OUP user menu

Radiofrequency Neurotomy for Low Back Pain: Evidence-Based Procedural Guidelines

W. Michael Hooten MD, David P. Martin MD, PhD, Marc A. Huntoon MD
DOI: http://dx.doi.org/10.1111/j.1526-4637.2005.05022.x 129-138 First published online: 1 March 2005

ABSTRACT

Objective. This review was undertaken to outline the procedural limitations of the randomized controlled trials (RCTs) of radiofrequency (RF) neurotomy for low back pain. Second, the literature related to patient selection, diagnostic testing, and the technique of performing lumbar spine RF neurotomy will be critically reviewed and analyzed. Based on these analyses, diagnostic and procedural guidelines will be proposed.

Design. A Medline and EMBASE search identified three RCTs and two systematic reviews of RF neurotomy for low back pain. A similar search identified pertinent literature related to the method of patient selection for a diagnostic block, the medial branch and L5 dorsal ramus comparative block, and the anatomical and technical parameters of lumbar spine RF neurotomy.

Results. Substantial procedural shortcomings were identified in all three RCTs. In the systematic reviews, these procedural limitations were not accounted for by the quality assessment of study design which resulted in an inaccurate estimate of clinical effectiveness. Analysis using likelihood ratios showed that screening criteria could increase the probability of zygapophysial joint pain before performing diagnostic blocks. Similar analysis showed that comparative medial branch blocks, rather than single blocks, must be used before RF neurotomy. Anatomical studies demonstrated that the shorter distal compared with the circumferential radius of the RF lesion necessitates placement of the electrode parallel to the course of the nerve along the base of the superior articular process.

Conclusions. The evidence-based procedural guidelines provide consistent criteria for multisite studies that could enroll a sufficiently large homogenous study cohort.

Introduction

Percutaneous radiofrequency (RF) neurotomy is a treatment for low back pain stemming from the zygapophysial joints of the lumbar spine. Potential candidates for this procedure are patients who experience relief of pain following diagnostic blocks of the lumbar zygapophysial joints. The procedure is performed by placing an insulated needle electrode with an exposed tip adjacent to the nerves that innervate the painful joint. A radiofrequency current applied to the electrode then heats the adjacent tissues and coagulates the nerve supply to the joint.

The efficacy of RF neurotomy for the treatment of low back pain has been evaluated in three randomized controlled trials (RCTs) [1–3]. Two of the trials demonstrated improvement in low back pain following treatment [1,2] while the third showed no benefit [3]. These results were captured by two systematic reviews which variously found the evidence for RF neurotomy to be either inconsistent or moderate [4,5]. These reviews and the original studies attracted numerous editorials and letters that either supported or questioned the validity of RF neurotomy for low back pain [6–11].

The systematic reviews addressed the conventional methodological aspects of the three studies such as method of randomization, concealment of treatment allocation, blinding, and intention to treat analysis. What the reviews did not address were various clinical procedural issues such as the validity of diagnosis in patient selection and the operative technique used. Each of these bears on the outcomes of the original studies and on the validity of the conclusions of the reviews.

Inspired by the work of Bogduk and Holmes [12], this review was undertaken to outline the procedural and technical limitations of the trials in order to demonstrate why these studies do not accurately determine the efficacy of RF neurotomy for low back pain. Second, we review the literature related to patient selection, diagnostic testing, and procedural considerations for RF of the lumbar spine. Based on this evidence, we propose diagnostic and procedural guidelines that could be used in future RCTs of lumbar spine RF. The results of the review provide a framework for designing more appropriate trials of the procedure.

Methods

A Medline and EMBASE search was performed using an OVID-based search strategy to identify RCTs as well as other studies related to low back pain stemming from the zygapophysial joints, and its treatment by RF neurotomy. The available trials of lumbar RF neurotomy were assessed with respect to the methods used to select patients and the techniques used to treat patients. The methods used in the trials were compared with data provided in other studies on the diagnosis of lumbar zygapophysial joint pain, the surgical anatomy of the medial branches of the lumbar dorsal rami, and the geometry of lesions produced by RF electrodes. This comparison served to identify deficiencies in the literature on the efficacy of RF neurotomy.

