Efficacy and safety of pulsed radiofrequency as a method of dorsal root ganglia stimulation for treatment of non-neuropathic pain: a systematic review

Background We systematically reviewed the evidence on the efficacy and safety of dorsal root ganglion (DRG) targeted pulsed radiofrequency (PRF) versus any comparator for treatment of non-neuropathic pain. Methods We searched MEDLINE, CINAHL, Embase, PsycINFO, clinicaltrials.gov and WHO clinical trial register until January 8, 2019. All study designs were eligible. Two authors independently conducted literature screening. Primary outcomes were pain intensity and serious adverse events (SAEs). Secondary outcomes were any other pain-related outcome and any other safety outcome that was reported. We assessed the risk of bias using the Cochrane tool and Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I). We conducted narrative evidence synthesis and assessed the conclusiveness of included studies regarding efficacy and safety. Results We included 17 studies with 599 participants, which analyzed various pain syndromes. Two studies were randomized controlled trials; both included participants with low back pain (LBP). Non-randomized studies included patients with the following indications: LBP, postsurgical pain, pain associated with herpes zoster, cervicogenic headache, complex regional pain syndrome type 1, intractable vertebral metastatic pain, chronic scrotal and inguinal pain, occipital radiating pain in rheumatoid arthritis and chronic migraine. In these studies, the PRF was usually initiated after other treatments have failed. Eleven studies had positive conclusive statements (11/17) about efficacy; the remaining had positive inconclusive statements. Only three studies provided conclusiveness of evidence statements regarding safety – two indicated that the evidence was positive conclusive, and one positive inconclusive. The risk of bias was predominantly unclear in randomized and serious in non-randomized studies. Conclusion Poor quality and few participants characterize evidence about benefits and harms of DRG PRF in patients with non-neuropathic pain. Results from available studies should only be considered preliminary. Not all studies have reported data regarding the safety of the intervention, but those that did, indicate that the intervention is relatively safe. As the procedure is non-destructive and early results are promising, further comparative studies about PRF in non-neuropathic pain syndromes would be welcomed.


Background
Chronic pain is one of the major public health issues worldwide and is one of the leading causes of years lived with disability [1]. Estimates on the prevalence of chronic pain in the general population vary, ranging from 11% [2] up to 64% [3]. These different estimates are mostly due to differences in the definition of chronic pain regarding the duration of symptoms (3 vs. 6 months) and the wording of questions used for assessing chronic pain [4]. Besides its major clinical impact and costs for the healthcare system, chronic pain impairs patients' quality of life, as well as their ability to work and function, causing massive indirect socioeconomic costs worldwide [5]. Chronic pain asserts this major impact on individuals, health systems and society because of inadequate treatment modalities.
Pulsed radiofrequency (PRF) emerged as a therapeutic treatment for various painful conditions, including both neuropathic and non-neuropathic pain [6][7][8]. PRF has been described as "a non-neurodestructive therapy in pain management " [9]. PRF is a minimally invasive intervention, which involves the application of pulses of electric current, created at the tip of an electrode, without a harmful increase in the temperature [9].
It has been suggested that a dorsal root ganglion (DRG) is a desirable target for the treatment of pain [10]. PRF application close to dorsal root could alleviate neuropathic pain [11]. However, we have observed an increasing number of studies on chronic pain, reporting use of DRG targeted PRF treatment of non-neuropathic pain in humans. Therefore, we aimed to conduct a systematic review about the evidence on the efficacy and safety of DRG targeted PRF treatment of nonneuropathic pain.

Study design
We published a systematic review protocol a priori in the PROSPERO database (registration number: CRD42017076502). Since the original protocol covered extremely wide scope and heterogeneous interventions, subsequently we divided the original protocol into a separate assessment of DRG targeted electrical field stimulation (EFS) [12] and PRF. The systematic review was performed following the PRISMA statement and Center for Reviews and Dissemination (CRD) manuals.

