Introduction

Chronic low-back pain (LBP) is related to disability and work absence and accounts for high social and health care costs in western societies [1]. The management of chronic LBP comprises a range of different intervention strategies including surgery, pharmacological interventions and non-medical interventions such as exercise, behavioural therapy and alternative therapies. Non-surgical interventions, such as injection therapy and denervation procedures, have rapidly increasing rates of utilization and associated costs [2]. Injection therapy for chronic LBP involves injections of medications, irritants, or proteolytic enzymes into soft tissues outside or within the spine. Denervation procedures involve the application of various types of thermal or radiofrequency energy within the spine.

Over the past years, a substantial number of randomized clinical trials have been published on injection therapy and denervation procedures. In order to appropriately evaluate the effects of these interventions in a systematic review, it is important to make valid comparisons that are homogeneous with regard to study population, indications, interventions and different anatomical structures being targeted, controls, outcome measures and follow-up. Based on the available literature, this overview presents the current evidence on injections into and outside the spine and radiofrequency or thermal denervation procedures for non-specific chronic LBP.

Objectives

The objective of this review was to determine the effectiveness of injection therapy (including injections into soft tissues outside or within the spine and chemonucleolysis), and radiofrequency or thermal denervation procedures for chronic LBP.

Criteria for considering studies for this review

Types of studies

Only randomized controlled trials with at least 1 day of follow-up were considered in this systematic review.

Types of participants

In order to be included in this review, participants of the randomized controlled trials (RCTs) must fulfil the following inclusion criteria: adult subjects (≥18 years of age) with chronic (>12 weeks duration) LBP (including subjects with radiculopathy or any other non-specific degenerative pathology, such as osteoarthritis). Exclusion criteria were (1) trials including subjects with specific LBP caused by pathological entities, such as vertebral spinal stenosis, ankylosing spondylitis, scoliosis, or coccydynia; (2) post-partum LBP or pelvic pain due to pregnancy; (3) post-operative studies; (4) prevention studies; and (5) abstracts or non-published studies.

Types of interventions

RCTs studying the following interventions were included in this overview: injection of medications or proteolytic enzymes (chemonucleolysis); and radiofrequency or thermal denervation procedures. All anatomical sites subject to injection therapy (intervertebral disc, facet joint, epidural space, intramuscular) and denervation procedures (facet joint, intervertebral disc, spinal nerves) for LBP were included. Additional interventions were allowed in all studies if there was a contrast for the injection therapy or denervation procedure in the study.

Types of outcome measures

To be included at least one of the following outcome measures should have been measured in the RCT: pain intensity [e.g. visual analog scale (VAS), numerical rating scale (NRS), McGill pain questionnaire], back specific functional status [e.g. Roland–Morris Disability Questionnaire, Oswestry Disability Index (ODI)], perceived recovery (e.g. overall improvement), and return to work (e.g. return to work status, sick leave days). The primary outcomes for this review were pain and functional status.

Search methods for identification of studies

Existing systematic reviews for the interventions were screened for studies fulfilling the inclusion criteria [3–6]. Then, the literature was searched in the Cochrane Back Review Group (CBRG) trial register from the last date onward for each of the interventions up to November 17, 2009.

References of relevant studies were screened and experts were approached in order to identify additional primary studies not identified in the previous steps. The language was limited to English, Dutch, French and German, because these are the languages that the authors are able to read and understand. The search strategy outlined by the CBRG was followed [7].

Methods of the review

Study selection

Two authors (NH, TK) independently screened the abstracts and titles retrieved by the search strategy and applied the inclusion criteria to all relevant abstracts. The full text version of an article was obtained if the title and abstract seemed to fulfill the inclusion criteria or if eligibility of the study was unclear. All full text articles from the existing reviews were compiled and independently screened for inclusion criteria by the authors. Any disagreements on study eligibility were resolved by discussion and a consensus meeting.

