Interventional procedure overview of radiofrequency denervation for osteoarthritic knee pain
Closed for comments This consultation ended on at Request commenting lead permission
Evidence summary
Population and studies description
This interventional procedures overview is based on approximately 2,896 patients from 1 systematic review and network meta-analysis, 2 systematic reviews and meta-analyses, 1 RCT, 1 long-term cohort study that was a single-arm extension of an RCT, 1 cohort study, 1 narrative review and 2 case reports. There was some overlap in studies included in the systematic reviews. Of these patients, approximately 2,047 had the procedure. This is a rapid review of the literature, and a flow chart of the complete selection process is shown in figure 1. This overview presents 9 studies as the key evidence in table 2 and table 3, and lists 48 other relevant studies in table 5.
Of the 9 key evidence studies, 3 were systematic reviews and meta-analyses. Wu (2022) is a comprehensive network meta-analysis of RCTs that compares the efficacy of all modalities of RF denervation with each other, non-RF comparators, and placebo. Chou (2021) is a meta-analysis of RCTs and observational studies of the treatment effect of RF denervation (before versus after treatment). Liu (2022) is a meta-analysis of RF denervation versus several comparators pooled into a single group.
One further comparative study was an RCT of conventional RF versus sham versus control (best supportive care only; Kumaran, 2019). This study was also included in the Liu (2022) meta-analysis.
Long-term data is reported in a long-term cohort study which was a long-term extension of an RCT (Lyman, 2022). This study reported 2-year outcomes of cooled RF denervation. An observational study (Chen, 2021) provides real-world outcomes of RF denervation and reports prognostic factors. A narrative review and 2 case reports are included to report unique safety events (McCormick, 2021; Mateev, 2021; Jorge 2019).
In the Liu (2022) systematic review, there were 7 studies from China, 2 from the US, and 1 each from the UK, Italy, Korea, Iran, Turkey, and Egypt. The locations of studies were not well reported in the other systematic reviews; of the other studies, Chen (2021) and Lyman (2022) were conducted in the US; Kumaran (2019) was conducted in the UK. Follow-up periods were typically 6 to 12 months.
Table 2 presents study details.
No. | Author, date Location | Patients (m:f) Studies | Age | Study design | Inclusion criteria | Intervention | Follow up |
1 | Wu, 2022 Various | 21 studies n=1,818 (657 non-RF comparator) | NR | Systematic review and network meta-analysis of RCTs | RCTs with at least 1 treatment arm using a RF denervation treatment and a comparator arm using placebo or other active treatment; Studies including patients with knee osteoarthritis; Studies containing pain or functional outcome scores; Studies fully reporting the numbers of patients and involved knees and the time of follow up; Studies completely reporting the RF denervation methodology, target, and number of electrodes; Studies diagnosing OA clinically and radiographically. | Conventional RF: 10 studies Pulsed RF: 11 studies Cooled RF: 2 studies (some studies used multiple types) Network meta-analysis consisted of all treatments vs. all comparators | Various; only 3 and 6 month results included in network meta-analysis. |
2 | Chou, 2021 Various | 20 studies n=605 148 male; 343 female | Means ranged from 53.3 to 77.2 | Systematic review and meta-analysis of RCTs and observational studies | Articles investigating the efficacy of pulsed, conventional, or cooled RF denervation techniques; Patients with knee osteoarthritis; VAS or NRS used to evaluate pain; Articles written in English or Chinese. | Conventional RF: 11 studies Pulsed RF: 8 studies Cooled RF: 2 studies Meta-analysis was conducted on before vs. after treatment | Up to 12 months |
