considered. The WOMAC score is widely applied in the evaluation of hip and knee OA for assessment of the activities of gait, daily living, general health, functional mobility, and quality of life, and has been identified as one of the highest performing outcome measures in terms of validity, reliability, interpretability, and responsiveness [41‐43]. Furthermore, although the results for the effectiveness and safety of RF treatment were consistent between the current meta-analysis and previous meta-analyses by Zhang et al. [40] and Li et al. [44], the heterogeneity requires further analysis. In the present study, subgroup and sensitivity analyses were performed on RCTs to test the robustness of the pooled results and explore the potential sources of heterogeneity in our meta-analysis.

The present meta-analysis including 1009 patients from 15 RCTs demonstrated that the use of RF treatment seemed to be effective at increasing the patient’s degree of satisfaction with the treatment effectiveness after 12-week treatment, although no statistical significance was seen at the 4-week follow-up after treatment. In addition, no SAEs were observed in any patients who received RF therapy. From our results, RF shows a seemingly excellent curative effect in patients with knee OA.

The pulse generator of RF is a simple electrode structure that generates an electromagnetic field when an electric current is passed through. RF pulse generators are able to generate frictional heat in the surrounding tissue through ionic (Na⁺, K⁺ and Cl⁻) oscillating motion repeatedly in the presence of an electromagnetic field, which in turn causes thermal destruction of the nerves and interruption of the pain impulses [45]. Meanwhile, the use of fluoroscopic or ultrasonographic guidance could ensure smooth, painless, and precise introduction of the RF cannula into the joint. This characteristic explains the immediate pain relief effect of RF therapy. However, rehabilitation training should be conducted after alleviation of the pain to increase muscle strength in the lower extremities. Patients with long-term knee joint pain may be afraid of the recurrence of pain and, thus, limit activities involving the knee joint in their daily lives [33]. Such limitation of activity may influence the measurement of patient-rated recovery, leading to a response of no significant difference in GPE scales in the short-term follow-up period. As the knee pain eases and the amount of functional activity increases, such as going up and down stairs, patients are more likely to consider their disease condition improved and report high life-satisfaction in the long-term follow-up period [33].

Our results indicated that RF treatment significantly improved knee function as assessed by the WOMAC at 4, 12, and 24 weeks after treatment, rather than the OKS. The OKS is a 12-item patient-reported outcome questionnaire regarding an individual’s level of function, activities of daily living, and how they have been affected by pain over the preceding 4 weeks. The OKS was specifically designed and developed to assess function and pain after total knee replacement (TKR) surgery or TKA [46, 47]. The WOMAC is a widely used, proprietary set of standardized questionnaires for evaluating the condition of patients with osteoarthritis of the knee and hip, including pain, stiffness, and physical functioning of the joints [48, 49]. The WOMAC is more sensitive to changes than the OKS following nonsurgical interventions for knee OA, including the effects of physical therapy, weight loss, electrotherapy corticosteroid injection, intra-articular hyaluronic acid injection, and autologous chondrocyte implantation [50‐52]. In the present study, RCTs with patients who had undergone knee arthroplasty or arthroscopic surgery were excluded. Thus, it was better to apply the WOMAC to assess knee function in patients with knee OA.

Our meta-analysis suggested that pain and knee function might be alleviated by RF therapy, but the results for these curative effects were highly heterogeneous. Although subgroup analysis and sensitivity testing were performed on the pooled results, the heterogeneity was not effectively improved, which may be attributed to the following factors: (1) the inclusion criteria of Kellgren-Lawrence OA grade: the K-L grades of participants in each RCT, which ranged from 1 to 4, were relatively different. Some studies only included patients with relatively severe OA (K-L grade III-IV), while some studies included patients with relatively mild OA (K-L grade I-III). This difference in the degree of progression of knee OA will inevitably affect the difference in the efficacy of RF across the studies; (2) the intervention parameters for RF: when we re-examined the included trials, we found considerable differences in terms of RF protocols related to cycles, total operation time, and temperature, for example. Above all, differences existed even within RFA or PFA procedures [53]; (3) DNB prior to the RF procedure: innervation of the knee joint is complex and a DNB may have a role in predicting response to a RF procedure. Most of the included studies did not describe the use of DNB prior to the RF procedure. In addition, the cutoff values for the duration and amount of pain relief following DNB were not standardized within the five RCTs that reported DNB; and (4) others: the effects of RF may also be affected by the presence of local anesthetic, gender, mental health disorders, diabetes mellitus, and other conditions [54, 55].