Harms of selected spinal and paraspinal injections and denervation procedures for chronic non-cancer spine pain: a systematic review and meta-analysis of non-randomised studies

Abstract

Objective

To summarise the evidence on long-term and infrequent harms following selected spinal and paraspinal injections and denervation procedures for chronic non-cancer spine pain.

Design

Systematic review and meta-analysis.

Data sources

MEDLINE, EMBASE and Cumulative Index to Nursing and Allied Health Literature from inception to October 2023.

Study selection

Non-randomised studies reporting on harms of selected interventional procedures administered to adults living with chronic axial or radicular non-cancer spine pain with ≥4 weeks of follow-up.

Data extraction and synthesis

A parallel guideline panel provided input on the scope, design and interpretation of this systematic review, including selection of adverse events for consideration. Systematic literature screening, data abstraction and risk of bias appraisal were conducted independently and in duplicate by pairs of reviewers. We used random-effects models for all meta-analyses and the Grading of Recommendations Assessment, Development and Evaluation approach to evaluate the certainty of evidence.

Results

We included 60 longitudinal studies (56 non-comparative, 4 comparative) that enrolled 4966 patients with chronic non-cancer spine-related pain. 31 studies investigated radiofrequency ablation or denervation, 22 epidural injections and 11 joint injections or nerve blocks. Low certainty evidence suggests that joint targeted steroid injection and epidural steroid injection for chronic spine pain may result in temporary altered level of consciousness (incidence: 2.1%; 95% CI 1.1% to 4.0%), joint radiofrequency nerve ablation, joint targeted steroid injection and epidural injection of local anaesthetic and steroids may result in deep infection (incidence: 0.7%; 95% CI 0.3% to 2.0%), epidural steroid injection, joint radiofrequency nerve ablation and joint targeted injection of local anaesthetic and steroids may result in dural puncture (incidence: 1.4%; 95% CI 0.5% to 4.3%), and dorsal root ganglion radiofrequency and joint radiofrequency nerve ablation with or without joint-targeted injection of steroids may result in prolonged pain or stiffness (incidence: 8.6%; 95% CI 6.3% to 11.6%). Several interventional procedures may result in metabolic complications and prolonged sensory deficits, but the supporting evidence was only very low certainty. Most complications resolved spontaneously or with conservative management.

Conclusions

Low certainty evidence suggests that several common interventional procedures for chronic spine pain show risk of deep infection, dural puncture, temporary altered level of consciousness and prolonged pain or stiffness. Other harms are uncertain due to very low certainty evidence, and catastrophic outcomes were not reported in the small studies that contributed to our analyses.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• We conducted a comprehensive search for eligible studies informing harms associated with selected spinal and paraspinal injections and denervation procedures for chronic, non-cancer spine pain.

• We used the Grading of Recommendations Assessment, Development and Evaluations approach to appraise the certainty of evidence.

• The evidence for all adverse events was only low or very low certainty, suggesting that further research is highly likely to have an important impact on confidence in effect estimates.

• Both the retrospective design of most studies and the small numbers of patients enrolled in studies eligible for review, with none reporting very rare but catastrophic harms that can occur after interventional procedures (eg, paralysis following spine injection), likely result in underestimation of the risk of harms.

Introduction

Chronic, non-cancer spine pain is a leading cause of morbidity, years lived with disability and productivity loss globally.1,3 In 2020, low back pain affected 619 million people worldwide and the number of individuals affected by 2025 is estimated to exceed 840 million.⁴ Globally, in 2020, neck pain affected 203 million people with an estimated 20% developing chronic concerns.⁵ The economic burden associated with low back pain and its management exceeds US$100 billion per year.16,8

Clinicians frequently offer patients living with chronic spine pain interventional procedures, particularly in North America, such as joint or epidural injections with corticosteroids or anaesthetics, medial branch blocks or radiofrequency ablation.^{9 10} These procedures may be associated with rare but serious harms, such as deep infection or paralysis.11,14 Several systematic reviews of randomised controlled trials have explored the safety and effectiveness of interventional procedures for chronic non-cancer spinal pain15,17; however, randomised trials in this area typically enrol small numbers of patients followed for short time frames, limiting their ability to inform infrequent harms.

We conducted a systematic review and meta-analysis of non-randomised studies to summarise the evidence on adverse events of interventional procedures for chronic non-cancer spine pain. Our findings, as well as a second systematic review,¹⁸ informed a parallel BMJ Rapid Recommendation addressing common interventional procedures for chronic spine pain.¹⁹ This evidence synthesis is part of the BMJ Rapid Recommendations project, a collaborative effort from the MAGIC Evidence Ecosystem Foundation (www.magicevidence.org) and the BMJ.²⁰

Methods

We reported our systematic review and meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Harms Checklist,²¹ and registered our protocol on the Open Science Framework (https://osf.io/uebfs).