The literature on diagnosis of lumbar zygapophysial joint pain was harvested to develop an algorithm, based on Bayesian statistics, designed to provide a valid diagnosis of lumbar zygapophysial joint pain efficiently. The literature on surgical anatomy and lesion geometry was used to produce recommendations on how lumbar RF neurotomy should accurately be performed.

Results

On the treatment of lumbar zygapophysial joint pain by RF neurotomy, the literature search found the three controlled trials [1–3] and two systematic reviews [4,5] already in hand, and the observational study of RF neurotomy by Dreyfuss et al. [13]. It found two studies on the surgical anatomy of the procedure [14,15] and two on the nature of RF lesions [16,17]. On the diagnosis of lumbar zygapophysial joint pain, the search found 8 studies that provided analytical data [18–25].

Diagnosis of Lumbar Zygapophysial Joint Pain

In the controlled studies of lumbar RF neurotomy, investigators used a variety of strategies to make a diagnosis. Some used preliminary screening tests. Others used a variety of diagnostic blocks.

Gallagher et al. [1] screened patients according to eight clinical criteria that had not been validated: tenderness; exacerbation of pain on extension, rotation, exercise, or standing; pain not exacerbated by coughing or sneezing; pain referral proximal to the knee; and radiological evidence of zygapophysial joint degeneration. Patients who satisfied four or more of these criteria became eligible for diagnostic blocks of their zygapophysial joints. In the studies of van Kleef et al. [2] and Leclaire et al. [3] no screening criteria were applied. All patients with chronic nonspecific low back pain were eligible for diagnostic blocks.

The diagnostic blocks used in each of the trials differed. Leclaire et al. [3] used intra-articular blocks, Gallagher et al. [1] used a combination of periarticular and intra-articular blocks while van Kleef et al. [2] used medial branch blocks. In all three studies, only a single diagnostic block was used. Controlled blocks were not used. In only the study of van Kleef et al. [2] were segmental levels of the blocks reported. Only van Kleef used a validated scale to assess the response to blocks [2]. Across the three studies, the time of assessment of response to blocks varied from 30 minutes [2] to 1 week [3] after the block. In the study of Leclaire et al. [3], blocks were performed and responses were assessed by at least 30 different physicians. These findings are summarized in Table 1.

View this table:
Table 1

Clinical characteristics of randomized controlled trials of radiofrequency (RF) neurotomy for low back pain (LBP)

AuthorsNumber of Subjects and Follow-up PeriodsPain CharacteristicsDiagnostic BlockPain Assessment Following BlockRF Electrode PositionRF Neurotomy Parameters
TreatControlFollow-up
Gallagher et al.118121 month, 6 monthsLBP criteria specified, pain > 3-month durationIntra-articular and periarticular zygapophysial joint injection, levels not specified, single block performed by authors“Good or equivocal response” 12 hours after block, assessed by authorsJunction of pedicle and transverse process, not fluoroscopically confirmedElectrode type not specified, treatment duration 90 seconds, temperature 80°C
van Kleef et al.215162 months“Nonspecific” LBP > 1-year duration, pain distribution not specifiedMedial branch, L3, L4, L5, two levels performed at each affected joint, single block performed by authors50% reduction 30 minutes after block, measured on 4-step Likert scale, assessed by authorsCephalad to junction of transverse and superior articular processes L3–L5, junction sacral ala and S1 articular process L5–S1, fluoroscopically confirmed10-cm 22-gauge electrode 5-mm exposed tip, treatment duration 60 seconds, temperature 80±C
Leclaire et al.335311 month, 12 monthsLBP > 3-month duration, pain distribution not specifiedIntra-articular zygapophysial joint injection, levels not specified, single block performed by 30 referring physiatrists“Significant relief” for 24 hours during week following block, assessed by referring physiatristMedial branch of posterior rami, anatomical position not described, proximal and distal neurotomy, minimum 2 levels L4–5, L5–S1, electrode fluoroscopically placed10-cm 22-gauge electrode 5-mm exposed tip, treatment duration 90 seconds, temperature 80±C