Risk of bias assessment

Two authors (NH, TK) independently assessed the risk of bias (RoB) of all eligible studies using the criteria list advised by the CBRG, which consists of 11 items [7]. Items were scored as ‘positive’ if they fulfilled the criteria, as ‘negative’ when there was a clear RoB, and as ‘inconclusive’ if there was insufficient information. Differences in assessment were discussed during a consensus meeting. A total score was computed by adding the number of positive scores, and low RoB was defined as fulfilling six or more (more than 50%) of the 11 internal validity criteria. Empirical evidence has shown that studies fulfilling <6 items report higher treatment effects than studies fulfilling 6 or more items [8].

Data extraction

A standardized form was used for data extraction which included collection of descriptive data on the study population and the type of intervention, as well as quantitative data regarding the outcome measures. Data on the characteristics of the study population (gender, age), type of therapy and control treatment, adverse events and complications were also collected.

Data analysis

If studies were clinically homogeneous regarding study population, types of treatment, outcomes and measurement instruments, a meta-analysis was performed. If possible, the weighted mean difference (WMD) was calculated because this improves the interpretability of the results. If a WMD was not possible the standardized mean difference (SMD) was calculated. If trials reported outcomes as graphs, the mean scores and standard deviations (SD) were estimated from these graphs. If SD were not reported, they were calculated using the reported values of the confidence intervals, if possible. If the SD of the baseline score was reported, we used the ratio between the baseline score and SD to calculate the SD for other follow-up moments. Finally, if none of these data were reported, an estimation of the SD was based on study data (population and score) of other studies. In order to correct for error introduced by “double-counting” of subjects of “shared” interventions (i.e. 2 comparisons within 1 study that used the same control group as contrast) in the meta-analyses, the number of subjects in the control group was divided by the number of comparisons that this one study added in the meta-analyses. For the comparisons where studies were too heterogeneous, no meta-analysis was performed.

Quality of the evidence

Grades of recommendation, assessment, development and evaluation (GRADE) profiles were used to evaluate the overall quality of the evidence and the strength of the recommendations [9]. The quality of the evidence for a specific outcome was based upon five principal factors: (1) limitations (for example due to study design), (2) inconsistency of results, (3) indirectness (e.g. generalizability of the findings), (4) imprecision (e.g. sufficient data) and (5) other considerations, such as reporting bias. The overall quality was considered to be high when multiple RCTs with a low RoB provide consistent, generalizable, and precise data for a particular outcome. The quality of the evidence was downgraded by one level when one of the factors described above was not met [9]. Single studies were considered inconsistent and imprecise (i.e. sparse data) and provide “low quality evidence”, which could be further downgraded to “very low quality evidence” if there were also limitations in design or indirectness. The following grading of quality of the evidence was applied [10]:

High quality:

Further research is very unlikely to change our confidence in the estimate of effect.

Moderate quality:

Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low quality:

Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low quality:

We are very uncertain about the estimate.

Results

Description of studies

Study selection

From three existing reviews a total of 83 references were screened for eligibility. Of these 83 articles, 11 studies fulfilled the inclusion criteria and were included in this overview (Fig. 1). The most important reason for exclusion was inclusion of acute or sub-acute patients in the study or lack of information about the duration of the complaints prior to the study. Additionally, 14 potentially relevant titles and abstracts were identified in the CBRG trial register for injection-related interventions and screened for potential inclusion. All of these abstracts fulfilled the inclusion criteria and were therefore included. Two additional studies were found by forward citation searching of included studies.

Fig. 1
figure 1

Flow diagram of selection process of studies on injection therapy for chronic LBP

Full size image

In total 27 studies were included in this review. The study characteristics of all included (Appendix 1) and excluded studies (Appendix 2) are summarized in appendices.

Injection therapy

Intervertebral disc

One RCT was identified which evaluated injection therapy aimed at the intervertebral disc in chronic LBP patients [11]. The authors compared chemonucleolysis (intradiscal injection of chymopapain which digests the nucleus pulposus) to discectomy in patients with disc herniation which was confirmed by myelography [11]. This trial enrolled patients with both sub-acute and chronic LBP, but presented selected outcomes for the subgroup of patients (n = 58) with a duration of symptoms >6 months.