Guideline panel involvement

Our parallel guideline panel was comprised of 22 members, including 10 clinical experts (4 physiatrists, 1 rheumatologist, 2 anaesthesiologists focused on pain medicine, 1 clinical pharmacologist, 1 physiotherapist, 1 general internist), four patient partners and eight methodologists (four of whom were also front-line clinicians). The panel decided the scope of the guideline (eg, which interventional procedures would be considered) and led the selection and categorisation (type and severity (moderate or serious)) of adverse events for consideration. To ensure feasibility, the panel excluded purported regenerative therapies such as prolotherapy, platelet-rich plasma and stem cell injections.

We extracted all adverse events from eligible studies, which were then categorised by six clinical experts on the parallel BMJ Rapid Recommendations guideline panel who had experience administering interventional procedures for chronic spine pain, blinded to results. We used an online survey to have panel members independently rank each category of harm according to patient-importance on a 12-point scale, from 12 (most important) to 1 (least important). We prioritised presentation of harms identified by the guideline panel as most important to patients (weighted score of ≥6/12) and reported by at least five studies.

Data sources

We conducted systematic searches for eligible studies in Medline, EMBASE and Cumulative Index to Nursing and Allied Health Literature from inception to October 2023, without language restrictions. An experienced medical librarian (RC) developed all database-specific search strategies (online supplemental appendix 1). We reviewed the reference lists of included studies to ensure we had not omitted any relevant evidence.

Study selection

Using standardised pilot-tested forms, three pairs of trained reviewers, all of whom were medical residents or physicians in training (FM, SA, FK, CL, PH and TL), screened titles and abstracts of identified citations independently and in duplicate. Following title and abstract screening, reviewers screened full texts of potentially eligible studies. Disagreements were resolved by reviewers through discussion or adjudication with a third reviewer. For methodological disagreements, the third reviewer was a senior methodologist (JWB), and for clinical disagreements the third reviewer was one of the clinical experts on our guideline panel. Title and abstract screening as well as full-text review was completed using DistillerSR (Evidence Partners, Ottawa, Canada; http://systematic-review.net).

We included all non-randomised studies in which: (1) at least 80% of participants were adult patients (>18 years old), presenting with chronic (>12 weeks or explicitly defined by study authors as ‘chronic’) axial and/or radicular, non-cancer spine pain, (2) patients received an interventional procedure considered by our guideline panel (ie, injection of local anaesthetic, steroids or their combination into the cervical or lumbar facet joint, or sacroiliac joint; epidural injections of local anaesthetic, steroids or their combination; radiofrequency of dorsal root ganglion; radiofrequency denervation of cervical or lumbar facet joints, or the sacroiliac joint; and paravertebral intramuscular injections of local anaesthetic, steroids or their combination) and (3) reported harms as defined by the authors (online supplemental appendix 2).

We excluded studies: (1) with fewer than 25 patients, which would not appreciably contribute to pooled estimates and would be too small to reliably estimate the incidence of adverse events, (2) with less than 1-month follow-up, (3) that only reported surrogates of patient-relevant harms or (4) systematic reviews or other study types that did not report primary data.

Data extraction

Using standardised pilot-tested data collection forms, pairs of reviewers working independently and in duplicate extracted the following information from eligible studies: (1) study design, (2) study and patient characteristics (country, sample size, mean age, per cent female, condition or diagnosis, funding source), (3) intervention details (description of intervention, specialisation of providers, use of image guidance, whether a positive diagnostic block was required before treatment, follow-up length) and (4) outcome data for all patient-important harms (description of adverse event reported, number of participants experiencing the outcome). For studies that reported on adverse events at multiple time points, we extracted data for the longest point of follow-up. We attempted to acquire missing data by contacting study authors.