The procedures used in these trials differ considerably from what has been described in the literature concerning the diagnosis of lumbar zygapophysial joint pain. That literature pertains to clinical screening and to the conduct of diagnostic blocks and their validity. The ensuing discussion focuses on the pertinent findings from the literature related to the method of patient selection for a diagnostic block and the importance of performing medial branch and L5 dorsal ramus comparative blocks (Table 2). Furthermore, these data underlie how a valid diagnosis of lumbar zygapophysial joint pain can be made efficiently.

View this table:
Table 2

Evidence-based outcomes for standardizing the methodology of radiofrequency (RF) neurotomy for low back pain (LBP)

Research ParameterAuthorsStudy InterventionOutcome
Patient selection criteriaRevel et al.19Clinical predictors of response to intra-articular zygapophysial joint block:
  • LBP relieved by recumbency plus four of the following:

  • age > 65 years or

  • LBP not exacerbated by either cough, hyperextension, forward flexion, rising from flexion, or extension-rotation

Sensitivity 0.92, Specificity 0.80
Comparative blocksKaplan et al.23Medial branch and L5 dorsal ramus block ability to anesthetize lumbar zygapophysial jointsFalse-negative rate 0.11
Dreyfuss et al.24Needle position and specificity of medial branch and L5 dorsal ramus block
  • No aberrant contrast spread with medial branch block when needle between upper border transverse process and mamillo-accessory ligament (low position)

  • 33% aberrant spread with medial branch block when needle at superior medial end of transverse process (high position)

  • 20% aberrant spread with L5 dorsal ramus block regardless of high or low needle position

  • 8% of injections were intravenous

Schwarzer et al.25Intra-articular and medial branch blocks using comparative technique to identify false-positive response rate of single diagnostic blocksFalse-negative response undetermined
Reported estimate of sensitivity 0.95
Reported estimate of specificity 0.61
Dreyfuss et al.13Comparative medial branch block performed as part of prospective trial of RF neurotomyFalse-negative response undetermined
RF anatomical and lesioning parametersBogduk et al.16Cadavaric study, anatomy of L1–L4 medial branch of the dorsal rami and L5 dorsal ramusL1–L4, medial branch fixed proximal at origin from dorsal ramus at and distally under mammillo-accessory ligament at caudal edge of superior articular process L5–S1, dorsal ramus passes over ala of sacrum in bony groove at base of superior articular process and sacral ala
Bogduk et al.16Lesion size produced in ex vivo muscle tissue by 22 and 18 gauge electrodes with 4-mm and 5-mm exposed tip, respectively, at 80°C for 2 minutes22 gauge, mean circumferential and distal radius 1.9 ± 0.3 mm and 1.1 ± 0.3 mm; 18 gauge, mean circumferential and distal radius 2.2 ± 0.3 mm and 1.4 ± 0.5 mm
Goldberg et al.17Lesion size variation in ex vivo muscle and liver samples with change in electrode gauge, exposed tip length, lesioning duration, temperatureElectrode gauge, lesion radius and length increased from 24 to 18 gauge, no further increase from 18 to 15 gauge; Exposed tip length, lesion radius increased with increased exposed tip length from 0.5 to 1.0 cm, no further increase at 2 cm; Duration, at 1 and 2 minutes time 64% and 73% maximal radius achieved, respectively; Temperature, lesion size unchanged at 80°C and 90±C