Zygapophysial joints (facet joints)

Eight trials, three of which were placebo controlled [12–14], evaluated intra-articular or peri-articular facet joint injections with corticosteroids (methyl-prednisolone, triamcinolone, or betamethasone) [12, 15], local anaesthetic (lidocaine, bupivacaine, or lignocaine) [14], sodium hyaluronate [16], Sarapin (a suspension of powdered pitcher plant in alkaline solution) [17], or a combination of these [13, 18, 19]. Three of these trials confirmed that pain was originating from the facet joint by including only patients with a positive response to local anaesthetic nerve block [12, 17, 18]. One trial used radiographic confirmation of facet joint arthritis as an inclusion criterion [16] while the remaining four trials used clinical criteria (e.g. unilateral pain, pain with sustained postures) to identify patients with facet joint pain [13–15, 19].

Epidural space

Three RCTs were identified that evaluated injections of corticosteroids [20, 21] or anaesthetic [22] into the epidural space. There were no placebo-controlled trials of injections into the epidural space. Comparisons in the included trials were made with benzodiazepine [21], spinal endoscopy [20] or between ropivacaine and bupivacaine [22]. One study excluded patients with a previous history of spinal surgery [20].

Spinal muscles (local injections)

Two RCTs were identified that evaluated local intramuscular injections for non-specific chronic LBP. One study compared injections of vitamin B12 with placebo [23], while the other compared injections of botulinum toxin A with placebo [24].

Denervation procedures

Intervertebral disc

Six RCTs were identified which evaluated denervation procedures [25–30] for chronic LBP targeted at or around the intervertebral disc. Five of the trials only included patients with positive responses to either analgesic [25] or provocative [26–29] discography.

Three trials evaluated the use of percutaneous intradiscal radiofrequency thermocoagulation (PIRFT). PIRFT involves the placement of an electrode or catheter into the intervertebral disc and applying an alternating radiofrequency current to reduce nociceptive input from the disc. Two of the trials were placebo controlled [25, 28], while the other compared high- with low-intensity PIRFT [26].

Two trials were identified which compared the use of intradiscal electrothermal therapy (IDET) with a placebo [27, 29]. IDET is similar to PIRFT and involves the insertion of an electrode into the annulus or nucleus of the intervertebral disc and application of electrothermal energy to alter the pain receptors.

One RCT evaluated radiofrequency denervation of the ramus communicans nerve compared with placebo denervation in patients who failed to respond to IDET [30]. This denervation is performed outside the intervertebral disc, unlike IDET and PIRFT.

Zygapophysial joints (facet joints)

Six trials compared radiofrequency denervation of the medial branch of the dorsal ramus with a sham denervation procedure [31–36], and one trial compared radiofrequency denervation of the dorsal root ganglion with a sham procedure [37]. The trial by Tekin and colleagues [34] included comparisons between conventional, pulsed, and sham radiofrequency denervation. All trials except one [32] included only patients with a positive response (~50–80% pain relief) to selective anaesthetic nerve blocks. The trial by Leclaire and colleagues [32] included patients who experienced significant relief of their LBP for at least 24 h during the week after intra-articular facet injections under fluoroscopy. This inclusion criterion was considered to be insufficiently sensitive and likely to have included patients with pain of non-facet joint origin [32]. Therefore, this trial was not included in the primary analysis, but a sensitivity analysis was performed to evaluate the addition of this study to the meta-analyses.

Risk of bias in included studies

The results of the RoB assessment are shown in Table 1. 18 studies (66%) had a low RoB. All studies were described as randomized; however, only nine studies (33%) used an adequate randomization procedure in combination with an adequate concealment of treatment allocation. In only four studies (15%) co-interventions were avoided or similar between groups. Many studies had acceptable compliance (20 studies; 77%) or acceptable drop-out rates (21 studies; 77%) or both (15 studies; 58%).