Risk of bias

Available instruments to assess risk of bias of non-randomised studies, including the Risk Of Bias In Non-randomized Studies of Interventions, have shown poor inter-rater and inter-consensus reliability and high evaluator burden.²² We therefore assessed risk of bias using modified criteria from the Users’ Guides to the Medical Literature,²³ which have been used in several prior reviews of observational studies and demonstrated high agreement between reviewers.24,26 The following criteria were used to assess the risk of bias in observational cohort studies with two or more arms: (1) selection of both cohorts from the same population, (2) representativeness of the target population, (3) accuracy of assessment of adverse events and (4) infrequent (<20%) missing outcome data. We used the following criteria to assess the risk of bias for single-arm observational studies: (1) representativeness of target population, (2) accuracy of assessment of adverse events and (3) infrequent (<20%) missing outcome data. Response options for each item were ‘definitely yes’ (assigned a low risk of bias), ‘probably yes’ or ‘probably no’ (assigned intermediate risk of bias) and ‘definitely no’ (assigned a high risk of bias). We rated studies as low risk of bias overall when all domains were at low risk of bias, or if most domains were low risk of bias and one domain was rated at intermediate risk. We rated studies as intermediate risk of bias if all or most domains were at intermediate risk, and high if one or more domains were rated as high risk of bias.

Data synthesis

We conducted analyses for comparative and non-comparative studies separately. We prioritised presentation of data on all adverse events reported by at least five studies and ranked by our guideline panel as among the most patient-important, defined as an average score of ≥6 out of 12. To address this question, clinicians on the panel considered their experiences conducting shared decision-making with patients. Patients on the panel considered their own experience and those of other patients attending interventional procedure clinics with whom they had interacted. Our unit of analysis was the number of patients experiencing an adverse event and not the number of events. Details on all adverse events captured in our review are reported in online supplemental appendix 3.

We reported adverse events as binary outcomes. Where possible, we reported risk differences with associated 95% CIs for comparative studies. We conducted meta-analyses for all adverse events reported by at least two studies using one-stage random-effects models, which consider both within- and between-study variability.²⁷ We performed sensitivity analyses using a Hartung-Knapp random-effects model.²⁸ When studies reported multiple events within the same category of adverse event but did not confirm independence, we did not add events but instead used the most frequently endorsed event to avoid clustering (eg, Al-Samman et al²⁹ reported that of 52 patients undergoing epidural steroid injection, 7 experienced accidental intradural puncture resulting in headache, 1 developed vomiting and 1 experienced drowsiness; we used 7/52 for the event rate of dural puncture).

In consultation with the parallel BMJ Rapid Recommendations guideline panel, we prespecified three subgroup hypotheses to explain heterogeneity between studies: (1) risk of bias (high and intermediate vs low risk of bias), (2) type of interventional procedure and (3) use of image guidance versus not. We conducted subgroup analyses only if there were two or more studies in each subgroup. We performed tests for interaction to establish whether subgroups differed significantly from one another and assessed the credibility of significant subgroup effects using the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN) criteria.^{30 31} All analyses were conducted using the metafor package in R (V.3.5.1).³²

Certainty of evidence

We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to evaluate certainty of evidence.^{33 34} With this approach, observational studies reporting on treatment effects begin at low certainty evidence but can be further rated down based on indirectness, imprecision, inconsistency, risk of bias and publication bias. The certainty of evidence can also be rated up one or two levels on the basis of a strong association, a dose-response gradient or when all plausible confounders or other biases increase our confidence in the estimated effect.³⁵ To assess the strength of association between the outcome and intervention, we provided each type of harm reported among studies eligible for review to the 10 clinical experts on our guideline panel without any other data (see Acknowledgements). These clinicians determined that the strength of association was strong for deep infection and dural puncture occurring soon after an interventional procedure. In other words, they were confident that these events would be highly unlikely to have occurred (eg, spontaneous dural puncture) unless an interventional procedure was administered. They determined that the strength of association was moderate for temporary altered level of consciousness, prolonged (>48 hours) sensory deficits and prolonged (>48 hours) pain or stiffness. Specifically, these events could have occurred spontaneously after a procedure but it would be unlikely without an interventional procedure being administered.

We assessed heterogeneity across studies contributing to our pooled estimates using both statistical tests and visual inspection of forest plots (ie, differences in point estimates and overlap of CIs). We estimated between-study variance with tau-squared (τ²) which provides an absolute measure of heterogeneity. We considered τ²<0.05 as low, between 0.05 and 0.1 as moderate, and >0.1 as substantial between-study heterogeneity.³⁶ We rated down for heterogeneity when τ² was greater than 0.1, and did not when values were less than 0.05; when τ² scores were between 0.05 and 0.1 we reviewed forest plots to determine if studies contributing to pooled estimates showed important heterogeneity and, if so, rated down. We did not rate down the same effect estimate twice for both inconsistency and imprecision, when inconsistency was the cause of imprecision.