In a prospective study, clinical predictors of response to intra-articular zygapophysial joint anesthesia included advanced age, low back pain relieved by recumbency, and low back pain not exacerbated by coughing, hyperextension, forward flexion, rising from flexion, or extension-rotation [18]. These clinical characteristics were further evaluated in a double-blind RCT [19]. When five of these seven clinical characteristics were present, in addition to relief of low back pain with recumbency, the sensitivity and specificity was 0.92 and 0.80, respectively. Relief of low back pain was defined as a 75% reduction in pain following an intra-articular zygapophysial joint injection with 1 mL of 2% lidocaine. These clinical characteristics were intended to be used as selection criteria for RCTs that involved patients with zygapophysial joint pain. Important limitations related to the application of the criteria in a RF trial include the use of an intra-articular versus a medial branch block. Other important shortcomings include the use of a single diagnostic block and administration of periarticular corticosteroids. Despite these limitations, the screening criteria of Revel et al. [19] is an easily administered bedside instrument that could increase the probability of zygapophysial joint pain among patients referred for a diagnostic block, as evidenced by use of the likelihood ratio (LR).

The LR (sensitivity/[1 − specificity]) of a diagnostic test is used to calculate the post-test probability, or post-test odds, of disease among patients with a positive test result as follows: Embedded Image

The prevalence of the disorder among patients to be tested serves as the pretest probability of disease. In order to use the above equation, the pretest probability must be converted to odds, termed the pretest odds. Likewise, the post-test odds can be converted to a probability. When diagnostic tests are used in succession, the post-test probability of the first test can function as the pretest probability for a subsequent test. Using the sensitivity (0.92) and specificity (0.80) reported in the study of Revel et al. [19], the LR ratio for the screening criteria was calculated to be 4.6. The 95% confidence intervals (95% CI) for the sensitivity, specificity, and LR could not be calculated from the published data. The estimated prevalence of zygapophysial joint pain among patients with low back pain ranges from 15% to 40%[20,21]. Assuming the true prevalence is between these two values, the mid-point of 27% was used to estimate the pretest probability (0.27) of zygapophysial joint pain. Following conversion of the pretest probability to pretest odds (0.37), the post-test odds was calculated as follows: Embedded Image

The post-test odds convert to a post-test probability is 0.63. The post-test probability of 0.63 now represents the pretest probability of zygapophysial joint pain before performing diagnostic blocks. Thus, the screening criteria of Revel et al. [19] increased the pretest probability of disease from 0.27 to 0.63. If the screening criteria were used to identify potential subjects for a clinical trial of RF neurotomy, fewer patients would need to undergo diagnostic blocks.

The primary nociceptive afferents of the zygapophysial joint can be blocked by anesthetizing the medial branch of the lumbar dorsal rami or the L5 dorsal ramus at the level of the targeted joint and one level cephalad. While both an intra-articular and medial branch block can anesthetize the zygapophysial joint [22], improvement in low back pain following an intra-articular block does not provide anatomically based evidence that a subsequent neurotomy of the nociceptive afferents of the joint would produce a similar response. Important considerations related to the use of a medial branch block as a diagnostic procedure before RF neurotomy include the ability of the block to anesthetize the zygapophysial joint, the specificity of the block to anesthetize the medial branch as opposed to anesthetizing adjacent neural structures, and the LR of a single diagnostic block. These three aspects of a medial branch block will be addressed separately.

In a single-blind RCT that involved 14 asymptomatic volunteers, lumbar medial branch blocks using 0.5 mL of 2% lidocaine were used to inhibit a provocative pain stimulus in the study participants [23]. The pain stimulus was generated by distention of the zygapophysial joint capsule with contrast medium. A false-negative response rate of 0.11 was identified when anesthetization of the medial branches failed to inhibit the provocative pain stimulus. The etiology of the false-negative rate remained undetermined. However, it was postulated that failure to deliver an adequate concentration of local anesthetic around the targeted nerve, because of diffusion or uptake by adjacent venous structures, could have contributed to the false-negative rate.