Table 1 Risk of bias assessment of injection therapy studies
Full size table

Effects of intervention

Feasibility of statistical pooling

As stated in the methods section, statistical pooling was only considered if subgroups of studies were clinically homogeneous, and the authors provided sufficient information on study characteristics, outcome measures, and study results. After reviewing the included study characteristics, only two treatment subgroups (IDET vs. placebo and facet joint denervation vs. placebo) were sufficiently clinically homogeneous to perform statistical pooling (Table 2).

Table 2 Pooled effect estimates for injection therapy and denervation procedures
Full size table

Injection therapy

Intervertebral disc

Chemonucleolysis versus other treatment

In one RCT with a high RoB, chemonucleolysis (via intradiscal chymopapain injection) was compared with discectomy [11]. A subgroup of 68 patients had duration of symptoms >6 months prior to the start of the trial. There was no difference between groups in percentage of patients whose pain had disappeared or improved (87 vs. 85%) 12 months post-treatment. Within 12 months post-treatment, 25% of the chemonucleolysis group were considered treatment failures and underwent surgery.

There is very low quality evidence (1 RCT; n = 68; limitations in design, inconsistency, imprecision) that chemonucleolysis is no more effective than discectomy over a long-term follow-up.

Zygapophyseal joint (facet joint)

Facet joint injections with corticosteroids versus placebo

Two RCTs, one with low RoB [12] and one with high RoB [13], compared the effects of facet joint injections with corticosteroids to placebo injections. There was insufficient data on pain and functional status in the Lilius study [13] to allow for statistical pooling of outcomes. In the Carette study [12], no significant differences were found between the groups at 1 and 3 months for pain, functional status or self-rated improvement. At the 6-month follow-up, significant differences were found in favour of the corticosteroid group [12]. The high RoB study [13] compared intra-articular and peri-capsular corticosteroid injections with placebo injections. No significant differences between the groups were reported for pain, disability or work attendance at either short- or intermediate-term follow-ups [13]. No side effects apart from transient pain were reported.

There is very low quality evidence (1 RCT; n = 97; inconsistency, indirectness, imprecision) that there is no significant difference in effect between facet joint injections with corticosteroids and placebo injections for short- to intermediate-term pain relief and improvement of function.

Facet joint injections with corticosteroids versus other treatment

Five RCTs compared the effects of corticosteroids injections into and around the facet joints with other treatments [15–19]. Because of the clinical heterogeneity of the reference treatments, pooling was determined to be unsuitable.

In a study with low RoB [19], intra-articular facet joint injections with corticosteroids and lignocaine were compared with facet nerve blocks using similar medication. The facet joint injections provided slightly better pain relief than facet nerve blocks, although statistical significance was only reached at 1 month, not immediately post-treatment or after 3 months [19].

Two RCTs with high RoB compared intra-articular facet joint corticosteroid injections with other treatments; one compared facet joint injections with a mixture of local anaesthetics and corticosteroids combined with a home stretching exercise program with the home stretching exercise program only [15]. No significant post-treatment differences between the groups were found for pain and disability. The other trial compared the effects of facet joint corticosteroid injections with intra-articular sodium hyaluronate injections. No significant differences in pain relief, disability and quality of life between the groups were found at different follow-up points over a 6-month period [16].

One RCT with low RoB [18] compared the effects of multiple medial branch blocks of corticosteroids combined with local anaesthetics with multiple medial branch blocks consisting of only local anaesthetics. No significant differences between the groups were found at 3, 6 or 12 months post-treatment. One RCT with high RoB [17] compared the effects of multiple medial branch blocks of corticosteroids combined with local anaesthetics and Sarapin with multiple medial branch blocks consisting of local anaesthetics and Sarapin. No significant differences between the groups were found for pain relief, overall health, functional status and return-to-work over more than 2 years of follow-up.