For assessing imprecision of adverse events, based on feedback from the guideline panel, we deemed a 5% difference (for comparative studies) or incidence (for single-arm studies) of moderate adverse evidence to be patient-important, and a 1% difference or incidence for serious adverse events, such as deep infection and dural puncture. We followed GRADE guidance for communicating our findings.³⁷ Our intent was to explore harms caused by interventional procedures; therefore, although our review was restricted to observational studies we used causal language to describe effects.³⁸ Guideline panel members interpreted the magnitude of adverse events and decided whether the observed incidence of harms was sufficient to affect patients’ decisions to pursue interventional procedures for chronic spine pain.

Patient and public partner involvement

Four patient partners (one woman and three men) were included as part of the guideline panel and contributed to the selection and prioritisation of outcomes, protocol development, interpretation of review findings and provided insight on patients’ values and preferences. Each of our patient partners was living with chronic spine pain and were selected to represent a range of experiences regarding interventional procedures. Three had received various interventional procedures for their chronic pain, with two finding relief and one that did not. The fourth lived with chronic spine pain but had no personal experience with interventional procedures.

Results

Study selection

We identified 13 123 citations, of which 582 were screened in full text and 93 met initial eligibility criteria. Thirty-three studies did not report usable information for adverse events (eg, reporting adverse events as a category without details about what specific events were considered, reporting that no adverse events occurred without any other details), leaving 60 studies eligible for review (figure 1).2939,97 We contacted authors of five studies to clarify if ≥80% of their participants had chronic pain, of which two responded in the affirmative.^{43 54}

Figure 1. Study selection process.

Description of studies

We included 60 longitudinal cohort studies (56 non-comparative, 4 comparative) in our analysis, that enrolled 4966 adult patients living with chronic axial pain, with a median of 62 participants (IQR 44–107) (online supplemental appendix 4). Of the 56 single-arm studies, 23 were prospective and 33 were retrospective. Of the four comparative studies, one was prospective,⁸¹ three were retrospective^{40 44 62} and none attempted to match cohorts for baseline factors. Follow-up ranged between 4 and 139 weeks. 26 studies (43%) were conducted in Europe, 21 studies (35%) in Asia, 7 studies (12%) in North America, 4 studies (7%) in South America and 1 each in Egypt and Australia. Most studies (38 of 60; 63%) did not report funding information, but of the 22 that did only 2 were industry funded (online supplemental appendix 5).

The median of the mean age reported among studies was 54 years (IQR 49–60). Among studies reporting sex distribution, 58% (2691/4646) of patients were female. The most commonly reported conditions were disc herniation (33%, n=20), spinal stenosis (15%, n=9) and degenerative lumbar disc disease (10%, n=6). The most frequently administered procedures were joint radiofrequency ablation/denervation (52%, n=31), epidural injections (37%, n=22) and joint injections or nerve blocks (18%, n=11); four studies explored more than one procedure. 19 studies (32%) reported that a positive diagnostic block was a requirement for patients before an interventional procedure was administered. 55 studies (92%) used some form of image guidance during procedures, while 5 studies (8%) did not. Of the four comparative studies, three^{40 44 81} compared joint injections or nerve blocks with epidural injections, and one⁶² compared epidural injection combined with radiofrequency ablation with epidural injection alone. The most common clinical specialisation of physicians administering interventional procedures was anaesthesiology and physical medicine and rehabilitation (online supplemental appendix 6).

Risk of bias

Of the 60 included studies, 43 (72%) were rated at intermediate risk of bias, 9 studies (15%) at high risk of bias and 8 studies (13%) at low risk of bias. Only 1 of 60 included studies was at low risk of bias for all criteria. Primary sources of bias across studies included limited representativeness of the target population and the lack of robust, well-delineated methods for adverse event assessment (online supplemental appendix 7). Adverse event reporting in most studies consisted of unprompted self-reporting by patients, often over long time frames, thus increasing the risk of recall bias.

Adverse event reporting

Clinical experts on our guideline panel categorised all reported adverse events into six domains (pain, stiffness and soreness, infection or inflammation, neurological deficits or injury, bleeding, other puncture-related complications and other systemic complications), and 21 subdomains (online supplemental appendix 3). The most patient-important harms reported by at least five studies were: (1) deep infection (eg, bone infection, joint infection, meningitis), (2) puncture of the membrane surrounding the spinal cord (the dura) and directly related symptoms, (3) prolonged (>48 hours) sensory deficits (eg, numbness or loss of sensation), (4) metabolic complications (eg, weight gain), (5) temporary altered level of consciousness (eg, syncope, presyncope, dizziness, drowsiness) and (6) prolonged (>48 hours) procedure-related localised pain or stiffness at the injection site (online supplemental appendix 8). Most adverse events resolved spontaneously without intervention. One patient who developed discitis required surgical fusion. Another patient developed spondylodiscitis that was resolved by antibiotic therapy. In one study, two patients developed neuritis following a radiofrequency denervation procedure, which resolved with oral prednisone.