Eligibility criteria Participants, intervention and study designs
We included primary studies with participants suffering from various painful conditions which are not currently classified as purely of neuropathic origin (i.e. nonneuropathic pain). In case that condition was defined of both origins, neuropathic and non-neuropathic, such as post-surgical pain or low back pain we included such condition. We excluded studies where PRF treatment was used for neuropathic pain as it is defined in the guidelines of the International Association for the Study of Pain (IASP). We used the IASP classification of chronic pain for ICD-11. We chose to include both randomized controlled trials (RCTs) as well as nonrandomized study designs (NRSDs) because we expected a few RCTs in this research area, and we wanted to provide a comprehensive picture of evidence in this field of research. We used Cochrane Handbook for Reviews of Interventions to define the design of included studies. Manuscripts that included more than 10 participants were classified as case series, while those that included less than 10 participants were defined as case reports [13]. We only included studies where PRF treatment was directed to the DRG, including a combination of PRF with other therapies. If the study only reported results about efficacy, and safety was not reported, we still included such a study to get comprehensive evidence synthesis regarding efficacy.

Outcome measures
Primary outcomes were: pain intensity and serious adverse events (SAEs). For primary outcome, we reported any outcome measures, as reported in included manuscripts. Secondary outcomes for efficacy were any other pain-related outcomes, and for safety any other safety data, including non-serious adverse events and other complications regarding tested intervention.

Search strategy and information source
We searched four databases: MEDLINE via Ovid, Embase via Ovid, CINAHL and PsycINFO via EBSCOhost (Supplementary Table 1). Databases were searched from the date of inception until January 8, 2019 with no restriction regarding the language. Records were then exported to the EndNote X5 citation software (Clarivate Analytics, Boston, MA, USA) and duplicates removed. Furthermore, reference lists of all included studies and their citations were downloaded from Web of Science and screened to find additional eligible studies. ClinicalTrials.gov and World Health Organization's International Clinical Trial Registry Platform (WHO ICTRP) were searched to identify ongoing studies.

Study selection
Reviewers independently screened each title/abstract of retrieved records as well as full-texts of retrieved studies for possible inclusion (authors LFH, IV, TM and SD participated in screening). Discrepancies were resolved by another author (DS).

Data extraction
Independent data extraction was performed by two authors for each data point (authors: IV, and TM or KV). We extracted the following data: the surname of the first author, year of publication, study design, details about intervention (treatment protocol and device used), comparator, inclusion and exclusion criteria, number of participants, baseline characteristics of participants, followup period, DRG level treated and outcomes about efficacy and safety.

Risk of bias assessment
We used the Cochrane Risk of Bias (RoB) tool (version from 2011) to assess RoB in RCTs and the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool for cohort type studies. RoB was analyzed independently by two authors (IV, and SD or KV). Discrepancies were resolved by another author (LP).

Synthesis of results
Due to the heterogeneity of included studies, it was not possible to conduct a meta-analysis, even though we have planned to do it in our study protocol. For this reason, we conducted a narrative and tabular synthesis of results. We also conducted an analysis of conclusiveness about efficacy and safety of the treatment in the abstracts of included studies. We divided conclusiveness statements into five categories: positive conclusive (favorable conclusion in favor of PRF), positive inconclusive (favorable conclusion, but with a note about insufficient or low quality evidence), negative conclusive (PRF not beneficial), negative inconclusive (PRF not beneficial, but with a note about insufficient or low quality evidence) and not reported.

Results
The flow chart in Fig. 1 shows the number of records analyzed in each screening phase. We screened 63 manuscripts in full text, and we finally included 17 manuscripts in this systematic review. Excluded studies, and reasons for their exclusion, are listed in Supplementary  Table 2. The characteristics of the included studies are detailed in Table 1.