3 | Liu, 2022 Various | 15 studies n=1,009 (506 non-RF comparator) 309 male; 594 female | Means ranged from 47.8 to 70.9 | Systematic review and meta-analysis of RCTs | Patients were diagnosed with knee osteoarthritis; Patients in the experimental group received RF therapy; The trial had a control group; Included the following outcome measurements: VAS or NRS, WOMAC, Oxford Knee Score, Global Perceived Effect scale, and adverse effects at different time points after treatment; Studies were RCTs. | Conventional RF: 9 studies Pulsed RF: 4 studies Cooled RF: 2 studies Meta-analysis was conducted on RF denervation vs. comparator | Up to 24 weeks |
4 | Kumaran, 2019 UK | n=45 (15 conventional RF; 15 sham; 15 control) 18 male;27 female | Mean RF 63; mean sham 63; mean control 60 | RCT | Symptomatic for a minimum of 6 months; A prior clinical and/or radiological diagnosis of osteoarthritis knee meeting the American College of Rheumatology criteria. | Conventional RF | 3 months |
5 | Lyman, 2022 US | n=32 Male:female breakdown NR | NR | Prospective cohort study (Single arm, multicentre, long-term extension of an RCT [Chen, 2020 in the appendix]) | Baseline NRS score of 6 or more (usual daily pain) for the index knee was required for enrolment in the RCT Baseline of score of 2 or 3 on WOMAC question A1 (pain while walking on flat surface) and a baseline mean score of 1.5 to 3.5 on all five questions of the WOMAC subscale A (pain) Only people initially randomised to cooled RF denervation were eligible for the extension. | Cooled RF | Up to 24 months |
6 | Chen, 2021 US | n=265 96 male;169 female | Mean 64.3 | Retrospective, multicentre, cohort study | Patients with a primary complaint of knee pain treated with a radiofrequency procedure(s). | Conventional RF: n=103 Pulsed RF: n=7 Cooled RF: n=151 | 3 months or more |
Safety studies – the following studies were included to show unique safety events | |||||||
7 | McCormick, 2021 | 5 studies n=8 Male:female breakdown NR | NR | Narrative review | NR | Conventional RF: 2 study Cooled RF: 3 studies | NR |
8 | Matveev, 2021 US | n=1 Male | 49 | Case report | NR | Conventional RF | NR |
9 | Jorge, 2019 US | n=1 Male | 76 | Case report | NR | Conventional RF | NR |
Procedure technique
All 3 modalities of RF denervation were included in the overview. In addition, the Wu (2022) network meta-analysis included comparisons between polarity of RF treatment (bipolar vs. monopolar) and target of RF (genicular nerve vs. saphenous nerve vs. articular cartilage). The Chen (2021) cohort study also identified treatment characteristics associated with a positive outcome.
Efficacy
Pain relief
Six studies reported on pain relief.
In the Wu (2022) network meta-analysis, pain relief (as measured by VAS) was assessed at 3 and 6 months. At 3 months, the network was comprised of 16 RCTs (16 interventions) of 1,401 patients, for a total of 120 paired estimates (17 direct and indirect evidence; 103 indirect evidence only). All treatments except for exercise had statistically significantly decreased VAS compared to placebo at 3 months. According to the surface under the cumulative ranking curve (SUCRA), the treatment rankings were as follows (best to worst; MD of treatment vs. placebo [95% CI] shown):
Pulsed bipolar articular cavity RF, -3.8 (-4.8 to -2.8)
Conventional monopolar genicular nerve RF, -3.5 (-3.8 to -3.2)
Pulsed monopolar articular cavity RF combined with intra-articular platelet-rich plasma injection, -3.5 (-4.3 to -2.6)
Intra-articular erythropoietin injection, -3.1 (-4.5 to -1.7)
Intra-articular platelet-rich plasma injection, -2.8 (-3.7 to -1.9)
Pulsed monopolar genicular nerve RF, -2.8 (-3.2 to -2.5)
Pulsed monopolar articular cavity RF, -2.8 (-3.8 to -1.8)
Intra-articular anaesthesia, -2.0 (-2.5 to -1.5)
Intra-articular dextrose injection, -2.0 (-3.4 to -0.6)
Intra-articular sodium hyaluronate injection, -1.8 (-2.3 to -1.3)