Deep infection adverse events

Five longitudinal, non-comparative studies including 539 patients reported on the incidence of deep infection following interventional procedures.^{45 48 61 67 83} Three studies examined joint radiofrequency ablation or denervation,^{45 48 83} one study examined epidural injections with both corticosteroid and anaesthetic⁶⁷ and one study examined joint injection with corticosteroid.⁶¹ All five studies used image guidance. Specific outcomes considered under deep infection included discitis, spondylodiscitis, septic sacroiliitis, articular abscess, osteomyelitis and meningitis. We found low certainty evidence that several interventional procedures for chronic spine pain may result in deep infection (0.7%, 95% CI 0.3% to 2.0%; τ²=0) (figure 2, table 1).

Figure 2. Incidence of deep infection among people living with chronic spine pain following receipt of an interventional procedure. HK, Hartung-Knapp.

Table 1. GRADE evidence profile for adverse events associated with interventional procedures for chronic spinal pain in single-arm observational studies.

Adverse event	Types of interventional procedures administered (no. of studies)	No. of studies (range of follow-up)	No. of patients	Incidence (95% CI)	Reasons for downgrading	Certainty of evidence
Deep infection	Joint radiofrequency ablation (n=3). Epidural injection of corticosteroid and local anaesthetic (n=1). Joint corticosteroid injection (n=1).	5 (12–90 weeks)	539	0.7% (0.3% to 2.0%)	Risk of bias* Imprecision†	Low‡
Dural puncture	Epidural corticosteroid injection (n=5). Joint radiofrequency ablation (n=2). Nerve blocks (n=1).	8 (12–52 weeks)	1097	1.4% (0.5% to 4.3%)	Risk of bias* Imprecision†	Low‡
Temporary altered level of consciousness	Epidural steroid injection (n=9). Joint corticosteroid injection (n=1).	10 (12–52 weeks)	1021	2.1% (1.1% to 4.0%)	Risk of bias*	Low§
Prolonged (>48 hours) sensory deficits	Dorsal root ganglion radiofrequency (n=2). Joint radiofrequency ablation (n=2). Joint radiofrequency ablation with corticosteroid injection (n=1).	5 (12–104 weeks)	363	6.8% (2.3% to 18.6%)	Risk of bias* Imprecision†	Very low§
Prolonged (>48) pain or stiffness	Dorsal root ganglion radiofrequency (n=3). Joint radiofrequency ablation (n=12). Joint radiofrequency ablation with corticosteroid injection (n=1).	16 (8–105 weeks)	1132	8.6% (6.3% to 11.6%)	Risk of bias*	Low§

^* Uncertain representativeness of the study population and/or accuracy of the assessment of adverse events.

^† The associated estimate of precision includes the patient-important threshold (ie, 1% for serious fatal adverse events, such as deep infection and dural puncture; 5% for moderate adverse events).

^‡ Certainty of evidence began as high based on a strong association between the intervention and observed harm and plausible confounders or other biases increasing confidence in the estimated effect.

^§ Certainty of evidence began as moderate based on a moderate association between the intervention and observed harm.

GRADE, Grading of Recommendations Assessment, Development and Evaluation.

Dural puncture and directly related adverse events

Eight longitudinal, non-comparative studies including 1097 patients reported on the incidence of dural puncture and directly related symptoms including postpuncture headache, intracranial hypotension secondary to dural puncture and cerebral spinal fluid aspiration following interventional procedures.^{29 51 53 58 60 71 72 91} Five studies examined epidural corticosteroid injections,^{29 51 53 60 71} two studies considered radiofrequency ablation^{72 91} and one study examined nerve blocks.⁵⁸ All studies used image-guidance for procedures except for one examining epidural steroid injections.⁷¹ We found low certainty evidence that several interventional procedures for chronic spine pain may result in dural puncture or related adverse events (1.4%, 95% CI 0.5% to 4.3%; τ²=1.5) (figure 3, table 1). In total, 23 of 1097 patients experienced dural puncture or related symptoms.

Figure 3. Incidence of dural puncture among people living with chronic spine pain following receipt of an interventional procedure. HK, Hartung-Knapp.

Two comparative studies including 234 patients examined joint steroid injection or nerve block with epidural steroids, and reported on dural puncture and directly related symptoms.^{40 44} In both studies, neither patients receiving joint injections/nerve blocks or epidural injections reported any dural punctures or related adverse events (online supplemental appendix 9). Image guidance was used for both intervention arms.