There is low quality evidence (1 RCT; n = 86; inconsistency, imprecision) that intra-articular facet joint injections are slightly more effective than facet nerve blocks for pain relief in the intermediate term. There is very low quality evidence (1 RCT; n = 70; limitations in design, inconsistency, imprecision) that intra-articular facet joint injections add no benefit to a home exercise stretching program in terms of pain and disability over the short term. There is very low quality evidence (1 RCT; n = 60; limitations in design, inconsistency, imprecision) that there is no significant difference in effect between intra-articular facet joint corticosteroid injections and intra-articular injections of sodium hyaluronate on pain and disability over a long-term follow up. There is very low quality evidence (1 RCT; n = 120; inconsistency, indirectness, imprecision) that the addition of corticosteroids does not increase the effectiveness of facet joint nerve blocks with local anaesthetic. There is very low quality evidence (1 RCT; n = 73; limitations in design, inconsistency, indirectness, imprecision) that the addition of corticosteroids does not increase the effectiveness of facet joint nerve blocks with anaesthetic and Sarapin in chronic LBP.

Facet joint injections with local anaesthetic versus placebo

One RCT with low RoB [14] compared intra-articular facet joint injections with lidocaine to intra-articular facet joint injections with saline. In both groups these injections were followed by an injection of corticosteroid (cortivazol) near the joints. The lidocaine group had significantly higher pain relief post-treatment than the saline group [14].

There is low quality evidence (1 RCT; n = 80; inconsistency, imprecision) that intra-articular facet joint injections with lidocaine combined with peri-articular corticosteroid injections are more effective for short-term pain relief than placebo.

Epidural space

Epidural corticosteroid injections versus other treatments

In an RCT with high RoB [21], an epidural injection with a corticosteroid and dextrose solution was compared with an intrathecal benzodiazepine with dextrose injection. Two weeks and 2 months post-treatment, no significant differences between the groups were reported for pain relief or general improvement.

One RCT with low RoB [20] compared caudal epidural local anaesthetic and steroid injection with targeted epidural local anaesthetic and steroid placement with a spinal endoscope. No significant differences were found between the groups for any of the outcome measures at any of the times. In all patients in the endoscope group, post-treatment low-back discomfort was experienced but this was not persistent.

There is very low quality evidence (2 RCTs; n = 88; limitations in design, imprecision, inconsistency) that epidural corticosteroid injection is not significantly different to benzodiazepine injection or targeted epidural placement for pain relief over the short to intermediate term.

Epidural injections with local anaesthetics versus other treatments

One RCT with low RoB [22] compared the effects of epidural blocks with ropivacaine with epidural blocks with bupivacaine. Eight single shot epidural injections followed by active physiotherapy were performed in all patients. There were no significant differences found between the groups in post-treatment analgesia. There were three cases of short episodes of headache post-injection.

There is low quality evidence (1 RCT; n = 40; imprecision, inconsistency) that there is no significant difference in analgesia provided by blocks of ropivacaine or bupivacaine in the short term.

Lumbar musculature

Intramuscular injections with vitamin B12 versus placebo

In one RCT with high RoB [23], the effects of intramuscular vitamin B12 injections were compared with intramuscular placebo injections. Post-treatment, there were significant improvements for pain and disability in favour of the vitamin B12 group.

There is very low quality evidence (1 RCT; n = 60; limitations in design, inconsistency, imprecision) that intramuscular vitamin B12 injections are more effective than intramuscular placebo injections for short-term pain relief and improvement of function.

Intramuscular injections with botulinum toxin A versus placebo

In one RCT with low RoB, intramuscular injections of botulinum toxin A were compared with intramuscular placebo injections of saline [24]. At 3 weeks follow-up, the degree of pain relief was significantly different between groups in favour of the botulinum toxin A group. At 8 weeks patients in the botulinum toxin A group had significantly more pain relief and better ODI scores than the placebo group.