Low back pain
In this group, there were 5 studies with a total of 328 participants, including two RCTs with 28 participants in one [14] and 60 participants in another one [15], one retrospective cohort study including 29 participants [16], and two before and after comparisons with 84 participants in one [17] and 127 participants in another [18].
Trial by Holanda et al. [14] included 28 participants which were randomized in three groups: PRF treatment group with the probe directed through the needle in the second lumbar intervertebral foramen (N = 11), lidocaine injection group (N = 7) and laser irradiation treatment group (N = 10). All participants from the lidocaine injection group and laser irradiation group reported a 100% reduction in visual analogue scale (VAS) scores immediately after the treatment, while participants from the PRF group reported a 62.5% reduction in pain. At 1month follow-up laser irradiation group had a 55.5% reduction in pain; lidocaine injection group 62.5% reduction and PRF group only 20% [14].
An RCT by Lee et al. [15] analyzed predictive value and cost-effectiveness of the use of diagnostic blocks before PRF treatment. They included 60 participants suffering from LBP with or without lower-limb pain, randomized into two groups. In one group (N = 30) participants received DRG blocks with 1 ml of 2% bupivacaine and 1 ml of 2% triamcinolone, and those who had at least 50% improvement were scheduled for PRF treatment. The other group (N = 30) received only PRF treatment without DRG blocks. Limited low back pain was treated with DRG block or PRF applied to the L2 DRG; lower -limb pain was treated with PRF applied to the L3-S1 DRG. The authors concluded that DRG blocks had no statistically significant impact on the results of PRF treatment, while their application resulted in overall higher medical costs [15].
Yang et al. [16] reported results of a retrospective cohort study that aimed to develop a patient-mounted navigated intervention (PaMNI) system for spinal diseases to evaluate the success of the PRF treatment. The study also included a pilot clinical trial were the new PaMNI system (N = 16) was compared to conventional fluoroscopy (N = 13). In all patients, PRF treatment was delivered on the L4 DRG. Both groups showed a reduction in VAS scores 1 month after the treatment with no statistically significant difference between groups (P = 0.238). However, the study showed the feasibility and efficacy of the PaMNI system [16]. Before and after comparison by Hsu et al. [17] followed 84 participants up to 3 years to investigate the correlation between different types of lumbar lordosis with the outcomes of PRF treatment applied to L2 DRG in chronic low back pain. The analysis showed that after 3-year follow-up participants had a statistically significant reduction in low back pain, regardless of the type of lumbar lordosis [17]. The study by Tsou et al. [18], also followed participants for up to 3 years. They included participants who had low back pain with lower -limb pain (N = 78) or without it (N = 49). LBP was treated with PRF applied to the L2 DRG and lower-limb pain was treated with PRF applied to the L3-S1 DRG. Percentage of participants achieving at least 50% improvement in VAS scores was similar in both groups at 1-year follow-up, with 20 out of 45 participants (44.44%) in the group without lower -limb pain and 34 out of 74 participants (45.95%) in the group with lower -limb pain [18].
None of the studies reported serious adverse events. Two studies reported minor complications: mild discomfort    Information not given.

Chronic migraine
Li 2018 [29] No comparator    -40-years-old woman pain lasting for 5 years Patient 2: -66-years-old women pain lasting for 1 year Complex regional pain syndrome Albayrak 2016 [24] no improvement with the combined use of medical therapy, physical therapy, and the rehabilitation program during the procedure [14] and leakage of the cerebrospinal fluid [17]. One study reported that there were no complications [18]. Two studies from this group did not report any outcomes regarding safety [15,16], but one of them provided a general warning about the radiation dose exposure [16].
In this group all studies reported positive statements regarding the efficacy of the treatment, four studies had positive conclusive statements [14,[16][17][18] while one study had positive inconclusive statement [15]. Only one study reported a positive conclusive statement about safety [18], one reported only specific adverse events that occurred [17], while others did not report any conclusion statements (Table 1 and Supplementary Table 3).

Post-surgical pain
Three studies explored PRF in postsurgical pain, with a total of 188 participants. In a cohort study of Albayrak et al. [19] there were 39 participants with postsurgical pain after total knee arthroplasty. In another cohort study, Cohen et al. [20] included 49 participants suffering from thoracic postsurgical pain. Fam et al. [21] included 100 women suffering from intercostobrachial neuralgia (ICBN) postmastectomy in a study designed as before and after comparison. Despite different etiology of postsurgical pain the majority of participants experienced a reduction in pain after the treatment (details are given in Table 1).
One participant from the study of Cohen et al. [20] had a serious adverse event that could not be related to procedure or treatment. Small pneumothorax was found during a routine scan after the PRF procedure. This participant was treated conventionally and monitored [20]. Pain at the site of the procedure was reported as a mild complication [21]. The third study reported that complications were not observed [19].
Two studies from this group reported positive conclusive statement for efficacy, while the conclusion for safety was not reported [19,20]. The study by Fam et al. [21] reported positive inconclusive statements for both efficacy and safety [21] (Table 1 and Supplementary  Table 3).