One comparative study examined image-guided combined epidural injection and radiofrequency ablation versus image-guided epidural steroid injection alone.⁶² No dural puncture or related adverse events were reported in the epidural injection arm, while two patients developed dural puncture or related outcomes in the combined treatment arm (risk difference [RD] 1.7%, 95% CI −1.9% to 5.9%; low certainty evidence) (online supplemental appendix 9).

Prolonged (>48 hours) sensory deficits

Five longitudinal, non-comparative studies including 363 patients reported on the incidence of prolonged (>48 hours) sensory deficits following interventional procedures for chronic spine pain.^{48 50 57 78 90} Two studies^{48 50} examined joint radiofrequency ablation, two studies^{57 78} examined dorsal root ganglion radiofrequency ablation and one study examined combined radiofrequency neurotomy and corticosteroid injection.⁹⁰ All studies used image guidance. We are uncertain whether interventional procedures for chronic spine pain cause prolonged sensory deficits due to very low certainty evidence (6.8%, 95% CI 2.3% to 18.6%; τ²=1.4) (online supplemental appendix 10, Table 1). Six of these patients developed numbness of the back lasting up to 3 weeks,⁹⁰ 10 patients developed transient sensory changes in the ipsilateral leg which resolved in 1 week,⁵⁷ 2 patients developed numbness in both legs lasting approximately 3 months,⁴⁸ 1 patient developed transient thigh numbness that resolved in 1 week⁷⁸ and 6 patients developed paraesthesia in their lower back and buttocks that resolved in 2–3 weeks.⁵⁰

Metabolic complications

Four longitudinal, non-comparative studies including 473 patients reported on the incidence of metabolic complications following interventional procedures.^{29 53 64 70 71} Three studies^{29 53 64 71} examined epidural steroid injections while one study⁷⁰ examined joint radiofrequency ablation. Three studies^{29 53 64} used image guidance. Specific metabolic complications considered included hyperglycaemia, weight gain and blood pressure changes, which were not possible to pool due to insufficient number of studies reporting on the same outcome.

Among the three studies that administered epidural steroids, Adigüzel et al (2017) reported that 1 of 62 patients (1.6%) developed elevated blood sugar,⁶⁰ Forrest reported that 23 of 28 patients (82%) experienced an average weight gain of 2.7 kg and 16 patients (57%) showed an average increase in resting blood pressure of 10/5 mm Hg⁷¹ and Al-Samman et al (2020) reported elevated blood pressure in 2 of 52 (3.8%) patients.^{29 39} Among the one study evaluating radiofrequency denervation of the sacroiliac joint among 77 patients, 1 person with diabetes developed hyperglycaemia requiring increased insulin use for 3 days.⁷⁰ Evidence supporting weight gain, elevated blood sugar and increased blood pressure after interventional procedures was only very low certainty.

Temporary altered level of consciousness adverse events

10 non-comparative, longitudinal studies examining 1021 patients reported on the incidence of temporary altered level of consciousness events following interventional procedures for chronic spine pain.^{29 41 47 51 52 58 64 73 80 87} Nine studies examined epidural steroid injections, with one study⁵⁸ examining joint steroid injections. All studies except one⁴¹ used image guidance. Specific altered level of consciousness events captured included vasovagal episodes, dizziness and drowsiness. We found low certainty evidence that epidural steroid or joint steroid injections for chronic spine pain may result in temporary altered level of consciousness (2.1%, 95% CI 1.1% to 4.0%, τ²=0.5) (online supplemental appendix 11, Table 1). All events were transient and resolved spontaneously.

Two comparative studies^{40 44} examining joint steroid injection or nerve block with epidural steroid injections reported on temporary altered level of consciousness. A total of 10 patients receiving epidural steroid injections across both studies developed a vasovagal episode immediately after the procedure. In the joint injection or nerve block arm, nine patients developed a vasovagal episode immediately after the procedure (RD −2.7%, 95% CI −9.8% to 4.4%; low certainty evidence) (online supplemental appendix 12).

Prolonged (>48 hours) procedure-related localised pain or stiffness adverse events

16 non-comparative, longitudinal studies examining 1132 patients reported on the incidence of prolonged localised pain or stiffness following interventional procedures for chronic spine pain.4546 48,50 54 57 63 74 75 83 85 86 88 90 91 All 16 studies examined radiofrequency procedures. Image guidance was used in all studies except one.⁸⁵ Three studies examined dorsal root ganglion radiofrequency,^{57 74 86} and one study examined joint radiofrequency ablation combined with nerve block.⁹⁰ The remaining 12 studies examined joint radiofrequency ablation. We found low certainty evidence that several interventional procedures for chronic spine pain may cause prolonged (>48 hours) procedure-related localised pain or stiffness (incidence=8.6%, 95% CI 6.3% to 11.6%, τ²=0.2) (online supplemental appendix 13, Table 1).