There is low quality evidence (1 RCT; n = 31; inconsistency, imprecision) that intramuscular botulinum toxin A injections are more effective for pain relief in the short- and intermediate-term than placebo.

Denervation procedures

Intervertebral disc

Percutaneous intradiscal radiofrequency thermocoagulation versus placebo

In one placebo-controlled trial (n = 28) with low RoB, no significant differences were found between PIRFT and sham PIRFT in pain VAS scores, global perceived effect, ODI, or a composite outcome of overall treatment success 8 weeks post-treatment [25]. In a second placebo-controlled trial (n = 20) with low RoB, only follow-up data which were collected after 6 and 12 months post-treatment were reported [28]. No significant differences were seen between the PIRFT and sham-PIRFT groups on pain intensity or functional status at either of these time points. Because of the variability in the timing of outcome measures between these two studies, a decision was made not to pool the results.

A third trial (n = 37) with high RoB found minimal improvement over 6 months on pain (VAS) and disability (ODI) with both lower- and higher- intensity of PIRFT. No significant differences were found between the groups at any of the follow-up assessments [26]. No complications or adverse events were reported in the placebo-controlled trials. One patient was excluded from the analysis of Ercelen et al. [26] because of discitis.

There is low quality evidence (1 RCT; n = 28; inconsistency; imprecision) that there is no difference between PIRFT and placebo in pain and functional status outcomes over an intermediate term of follow-up. There is also low quality evidence (1 RCT; n = 20; inconsistency; imprecision) that there is no difference between PIRFT and placebo in pain and functional status outcomes over a long-term follow-up.

Intradiscal electrothermal therapy versus placebo: pain

In patients with a positive response to provocative discography, two small (n = 55 and n = 56), low RoB, placebo-controlled randomized trials evaluated IDET and both provided sufficient data for pooling [27, 29]. Both studies measured pain with the SF-36 Bodily Pain Index (100-point scale). The Chi-square value for homogeneity of the WMD was 0.09 (P > 0.05), indicating statistical homogeneity among these studies.

There is low quality evidence (2 RCTs; n = 111; indirectness, imprecision) that IDET is more effective than placebo for pain relief over a long-term (6 months) follow-up (WMD −7.84; 95% CI −14.96 to −0.72) (Table 2, Analysis 01.01).

Intradiscal electrothermal therapy versus placebo: functional status

The same two RCTs [27, 29] provided ODI scores on a 100-point scale which allowed statistical pooling. The Chi-square value for homogeneity of the WMD was 1.37 (P > 0.05), indicating statistical homogeneity among these studies.

There is low quality evidence (2 RCTs; n = 111; indirectness, imprecision) that IDET is no more effective than placebo in improving functional status over a long-term (6 months) follow-up (WMD −4.93; 95% CI −10.11 to 0.25) (Table 2, Analysis 01.02).

In patients unresponsive to treatment with IDET, one high RoB RCT found radiofrequency denervation of the ramus communicans nerve was associated with better VAS pain, SF-36 bodily pain and SF-36 physical function scores after 4 months compared to sham denervation [30]. In one RCT, four patients who underwent IDET experienced transient radiculopathy (<6 weeks) [27]. No other serious adverse events were reported in the three trials.

Zygapophyseal joint (facet joint)

Radiofrequency denervation of facet joints versus placebo: pain

Five RCTs provided sufficient data on pain VAS scores to allow for pooling over a short-, intermediate- or long-term follow-up [31, 33–36]. All studies included only patients with a positive response (~50–80% pain relief) to local anaesthetic nerve block. One RCT was not included in the primary analyses due to clinical heterogeneity of patient selection procedures [32].

For short-term outcomes (<4 weeks), the Chi-square value for homogeneity of the WMD was 1.07 (P > 0.05), indicating statistical homogeneity between two of the RCTs. There is low quality evidence (2 RCTs; n = 90; indirectness, imprecision) that radiofrequency denervation of lumbar facet joints is more effective than placebo for pain relief over a short-term follow-up (WMD −18.15; 95% CI −24.21 to −12.09) (Table 2, Analysis 02.01).