Pulsed monopolar saphenous nerve RF, -1.6 (-2.1 to -1.1)
Intra-articular corticosteroid injection, -1.6 (-2.5 to -0.7)
NSAIDs, -1.4 (-1.7 to -1.0)
Exercise, -0.2 (-0.7 to 0.3)
Placebo, reference
At 6 months, the network was comprised of 10 RCTs (10 interventions) of 1,021 patients, for a total of 45 paired estimates (10 direct and indirect evidence; 35 indirect evidence only). All treatments except for NSAIDs had statistically significantly improved VAS compared to exercise at 6 months (placebo was not included in the network). According to the SUCRA, the treatment rankings were as follows (best to worst; MD of exercise vs. treatment [95% CI] shown):
Conventional bipolar genicular nerve RF, 5.5 (4.3 to 6.7)
Cooled monopolar genicular nerve RF, 4.7 (3.8 to 5.6)
Conventional monopolar genicular nerve RF, 3.5 (3.1 to 3.9)
Pulsed monopolar articular cavity RF combined with intra-articular protein-rich plasma, 3.3 (2.6 to 3.9)
Pulsed monopolar genicular nerve RF, 2.5 (2.2 to 2.8)
Intra-articular corticosteroid injection, 1.9 (1.1 to 2.7)
Intra-articular sodium hyaluronate injection, 1.9 (1.0 to 2.7)
Intra-articular platelet-rich plasma injection, 1.5 (0.4 to 2.6)
NSAIDs, 0.06 (-0.7 to 0.9)
Exercise, reference
In the Chou (2021) meta-analysis, there was a statistically significant decrease in pain from before treatment to 1, 3, 6, and 12 months after treatment. There was no statistically significant difference in pain relief between the RF modalities at any time point.
1 month (17 studies): SMD 3.25 (95% CI 2.56 to 3.93; I2=93%; p<0.00001)
3 months (14 studies): SMD 3.39 (95% CI 2.47 to 4.31; I2=96%; p<0.00001)
6 months (12 studies): SMD 4.84 (95% CI 3.62 to 6.03; I2=95%; p<0.00001)
12 months (4 studies): SMD 2.71 (95% CI 1.23 to 4.18; I2=94%; p=0.00003)
In the Liu (2022) meta-analysis, there was a statistically significant decrease in pain compared to control (either sham, intra-articular injection, or oral NSAIDs) at 1 to 2, 4, 12, and 24 weeks after treatment.
1 to 2 weeks (10 studies): WMD -1.72 (95% CI -2.17 to -1.30; I2=78%; p<0.00001)
4 weeks (9 studies): WMD 1.49 (95% CI -1.76 to -1.21; I2=66%; p<0.00001)
12 weeks (6 studies): WMD 1.83 (95% CI -2.39 to -1.26; I2=88%; p<0.00001)
24 weeks (studies): WMD 1.96 (95% CI -2.89 to -1.04; I2=97%; p<0.0001)
In the Kumaran (2019) RCT of conventional RF vs. sham vs. control, there was a statistically significant main effect for time (within group change) (F [2.1, 88]=16, p<0.001) and a significant interaction between group and time (F[4.2]=5.2, p=0.001). This infers that the type of intervention made a statistically significant difference to VAS pain, and that there was a statistically significant overall difference between pre, post, 1-month, and 3 months. The decrease in VAS score in the RF group was greater than published MCID estimates.
In the long-term cohort study, patients treated with cooled RF denervation had statistically significantly lower scores on the pain NRS at 18 and 24 months than at baseline (all p<0.0001). In Kaplan-Meier analysis, there was approximately a 50% chance that patients maintained a 50% or greater pain relief through 700 days after treatment (Lyman, 2022). Per the Global Perceived Effect questionnaire, 63% reported a perceived improvement in pain at 24 months (Lyman, 2022).
In the Chen (2021) cohort study, 162/265 (61.1%) patients had a positive outcome, defined as at least 30% pain relief lasting at least 3 months after RF denervation. In multivariate analysis, the following factors were associated with positive outcome:
Being obese, OR 3.68 (95% CI 1.66 to 8.19, p=0.001)
Not using opioids, OR 0.35 (95% CI 0.16 to 0.77, p=0.009)
Not being depressed, OR 0.29 (95% CI 0.10 to 0.82, p=0.02)
Use of cooled RF, OR 3.88 (95% CI 1.63 to 9.23, p=0.002)
Performing multiple lesions at each neural target, OR 15.88 (95% CI 4.24 to 59.50, p<0.001)
Composite knee function and pain measures
WOMAC
Four studies reported WOMAC outcomes.