The severity and duration of pain symptoms reported was variable. In one study, 27 patients developed pain and muscle stiffness lasting up to 1 month, with 6 patients experiencing pain aggravation resolving after 1 month.⁴⁵ Several patients developed mild-to-moderate injection or puncture site pain that lasted between 3 days and 5 weeks.^{46 48 50 57 75 85 90 91} Seven patients developed neuritis.^{49 63 88} One study reported two patients developed post-procedural pain of up to 5 days requiring additional CT-guided blocks with corticosteroids and local anaesthesia.⁸³ The longest duration of post-intervention symptoms was a mean of 91 days in two patients who suffered from mild leg pain with pedicle tract issues and were treated with oral pain medication.⁸⁶ The most severe procedure-related localised pain event was captured in one study in which two patients developed neuritis during the first week following a radiofrequency denervation procedure.⁴⁹ These patients reported ‘severe’ burning pain radiating down their leg and were managed with oral prednisone, 25 mg/day for 5 days, at which time their symptoms resolved.

Other adverse events

The incidence of other adverse events following interventional procedures for chronic spine pain are summarised in online supplemental appendices 14-29.

Subgroup analyses

There was insufficient variability among studies reporting on the incidence of deep infection to facilitate subgroup analysis. For dural puncture and directly related outcomes, we were able to conduct subgroup analysis by intervention type (epidural steroid injection compared with radiofrequency ablation, online supplemental appendix 30) and overall risk of bias (high compared with low risk of bias, online supplemental appendix 31). For prolonged sensory deficits, we conducted subgroup analysis by intervention type (dorsal root ganglion radiofrequency ablation compared with joint radiofrequency ablation) (online supplemental appendix 32). For temporary altered level of consciousness, we conducted subgroup analysis by risk of bias (intermediate compared with high risk of bias, online supplemental appendix 33). For prolonged localised pain, we were able to conduct subgroup analysis by high compared with low risk of bias, and by intervention type (joint radiofrequency ablation compared with dorsal root ganglion radiofrequency) (online supplemental appendix 34 and 35).

No significant subgroup effects were observed, except for prolonged localised pain in which dorsal root ganglion radiofrequency showed a larger incidence of events than joint radiofrequency ablation (14.8%, 95% CI 8.8% to 23.8% vs 7.8%, 95% CI 5.5% to 10.9%; p=0.04); however, this subgroup effect was only low credibility according to ICEMAN criteria (online supplemental appendix 36).

Discussion

Main findings

Our systematic review and meta-analysis found low certainty evidence that suggests interventional procedures for chronic spine pain may result in deep infection (following joint radiofrequency nerve ablation, joint targeted steroid injection and epidural injection of local anaesthetic and steroids), dural puncture (following epidural steroid injection, joint radiofrequency nerve ablation and joint targeted injection of local anaesthetic and steroids), temporary altered level of consciousness (following joint targeted steroid injection and epidural steroid injection) and prolonged pain or stiffness (following dorsal root ganglion radiofrequency and joint radiofrequency nerve ablation with or without a corticosteroid injection). We are uncertain of the incidence of prolonged sensory deficits and metabolic complications resulting from interventional procedures for chronic spine pain as supporting evidence was only very low certainty. Most complications resolved spontaneously or with conservative management.

Relevant literature

The most recent narrative reviews examining interventional procedures for chronic spinal pain were published in 2023.98,100 One review examining epidural steroid injections and facet joint injections reported an overall incidence of adverse events ranging from 0.04% to 13%, with most complications being reported as transient or mild; the incidence of severe adverse events was less than 1%.⁹⁸ The most common adverse event following epidural steroid injection was vasovagal reaction, ranging from 0.04% to 8%. Local bleeding was the most common adverse associated with facet joint blocks with an estimated incidence of 12%; injection-site pain was the second most common with an incidence of 6%. Spinal infection was reported to be a rare complication, with no studies reporting an incidence of greater than 0.07%.⁹⁸

Another review examining radiofrequency ablation concluded that serious complications were rare. This review found the incidence of transient paraesthesia ranged from 1% to 22%,⁹⁹ and that overall, radiofrequency ablation for chronic spinal pain was associated with a minor complication rate of 2.4%.⁹⁹

Our study is the first systematic review and meta-analysis to examine harms of several common interventional procedures for chronic spinal pain. Strengths include a comprehensive search for non-randomised studies, explicit eligibility criteria, screening of studies and collection of data in duplicate to increase reliability, and use of the GRADE approach to evaluate the certainty of evidence.