For intermediate-term outcomes (1–6 months), the Chi-square value for homogeneity of the WMD was 1.75 (P > 0.05). There is low quality evidence (2 RCTs; n = 112; indirectness, imprecision) that radiofrequency denervation of lumbar facet joints is no more effective than placebo for pain relief in the intermediate term (WMD −9.29; 95% CI −22.57 to 4.00) (Table 2, Analysis 02.02).

For long-term outcomes (6 months), the Chi-square value for homogeneity of the WMD was 4.58 (P > 0.05). There is low quality evidence (3 RCTs; n = 130; indirectness, imprecision) that radiofrequency denervation of lumbar facet joints is no more effective than placebo for pain relief in the long term (WMD −6.99; 95% CI −14.73 to 0.76) (Table 2, Analysis 02.03).

When the study by Leclaire and colleagues [32] was included in the analyses, the pooled WMD (95% CI) for pain intensity was −14.80 (−22.77 to −6.82) in the short-term and −3.85 (−16.27 to 8.57) in the intermediate term. While the addition of this study to the meta-analyses slightly altered the pooled WMD, it did not change the conclusions.

One RCT with low RoB [37] compared radiofrequency denervation of the dorsal root ganglion with sham denervation. No significant differences were found between groups at 3-month follow-up. Adverse events and complications did not differ between treatments, and no serious complications or side effects arose in either group. There is very low quality evidence (1 RCT; n = 83; inconsistency, indirectness, imprecision) that radiofrequency denervation of the dorsal root ganglion is no more effective than placebo for pain relief in the intermediate term.

Radiofrequency denervation of facet joints versus placebo: functional status

One RCT comparing conventional and pulsed radiofrequency denervation to placebo provided sufficient data on functional status outcomes (ODI 0–100 scale) to allow for statistical pooling [34].

There is very low quality evidence (1 RCT; n = 60; inconsistency, indirectness, imprecision) that radiofrequency denervation of lumbar facet joints is more effective than placebo for improvement of function in the short term (WMD −5.53; 95% CI −8.66 to −2.40) (Table 2, Analysis 02.04).

When the study by Leclaire and colleagues [32] was included in the analyses, the WMD (95% CI) became −3.45 (−7.68 to 0.77) in the short term and −6.57 (−17.00 to 3.85) in the intermediate term (when pooled with the study by van Kleef et al. [35]). While the addition of this study to the meta-analyses slightly altered the short-term pooled WMD, it also changed the GRADE profile and the overall quality of the evidence from very low to low.

Radiofrequency denervation of facet joints versus other treatment

In a study with low RoB [34], conventional radiofrequency denervation of the lumbar facet joints was compared with pulsed radiofrequency denervation. Both treatments improved pain VAS and ODI scores compared with placebo, with conventional denervation improving significantly more than pulsed denervation by 6 months and 1 year post-treatment.

There is very low quality evidence (1 RCT; n = 40; inconsistency, indirectness, imprecision) that conventional radiofrequency denervation is more effective than pulsed radiofrequency denervation of the facet joints for pain relief or improvement of function over the long term.

Discussion

In this review, 27 RCTs were included that evaluated the effectiveness of injection therapy and denervation procedures for chronic LBP.

The effectiveness of the different therapies

In this review we found only low to very low quality evidence to support the use of injection therapy and denervation procedures in chronic LBP patients. Low quality evidence means that further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