In the Wu (2022) network meta-analysis, WOMAC score was assessed at 3 and 6 months. At 3 months, the network was comprised of 14 RCTs (14 interventions) of 1,091 patients, for a total of 91 paired estimates (14 direct and indirect evidence; 77 indirect evidence only). All treatments except for exercise, NSAIDs, and pulsed monopolar saphenous nerve RF, had statistically significantly decreased WOMAC compared to placebo at 3 months. According to the SUCRA, the treatment rankings were as follows (best to worst; MD of treatment vs. placebo [95% CI] shown):
Pulsed bipolar articular cavity RF, -26 (-37 to -14)
Conventional bipolar genicular nerve RF, -25 (-34 to -16)
Pulsed monopolar articular cavity RF, -22 (-34 to -10)
Conventional monopolar genicular nerve RF, -20 (-29 to -12)
Pulsed monopolar articular cavity RF combined with intra-articular platelet-rich plasma injection, -18 (-27 to -9)
Intra-articular anaesthesia, -18 (-27 to -8.2)
Pulsed monopolar genicular nerve RF, -16 (-24 to -6.9)
Pulsed monopolar saphenous nerve RF, -12 (-30 to 5.1)
Intra-articular corticosteroid injection, -14 (-24 to -5.5)
Intra-articular sodium hyaluronate injection, -13 (-22 to -4.4)
NSAIDs, -7.5 (-16 to 1.2)
Exercise, -0.5 (-8.3, 9.4)
Placebo, reference
At 6 months, the network was comprised of 9 RCTs (9 interventions) of 821 patients, for a total of 36 paired estimates (10 direct and indirect evidence; 26 indirect evidence only). All treatments had statistically significantly decreased WOMAC compared to exercise at 6 months (placebo was not included in the network). According to the SUCRA, the treatment rankings were as follows (best to worst; MD of exercise vs. treatment [95% CI] shown):
Cooled monopolar genicular nerve RF, 33 (29 to 37)
Pulsed monopolar articular cavity RF combined with intra-articular platelet-rich plasma injection, 30 (27 to 33)
Conventional bipolar genicular nerve RF, 24 (20 to 28)
Conventional monopolar genicular nerve RF, 20 (18 to 22)
Pulsed monopolar articular cavity RF, not reported
Intra-articular corticosteroid injection, 13 (10 to 17)
Intra-articular sodium hyaluronate injection, 9.1 (6.1 to 12)
NSAIDs, 9.7 (7.2 to 12)
Exercise, reference
In the Liu (2022) meta-analysis, there was a statistically significant decrease in WOMAC compared to control (either sham, intra-articular injection, or oral NSAIDs) at 4, 12, and 24 weeks after treatment.
4 weeks (4 studies): WMD -10.64 (95% CI -13.11 to -8.17; I2=1%; p<0.00001)
12 weeks (4 studies): WMD -6.12 (95% CI -7.67 to -4.57; I2=0%; p<0.00001)
24 weeks (4 studies): WMD -10.89 (95% CI -12.28 to -9.51; I2=57%; p<0.00001)
In the Kumaran (2019) RCT of conventional RF vs. sham vs. control, there was a statistically significant main effect for time (within group change) (F [2.2, 91]=18, p<0.001) and a significant interaction between group and time (F[4.3]=2.7, p=0.031). This infers that the type of intervention made a statistically significant difference to WOMAC score, and that there was a statistically significant overall difference between pre, post, 1 month, and 3 months. The decrease in WOMAC score in the RF group was greater than published MCID estimates.