The specificity of lumbar medial branch or L5 dorsal ramus blocks was evaluated in a cross-sectional study involving 15 volunteers who underwent 120 medial nerve branch blocks at L2 through L5 [24]. Spread of 0.5 mL of contrast medium toward adjacent structures occurred during 10 of 30 injections performed at the L2 level when the needle tip was positioned on the “dorsal surface of the transverse process immediately below the medial end of its superior border”[24]. Aberrant spread of contrast was eliminated at the L3 and L4 levels when the needle was properly placed “midway between the upper border of the transverse process and the mamillo-accessory ligament”[24]. However, at L5, placement of the needle midway between the two described positions reduced, but did not eliminate, aberrant spread to the L5–S1 intervertebral foramen. Administration of contrast medium revealed intravenous uptake in 8% of injections.

A comparative intra-articular or medial branch block of a single lumbar zygapophysial joint refers to a series of two injections where local anesthetic agents of varying durations are administered. Following the first injection, patients who experience a reduction in pain intensity consistent with the clinical duration of the administered medication undergo a second confirmatory block using a local anesthetic agent with a different duration of action. When two or more zygapophysial joints are evaluated, each joint must first be separately blocked to demonstrate incomplete pain relief when only one joint has been anesthetized. Next, all suspected joints must be blocked simultaneously and, depending on the pain response, a multilevel confirmatory block must be performed. Assuming the initial and confirmatory injections were properly performed, patients who do not respond to the second block are identified as having experienced a false-positive response to the first block. When the comparative block technique is used, a positive response to both injections serves as the criterion standard for identifying patients with low back pain attributed to the zygapophysial joint. However, the potential false-positive response rate of the second or any number of subsequent confirmatory injections remains undetermined, which ultimately limits the specificity of this technique.

Comparative intra-articular or medial branch blocks were performed in an unblinded prospective study that involved 176 consecutive patients with chronic low back pain [25]. Seventy-nine patients received an intra-articular block (45%), 70 received medial branch blocks (40%), and 27 had both blocks performed (15%). The initial block was performed using 0.5 mL of 2% lidocaine where a positive response was defined as “complete” or “definite” relief of pain following the injection. Eighty-three of 176 patients (47%) had a positive response to the initial block while the 93 patients (53%) who had a negative response were excluded from further evaluation. Confirmatory blocks were performed 2 weeks later on 71 of 83 patients who responded to the initial block. The 12 patients who did not receive confirmatory blocks either had no recurrence of pain or had logistical difficulties that prevented return to the study site. However, these 12 patients were retained in the analysis of the 83 patients who responded to the initial block. Following the confirmatory block with 0.5 mL of 0.5% bupivacaine, visual analog scale (VAS) ratings were obtained 30 minutes, 1 hour, and then hourly for a total period of 8 hours. A positive response to the confirmatory block, defined as a 50% reduction in pain that was sustained for ≥3 hours, was reported by 26 of 83 patients (31%). The 57 of 83 patients (69%) who had a negative response to the confirmatory block, including the 12 patients who did not receive the second injection, were identified as having experienced a false-positive response to the initial lidocaine block.

Comparative medial branch blocks were used to identify patients with zygapophysial joint pain as part of a prospective trial of RF neurotomy [13]. Four hundred and sixty patients were initially screened of which 41 (9%) underwent medial branch blocks. In this study, the patients were blind to the type of local anesthetic administered. The initial block was performed with 0.5 mL of 2% lidocaine where a positive response was defined as an 80% reduction in pain for 1 hour following the injection. Twenty-two of 41 subjects (54%) experienced a positive response to the initial block. The 19 patients (46%) who had a negative response were excluded from further evaluation. A positive response to the confirmatory block, which was performed using 0.5 mL of 0.5% bupivacaine, was defined as an 80% reduction in pain for 2 hours following the injection. Fifteen of 22 patients (68%) responded to the confirmatory block while the seven patients (32%) who had a negative response were considered to have experienced a false-positive response to the initial lidocaine block.