From the 26 included RCTs, only two treatment subgroups (IDET vs. placebo, and radiofrequency denervation of facet joints vs. placebo) were considered to be clinically homogeneous enough to allow for statistical pooling. The results showed that IDET is more effective than placebo for pain relief over 6 months of follow-up and radiofrequency denervation of lumbar facet joints more effective than placebo for pain relief over a short-term (<4 weeks) follow-up. The quality of the evidence in both of these comparisons was downgraded to low quality due to imprecision and indirectness. Indirectness (lack of ability to generalize) refers to the extent to which the people, interventions and outcomes in the trials are not comparable to those defined in the inclusion criteria of the review. As all of the pooled RCTs selected patients for inclusion only if they had a positive response to either provocative discography (IDET vs. placebo) or local anesthetic nerve block (radiofrequency denervation of facet joints vs. placebo). While the selection of appropriate patients may be considered necessary for these therapies, the results cannot be generalized to all patients with chronic LBP. However, from the perspective of pain clinics or anesthesiologists these populations may be generalizable to their setting. In that case, the level of evidence is moderate, indicating that further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Consequently, further research is needed for a more confident estimate of effect.

Methodological considerations

The RoB assessment describes whether a RCT reported certain methodological features known to decrease bias when interpreting the results. However, despite the fact that the RoB of the included studies was generally low, many studies showed flaws regarding concealment of treatment allocation, care provider blinding, avoidance of co-interventions and performing intention-to-treat analyses. Overall, only low to very low quality evidence was found by this review. By using the GRADE approach, we were able to determine an overall judgment of the quality of the evidence not only from the limitations in design, but also considering aspects such as inconsistency among studies, imprecision and indirectness.

It is recommended that study features which are associated with a lower RoB in RCTs be considered when designing future trials of non-surgical interventional therapy, as well as adequate reporting of these features.

Adverse effects

In the majority of studies presented in this review, no adverse events or side effects associated with the treatments were reported. Transient symptoms such as increased low-back discomfort or paraesthesia were noted in a few studies. Epidural injections were associated with nausea and headache in some patients. In a study which used discectomy as a reference treatment [11], a dural defect with leakage of cerebrospinal fluid and a partial cauda equina were noted, though unsure whether these patients were in the injection therapy or surgery group. Most trials were small and not designed to evaluate adverse events, so no clear conclusion can be drawn regarding the risks of injection therapy and denervation procedures.

Strengths and limitations

Several biases can be introduced in systematic reviews by literature search and selection procedures. It is possible that in searching studies for this review, relevant but unpublished trials may have been missed, which are often likely to be small studies without positive results, leading to publication bias. However, because the majority of published trials was small and did not show a positive effect, publication bias does not seem to be a big problem in this review. Screening references of identified trials and systematic reviews may result in an over representation of positive studies in the review, because trials with a positive result are more likely to be referred to in other publications, leading to reference bias. Only studies published in English, Dutch, French or German were included in this review. It is not clear whether a language restriction is associated with bias [38].

Due to the study selection criterion used, including only trials of chronic (>3 months duration) LBP patients, this review included less studies than similar Cochrane reviews [3, 5]. Many excluded RCTs failed to explicitly state the duration of symptoms experienced by the patients, while others considered only the length of time patients were unresponsive to conservative treatment as part of their inclusion criteria. However, the main conclusions of our review appear to be similar to recent systematic reviews [3–6], which generally report a paucity of RCTs on injection therapy and denervation procedures for chronic LBP and insufficient evidence to support their use.

Implications for research

To conclude, we identified 27 RCTs that evaluated various types of injection therapy or denervation procedures at different locations for patients with chronic LBP. Most of the studies included in this review were small and had a low RoB, though there were methodological weaknesses, especially regarding concealment of allocation, co-interventions and use of intention-to-treat analyses. The quality of evidence was low or very low, indicating that further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. There is a need for future high-quality placebo-controlled RCTs with large sample sizes on injection therapy and denervation procedures.

Implications for practice

In patients with chronic LBP there is only low to very low quality evidence to support the use of injection therapy for pain relief and improvement of function. It cannot be ruled out that in carefully selected sub-groups of patients, such as those with a positive response to discography or local anaesthetic nerve block, certain interventional therapies may be of some benefit. Potential benefits must be weighed against possible adverse effects when deciding whether to provide injection therapy or denervation to chronic LBP patients.