In the Lyman (2022) long-term cohort study, the WOMAC total score for people treated with cooled RF denervation statistically significantly decreased from 64.4 (SD 14.7; n=32) at baseline to 29.3 (SD 25.3; n=32) at 18 months (p<0.0001) and 41.3 (SD 29.9; n=27) at 24 months after treatment (p=0.0007).
Other functional outcomes
Walking ability
One study reported walking ability outcomes.
In the Kumaran (2019) RCT of conventional RF vs. sham vs. control, there was a statistically significant main effect for time (F [2, 85]=15, p<0.001); however, there was no significant interaction between group and time for walking ability as measured by the TUG test. This infers that the type of intervention did not make a statistically significant difference to walking ability.
Knee ROM
One study reported walking ability outcomes.
In the Kumaran (2019) RCT of conventional RF vs. sham vs. control, There was a statistically significant main effect for time (within group change) (F [3, 126)=9.1, p<0.001) and a significant interaction between group and time (F(6)=2.6, p=0.023). This infers that the type of intervention made a statistically significant difference to knee ROM, and that there was a statistically significant overall difference between pre, post, 1 month, and 3 months.
Radiographic outcomes
One study reported radiographic outcomes.
In the Lyman (2022) long-term cohort study, at 24 months, 68.2% (15/22) of patients had no change in Kellgren-Lawrence grade, 22.7% (5/22) showed worsening of 1 grade, and 9.1% of patients (2/22) showed worsening by 2 grades.
Generic Quality of Life measures
EQ-5D
One study reported EQ-5D outcomes.
In the Lyman (2022) long-term cohort study, the total EQ-5D-5L score statistically significantly increased from baseline by 0.15 points at 18 months (p<0.0001) and 0.07 points at 24 months (p=0.0146).
Safety
Rates of complications and major complications
In the Wu (2022) network meta-analysis, adverse events were reported in 6 studies. There were 43 (8.4%) adverse events in 513 patients treated with RF that were probably related to treatment; 20 (3.9%) major adverse events were possibly related to RF.
In the Liu (2022) meta-analysis, there was no statistically significant difference in the risk of adverse events between patients treated with RF denervation and patients treated with control, RD 0.03 (95% CI -0.01 to 0.06; I2=7%; p=0.14; all 3 modalities, 13 studies).
Specific complications
Pain
In the Wu (2022) network meta-analysis, 5 people had pain and 7 people had postprocedural pain. These were major adverse events that were deemed possibly related to RF denervation.
Falls
In the Wu (2022) network meta-analysis, 5 people experienced falls. These were major adverse events that were deemed possibly related to RF denervation.
Stiffness
In the Wu (2022) network meta-analysis, 1 person had stiffness. This was a major adverse event that was deemed possibly related to RF denervation.
Swelling
In the Wu (2022) network meta-analysis, 2 people had swelling. These were major adverse events that were deemed possibly related to RF denervation.
Septic arthritis
In the McCormick (2021) narrative review, there was a report of 1 person who had septic arthritis after treatment with RF.
Pes anserine tendon injury
In the McCormick (2021) narrative review, there was a report of 1 person who had a pes anserine tendon injury after treatment with RF.
Skin burn
In the McCormick (2021) narrative review, there was a report of 1 person who had a skin burn after treatment with RF.
Foot drop
In the Matveev (2021) case report, there was a report of 1 person who had foot drop after treatment with RF.
Vascular injury
In the Jorge (2019) case report, there was a report of 1 person who had a vascular injury after treatment with RF.
Anecdotal and theoretical adverse events
Expert advice was sought from consultants who have been nominated or ratified by their professional Society or Royal College. They were asked if they knew of any other adverse events for this procedure that they had heard about (anecdotal), which were not reported in the literature. They were also asked if they thought there were other adverse events that might possibly occur, even if they have never happened (theoretical).
They listed the following anecdotal adverse events:
Numbness
They listed the following theoretical adverse events:
Osteomyelitis
Injuries to motor nerves
Post ablation neuritis
Charcot neuropathy
Three professional expert questionnaires for this procedure were submitted. Find full details of what the professional experts said about the procedure in the specialist advice questionnaires for this procedure.