Further analysis and interpretation of these results requires clarification of the important differences that exist between the two studies. The study of Schwarzer et al. [25] consisted of a consecutive series of patients while the cohort of Dreyfuss et al. [13] was composed of subjects actively recruited to participate in a research trial. A combination of unblinded intra-articular and medial branch blocks was used in the study of Schwarzer et al. [25] compared with the other study which used patient-blinded medial branch blocks. In comparing the results, a similar proportion of patients responded to the initial lidocaine block. However, the proportion of patients who responded to the second confirmatory block in the study of Dreyfuss et al. [13] was 0.68 compared with 0.31 in the study of Schwarzer et al. [25]. The difference in this response was statistically significant (chi-square 9.93, df = 1, P = 0.002). The most likely explanation for this difference was the more stringent response criteria used for the initial lidocaine block by Dreyfuss et al. [13] which required an 80% reduction in pain intensity, compared with a 50% reduction in the study of Schwarzer et al. [25]. Consequently, this increased the specificity and reduced the false-positive response rate of the initial diagnostic block in the study of Dreyfuss et al. [13]. While both groups of investigators identified an unacceptable false-positive response rate following the initial lidocaine block, further analysis of these data using LRs will demonstrate the necessity of performing comparative blocks before RF neurotomy.

The results of the initial and confirmatory blocks from the two individual studies [13,25] were organized in 2 × 2 contingency tables (Figure 1). However, careful review of the response data reveals that an LR can not be calculated because the number of responses that should occupy box c, which represents the false-negative rate, remains unknown [13,25]. This occurred because a second lidocaine block was not performed on the patients who failed to respond to the initial block. If a second lidocaine block had been performed, patients who were incorrectly classified as true-negative responders by the first block would have been correctly identified as false-negative responders. For purposes of calculating the positive and negative predictive values of a single diagnostic block, Schwarzer et al. [25] estimated the false-negative rate to be 5%. A more accurate estimate can now be obtained from the study of Kaplan et al. [23], which showed that medial branch blocks have a false-negative response rate of 0.11. When this proportion was used as an estimate of the false-negative rate for a single diagnostic block in the Schwarzer [25] and Dreyfuss [13] studies, the number of false-negative responses, rounded to a whole integer, was calculated to be 3 and 2, respectively. This reduced the reported number of true-negative responses in each study by the respective number of estimated false-negative responses.

Figure 1

2 × 2 contingency tables incorporating estimated false-negative responses with sensitivities, specificities, likelihood ratios, and 95% confidence intervals for (A) Schwarzer et al.25 and (B) Dreyfuss et al.13 comparative block responses.

The LRs and 95% CI [26] for each diagnostic block study are presented in Figure 1. The LR from the data of Dreyfuss et al. [13] is greater than the LR from Schwarzer et al. [25]. Because the sensitivity and false-negative response rate of both studies were estimated to be approximately 0.89 and 0.11, respectively, the difference in LRs can be explained by the variation in the specificity and false-positive response rate of each study. Using the LR of 3.1 from Dreyfuss et al. [13], and the estimated pretest probability of 0.27 (pretest odds 0.37), the post-test odds of zygapophysial joint pain following a single diagnostic block was calculated as follows: Embedded Image

The corresponding post-test probability was 0.53. Using the upper limit of the 95% CI for the LR and the upper limit of zygapophysial joint pain prevalence as the pretest probability (0.40), the post-test probability was 0.79. Conversely, when the lower limits of the 95% CI for the LR (1.6) and disease prevalence (0.15) were used, the post-test probability was 0.22. Therefore, these analyses demonstrate that a large proportion of patients treated with RF neurotomy on the basis of a positive response to a single diagnostic block do not have zygapophysial joint pain. Use of the LR to calculate the post-test probability provides epidemiological evidence that comparative blocks must be used in a clinical trial of RF neurotomy. Furthermore, this epidemiologically based evidence demonstrates that the use of single diagnostic blocks in the three RCTs of RF neurotomy invalidates the results of these trials because a large and indeterminate proportion of the study cohort did not have zygapophysial joint pain [1–3]. These findings, including the analysis of the screening criteria of Revel et al. [19], and the range of proportionate responses to comparative blocks [13,25], are summarized in Figure 2.