There were several limitations to our review. We surveyed our guideline panel using established methods^{101 102} to inform the patient-importance of harms identified in studies eligible for review, as a large survey of patients on this issue was not feasible. The non-comparative design of most studies for several outcomes precludes confident inferences regarding the proportion of adverse events that can be attributed to interventional procedures; however, confounding was accounted for in our certainty of evidence ratings.

Studies eligible for our review included several types of chronic non-cancer neck, back and sacroiliac pain. About one third of studies required a positive diagnostic block before treatment but this diagnostic method has unproven reliability or validity in establishing an anatomic diagnosis. The large majority of included patients presented with degenerative changes that are common among asymptomatic adults.¹⁰³ The presumption of a spinal cause of pain was not confirmed or strongly supported in the majority of subjects. An estimated 85% of chronic spine pain cannot be attributed to a specific cause,¹⁰⁴ and diagnostic blocks carry a false positive rate of up to 60%.^{105 106} In addition, there may be patient features that are associated with harms (eg, spinal pathology severity); however, exploring this would require source articles to present harms according to different risk stratification factors, or access to individual patient-level data—neither of which was available. Further, apart from dural puncture, the risk of harms reported among studies captured in our review would be expected to be similar across the selected spinal and paraspinal injections and denervation procedures that were considered.

Studies eligible for our review enrolled small numbers of patients (median=62), and none reported very rare but catastrophic outcomes that have been reported in case reports and large administrative databases, such as epidural abscess, meningitis, paraplegia and death.1113 107,111 Most studies eligible for our review were retrospective in nature, and adverse events were likely not comprehensively or systematically appraised among patients receiving common interventional procedures for chronic spine pain. A comparison of 120 surgical patient records in both the University HealthSystem Consortium (UHC) and American College of Surgeons National Surgical Quality Improvement Program (NSQIP) found that 28% of patients were reported as suffering complications in NSQIP but only 11% in UHC (eg, 13% of patients were reported as having a surgical site infection in NSQIP, but only 1% in UHC).¹¹² As such, our findings may underestimate the risks of harms.^{113 114} Alternately, most studies were also non-comparative, which may result in overestimation of subjective harms as the lack of a control group precludes ruling out contribution by confounding variables or events. Further, inadequate reporting of harms (eg, reporting on ‘adverse events’ without any other details) resulted in the exclusion of 33 of 93 studies from our review. We encourage study authors to clearly report details on specific adverse events to optimise interpretability of their findings and facilitate pooling across studies.

Implications

Our systematic review and meta-analysis found that evidence regarding long-term and serious harms of interventional procedures for chronic spine pain is only low or very low certainty. This lack of information complicates informed decision-making by patients considering this therapeutic option; however, all low certainty evidence suggests that several interventional procedures may cause serious harms, including deep infections and dural puncture. Contextualised guidance on this body of evidence is provided in the accompanying BMJ Rapid Recommendation.¹⁹

Future studies of harms following interventional procedures for chronic spine pain should collect information on adverse events in a systematic manner, including prompting patients for all plausible events. Risk of catastrophic outcomes was not informed by the small studies that were eligible for analyses, likely because of their rarity; however, while rare (eg, permanent paraplegia due to infection) serious harms are an important consideration when the evidence suggests little to no benefit on pain relief for the interventional procedures considered in our review compared with sham procedures.¹⁸

Conclusions

In this systematic review of observational studies, we found low certainty evidence that suggests interventional procedures for chronic spine pain may increase the incidence of several harms. Specifically, (1) deep infection following joint radiofrequency nerve ablation, joint targeted steroid injection and epidural injection of local anaesthetic and steroids; (2) dural puncture following epidural steroid injection, joint radiofrequency nerve ablation and joint targeted injection of local anaesthetic and steroids; (3) temporary altered level of consciousness following joint targeted steroid injection and epidural steroid injection; and (4) prolonged pain or stiffness following dorsal root ganglion radiofrequency and joint radiofrequency nerve ablation with or without a corticosteroid injection. The risk of prolonged sensory deficits is uncertain due to very low certainty evidence, and catastrophic outcomes were not reported in the small studies that contributed to our analyses.

Data availability statement

Data are available upon reasonable request.