Validity and generalisability
Nine studies were included in the key evidence summary, including 1 systematic review and network meta-analysis, 2 systematic reviews and meta-analyses, 1 RCT, 1 long-term cohort study that was an extension of an RCT, 1 cohort study, and 3 case series/reports.
All 3 modalities of RF denervation were captured in the key evidence studies.
Important efficacy data came from the systematic reviews and meta-analyses:
The Wu (2022), Chou (2021), and Li (2021) systematic reviews included a large number of RCTs. Chou (2021) additionally included observational studies.
Chou (2021) found that RF denervation has a statistically significant beneficial effect on osteoarthritic knee pain compared to baseline, at a magnitude greater than published MCID estimates. Furthermore, there was no statistically significant difference in pain relief between the modalities.
Wu (2022) found that whilst all modalities of RF denervation were typically more effective than placebo or exercise, patients responded better to the cooled modality than the conventional and pulsed modalities, and bipolar is more effective than monopolar for pain and function in conventional and pulsed modalities. Several RF modalities saw improvements in pain and WOMAC score greater than published MCID estimates when compared to placebo or exercise.
However, the authors caution that the number of studies, including for the cooled modality, is insufficient.
Liu (2022) found that RF denervation was statistically significantly better than control in several measures of knee pain and function. Interpretation of whether these statistical differences were clinically important was complicated due to the comparator arm consisting of sham and active comparators.
The authors conclude that thought the meta-analysis shows that RF denervation is efficacious, the clinical utility of RF denervation is currently poorly defined and that further studies are required.
Further efficacy evidence came from the Kumaran (2019) RCT. This UK-based study demonstrated statistical and clinical significance over sham and control in improving pain and function. However, this study was limited by small sample size and short follow-up.
The Lyman (2022) cohort study showed that the treatment effect of RF denervation can last up to 2 years. Some attenuation of treatment effect is expected due to nerve regrowth. This study had high attrition rates.
Long-term evidence was generally limited.
The Chen (2021) cohort study gave real-world evidence on the effectiveness of RF denervation, with 61.1% of patients having a positive outcome. Significant prognostic factors for a positive outcome included use of cooled RF and performing multiple lesions at each target.
In the Wu (2022) network meta-analysis, 5 and 4 out of 21 studies were judged to have 'some concerns' or to be at 'high risk' of bias, respectively. The main reason for having some concern of bias was a lack of description of the randomisation process and the selection of the reported result; the main reason for having a high risk of bias was the selection of the reported result.
The studies list various sources of funding, including from companies involved in the manufacture of RF devices. The sources of funding and conflicts of interest were not well described in the systematic reviews.
The following ongoing studies were identified:
Coolief Cooled Radiofrequency vs. Conventional Radiofrequency to Manage OA Knee Pain. Prospective, multicentre, single-blind RCT of cooled RF versus conventional RF. Enrolment: 153 patients. Estimated study completion: August 2022. NCT04145011
Radiofrequency Ablation in the Pain Management of Knee Osteoarthritis. Double-blind RCT of conventional RF versus pulsed RF versus intra-articular steroids. Estimated enrolment: 150 patients. Estimated completion: March 2023. NCT05303766
Ultrasound-Guided Pulsed Radiofrequency In The Treatment Of Patients With Osteoarthritis Knee (USPRFGENOAK). Triple-blind RCT of pulsed RF versus sham. Estimated enrolment: 142 patients. Estimated completion: December 2022. NCT02915120.
Comparison of Conventional and Cooled Radiofrequency of the Genicular Nerves in Patients With Chronic Knee Pain (COGENIUS). Prospective, multicentre, double-blind RCT of conventional RF versus cooled RF versus sham. Estimated enrolment: 400 patients. Estimated completion September 2026. NCT05407610
Innovations in Genicular Outcomes Registry (iGOR). Prospective registry of patients undergoing several pain therapies including RF denervation. Estimated enrolment: 2,000 patients. Estimated completion: March 2025. NCT05495334
How are you taking part in this consultation?
You will not be able to change how you comment later.
You must be signed in to answer questions