Pain associated with herpes zoster
A retrospective cohort study by Kim et al. [6] with 42 participants addressed PRF of DRG for pain associated with herpes zoster but before post-herpetic neuralgia (PHN) was established. The study analyzed two groups of participants; one received continuous epidural block (N = 22), and the other received PRF treatment (N = 20) after the acute phase of herpes zoster, but before it was well established, meaning between 30 and 180 days of the herpes zoster diagnosis. Participants from the continuous epidural block group received 0.187% ropivacaine at the rate of 1 ml per hour, while concentration and rate of administration depended on pain relief and adverse effects (mean concentration of ropivacaine and infusion rates used were 0.22 ± 0.07% and 1.82 ± 0.65 ml/hr). When satisfactory pain relief was achieved catheter was removed. Reduction in pain was significantly higher in the PRF group compared to a continuous epidural block group (P = 0.029) up to 6 months after the treatment [6]. From the safety aspect, only procedural pain was reported [6]. The study abstract had a positive conclusive statement about efficacy, while safety conclusion was not reported [6] (Table 1 and Supplementary Table 3).

Cervicogenic headache
The before and after comparison by van Zundert et al. [22] included 18 participants, of which 14 had pain related to non-neuropathic origin (their characteristics were reported separately in Table 1). Participants were followed for a mean time of 19.4 months (maximum follow-up time 2.5 years) [22]. Before study inclusion, participants received diagnostic nerve blocks with 0.5 mL of 2% lidocaine. Treatment outcomes were scored using a 7-point Likert scale.
Participants who had at least 50% pain relief were included in the study and received PRF treatment. Successful PRF treatment was defined as 6 (≥ 50% improvement) or 7 (≥ 75% improvement) points on 7point Likert scale (Global Perceived Effect good or very good). Participants from the group of non-neuropathic pain origin had successful treatment in 9 cases while treatment was not successful in 5 cases. The case report by Zhang et al. [23] described 2 participants who reported 100% pain relief lasting for 6 months after the treatment. Both studies reported that no complications occurred ( Table 1).
The study by van Zundert et al. [22] reported positive conclusive statements about both, safety and efficacy, while Zhang et al. [23] reported positive inconclusive statement about efficacy, while safety was not reported (Table 1 and Supplementary Table 3).

Complex regional pain syndrome
This group included only two case reports [24,25] with three participants included. Albayrak et al. [24] reported cases of two women with post-stroke complex regional pain syndrome (CRPS). Both patients used multiple treatment modalities before the PRF treatment, including medical therapy, physical therapy, rehabilitation program and transcutaneous electrical nerve stimulation (TENS). After PRF treatment, both participants had an immediate resolution of their symptoms that lasted up to 5 and 10 months which were final follow-up time points [24].
Apiliogullari et al. [25] reported a case of a 16-year-old girl suffering from CRPS due to sequelae of poliomyelitis, who did not respond to non-steroidal antiinflammatory drugs. However, after two PRF treatments (first applied at L5 and repeated after 2 weeks at L4 DRG) she reported immediate pain relief, with VAS scores going from 100 points down to 10, this effect remained for over 6 months of follow-up [25]. Both studies reported that no complications occurred (Table  1).
Both studies from this group reported positive inconclusive statements about efficacy, while the conclusion about safety was not reported [24,25] (Table 1 and Supplementary Table 3).

Intractable vertebral metastatic pain
The case series of Arai et al. [26] included 15 cases with vertebral metastatic pain, which demonstrated pain relief, defined as a 50% pain reduction from baseline values. Values on the numerical rating scale (NRS), measured during rest and upon movement, were significantly lower 3 weeks after the PRF treatment (P < 0.0001) [26]. From the safety aspect, there were no SAEs or other complications ( Table 1). The study reported positive conclusive statements about efficacy, while conclusion about safety was not reported [26] (Table 1 and  Supplementary Table 3).

Chronic scrotal and inguinal pain
Hofmeester et al. [27] reported the first case of using PRF to treat scrotal and inguinal pain after orchidopexy in a 13-year boy. PRF was performed at three levels (T12 -L2) after other treatment modalities have failed. The PRF of DRG led to an immediate and lasting pain alleviation of more than 70% as reported by the patient [27]. Information about safety was not reported. The study reported positive conclusive statements about efficacy, while the conclusion about safety was not reported [27] ( Table 1 and Supplementary Table 3).

Occipital radiating pain in rheumatoid arthritis
Lee et al. [28] reported PRF of the C2 DRG to treat occipital radiating headache in a 74-year old woman with rheumatoid arthritis. The patient has not complained of any occipital radiculopathy for 6 months, and the posterior neck pain has since been reduced to a visual analogue scale (VAS) score of three, from initial 6/10. Information about safety were not reported [28]. This study also reported positive conclusive statements about efficacy, while the conclusion about safety was not reported [28] (Table 1 and Supplementary Table 3).

Chronic migraine
Li et al. [29] reported a case of a 34-year old woman who suffered from chronic migraine with occipital pain. She underwent PRF treatment after the failure of other treatment modalities. The patient had complete pain relief with no symptoms 1 year after the treatment [29]. Details are given in Table 1. The study did not report conclusion about safety, while the conclusion about efficacy was positive inconclusive [29] (Table 1 and Supplementary Table 3).

Parameters of PRF treatment
Low back pain was a painful condition which had the most different treatment parameters among included studies, with a range of different values for amplitude (45 and 100 V), frequency (2 and 50 Hz) and duration of treatment (120, 240 and 300 s). Pulse width was only reported in one study [14] ( Table 2). In other studies parameters were similar, the majority had a pulse width of 20 ms, the amplitude of 45 V, frequency of 2 Hz and duration of 120 s ( Table 2). The temperature at the electrode tip was constant parameter, same in all studies, and set to 42°C in order to avoid tissue damage.

Participants' inclusion criteria
More than a half of included studies were before and after comparisons, case series or case reports where participants were included and scheduled for PRF treatment after failure of other treatment modalities and as a last treatment option (Table 3). On the other side, higherquality studies, such as RCTs and cohort type studies had clearly defined inclusion and exclusion criteria as well as described participants' baseline characteristics (Table 3).

Summary on the conclusiveness of the evidence
Among 17 included studies, 11 studies had positive conclusive statements about efficacy; the remaining had positive inconclusive statements. The majority of the studies did not provide conclusive statements regarding safety in the manuscript abstracts. Only three studies provided safety conclusiveness statementstwo indicated that the evidence was positive conclusive, and one positive inconclusive (Table 1 and Supplementary Table  3).

Risk of bias
Analysis of two included RCTs, with Cochrane Rob tool, indicated that the majority of the domains were judged with unclear RoB due to insufficient information about the used methodology (Supplementary Table 4, Fig. 2). Four non-randomized studies were eligible for assessment with ROBINS-I. The most common judgment for analyzed domains (12 domains out of 28 domains judged for these four studies) was serious RoB. Ten domains were judged with moderate RoB, and only 6 domains with low RoB (Supplementary Table 5, Fig. 2).

Studies awaiting classification
One RCT, which aims to study DRG thermal RF versus PRF for metastatic pain in the thoracic vertebral body on 69 participants, is classified as completed on July 30, 2018. Results were reported to Clinical Trials.gov but were returned to the authors after the quality control review so results are still not publicly available (NCT03204942). A trial that aimed to study superior hypogastric plexus block versus PRF for chronic pelvic cancer pain on 40 participants is classified as 'Not yet recruiting' since June 26, 2018 (NCT03228316). Studies awaiting classification are described in Supplementary  Table 6.

Discussion
In this systematic review, we included 17 studies about the treatment of several non-neuropathic chronic pain conditions with PRF directed to DRG. All studies presented positive conclusions (both conclusive and inconclusive) about the efficacy of the treatment. However, the studies were mostly non-randomized, with small sample sizes, and issues related to the risk of bias. Therefore, their results should only be considered as preliminary.
PRF was developed as a less destructive pain relief modality alternative to conventional radiofrequency (CRF) which can selectively block delta and C fibers [30]. The first report about the clinical effects of PRF on DRG was published relatively recently, in 1998. Due to its theoretical benefits, it was postulated that PRF could be particularly helpful in neuropathic pain [31]. However, we have observed in the literature that clinicians and researchers apply PRF to non-neuropathic chronic pain as well. Despite the number of studies found in the literature about the treatment of non-neuropathic chronic pain in humans with PRF, their findings cannot be generalized. In the studies that we have found, the PRF was usually initiated after other treatments have failed. We reported a similar issue in our recent systematic review in which we studied the efficacy and safety of EFS of DRG [12]. In that systematic review, we found only one RCT among 29 included studies; most of the studies were low-level of evidencenon-randomized study designs, including case series and case reports. The review about EFS of DRG also included few participants, with a median of 6 participants per study [12]. In this systematic review there were 17 included studies, with a median of 28 participants.
The paucity of large and high-quality studies in the field of DRG neuromodulation is likely due to the relative novelty of this approach for the treatment of pain. In this systematic review, about PRF of DRG in nonneuropathic pain, only two of 17 included studies were RCTs, and RoB judgment for the majority of their methodological aspects was unclear. Likewise, the most common assessment in non-randomized studies assessed with the ROBINS-I tool was that there was a serious risk of bias. Besides their suboptimal methodological reporting, the analyzed studies were relatively small. Even the two included randomized controlled trials were small; one included a total of 28 patients in 3 groups, and the other one included 60 patients in two groups.
The highest number of studies was found for the low back pain indication. However, we were not able to perform a meta-analysis due to clinical heterogeneity of the studies, as can be seen from characteristics of included studies, different comparators used in included trials, and different stimulation parameters. Differences in treatment approaches can result in different clinical outcomes.
Despite the low level of evidence, all of the analyzed studies sent positive conclusions to the research community in their abstracts. The majority of these conclusions were conclusive, i.e. they did not mention the need to conduct further studies on this subject. Despite the authors' positive conclusions regarding the tested interventions, caution is needed when advising DRG targeted PRF to chronic pain patients, because of the paucity of high-quality and high-level evidence. This intervention should be tested in large-scale, high-quality RCTs to truly test whether the intervention has expected benefits and harms. Until then, these studies should be treated as preliminary evidence only.
A broad focus of this systematic review could be considered as a limitation of this review, as we included any pain condition that fits the IASP criteria of nonneuropathic pain. Furthermore, we acknowledge that the examined studies included patients with various clinical conditions, and thus there is a possibility that the effectiveness of the treatment depends on underlying pathogenic mechanisms. However, as can be observed from our results, there were very few studies in each group of indications; the highest number of studies (five) was found for low back pain. Therefore, focusing on every single one of these indications in a separate systematic review would result in a high number of systematic reviews, with minimal results included. Furthermore, with this approach, we are giving readers a very wide and informative picture of all the non-neuropathic pain conditions that were reported in the literature as treated with DRG targeted PRF.
We have used IASP classification for definitions of non-neuropathic pain; these classifications are evolving and changing, so the included conditions may be categorized differently, depending on the time of categorization and reference classification used. Previous versions of chronic pain classification were to some extent insufficient for chronic neuropathic pain conditions since some conditions were not defined properly or were missing so we decided to use the updated version of classification since it is crucial to get the comprehensive evidence synthesis. According to the newest IASP classification that we used (ICD-11) when deciding about study inclusion we might have included some studies that in previous versions of classification were classified either as neuropathic pain or as the pain of mixed origin. We have included CRPS 1 [25], which is not considered neuropathic pain. In the study of Kim et al. [6] the authors studied the effects of DRG PRF beyond the acute phase of zoster, bur before PHN was well established (from 30 days to 180 days after zoster onset). The study of van Zundert [22] has excluded "signs of radicular compression".
It has been questioned before what is the value of systematic review including poor evidence and small studies [32]. However, such systematic reviews are valuable because they are highlighting the paucity of evidence and the low quality of available information [33]. Our systematic review is such a case. We even included two case reports with only one participant which may be considered anecdotal rather than firm evidence. It could be argued that such studies should not even be included in systematic reviews; however, we did not set any restrictions regarding number of participants in our study eligibility criteria. By showing that many clinicians and researchers have published small studies, with low-level evidence, about potential benefits of PRF in chronic non-neuropathic pain, we hope that trialists will be inspired to explore this intervention in studies that are considered high-level evidence.