Figure 2

Flow chart for LR of Revel et al.19 screening criteria, pre- and post-test probability of single diagnostic blocks, and the proportionate responses to comparative blocks for the Schwarzer et al.*25 and Dreyfuss et al.**13 studies.

Treatment of Lumbar Zygapophysial Joint Pain with Radiofrequency Neurotomy

The medial branch of the posterior primary ramus was the intended target of RF neurotomy in all three randomized trials [1–3]. The needle electrodes were placed under fluoroscopic guidance in the trials of van Kleef et al. [2] and Leclaire et al. [3]. However, radiographic confirmation of the electrode position was published in only the van Kleef study. A 22-gauge electrode with a 5-mm exposed tip was used in the trials of van Kleef et al. [2] and Leclaire et al. [3], but the electrode type was not specified in the study of Gallagher et al. [1]. A single neurotomy was performed in the trials of Gallagher et al. [1] and Leclaire et al. [3], while both a proximal and distal neurotomy was performed in the study of van Kleef et al. [2]. The RF neurotomies were created by elevating the temperature of the electrode to 80°C and the duration of temperature elevation varied from 60 seconds to 90 seconds [1–3].

Knowledge of the zygapophysial joint anatomy and innervation should govern electrode placement and selection of lesioning parameters. The anatomy of the L1 through L4 medial branch of the dorsal rami, and the L5 dorsal ramus, has been well demonstrated in cadaveric studies [14,15]. The course of the medial branch at L1 through L4 has been shown to be fixed proximally at its origin from the dorsal ramus at the superior aspect of the transverse process and distally as the nerve passes under the mamillo-accessory ligament at the caudal edge of the superior articular process. Between these two fixed positions, the course of the medial branch across the base of the superior articular process does not vary. At the L5 level, the dorsal ramus passes over the ala of the sacrum in the bony groove formed by the base of the superior articular process and ala of the sacrum. The caudal origin of the medial branch at this level is not accessible by a percutaneous approach therefore the target nerve at L5 is the dorsal ramus.

The superior portion of each lumbar zygapophysial joint receives innervation from the medial branch originating one level cephalad, while the inferior portion receives innervation from the medial branch originating at the same level [14,15]. These anatomical findings confirm that at the L1 through L4 levels the trajectory of the electrode should follow the course of the nerve along the base of the superior articular process with the tip positioned at the superior aspect of the groove formed by the junction of the transverse process and superior articular process. At the L5 level, the electrode should rest in the groove between the ala of the sacrum and the superior articular process. The final position of all electrodes should be verified radiographically.

Variations in RF stimulation parameters and the resultant change in lesion diameter and length were evaluated in two series of observational experiments [16,17]. In the first study, fifteen 18-gauge electrodes with a 5-mm exposed tip were inserted into an ex vivo muscle sample and stimulated for 120 seconds at 80–85°C [16]. The mean maximal circumferential radius of coagulated tissue was 2.2 ± 0.3 mm while the mean distal radius, measured from the tip of the electrode to the distal edge of the lesion, was 1.4 ± 0.5 mm. Under similar conditions, the mean maximal circumferential and distal radius of coagulation created by eleven 22-gauge electrodes with a 4-mm exposed tip was 1.9 ± 0.3 mm and 1.1 ± 0.3 mm, respectively. In the second series of observational experiments, the radius of tissue coagulation increased as the electrode size was increased from 24 gauge to 18 gauge [17]. However, no further increase in radius occurred when the electrode size was increased from 18 gauge to 15 gauge [17]. The lesion radius produced by an 18-gauge electrode plateaued at 4.5 mm as the tip exposure was increased from 0.5 cm to 1.0 cm. As the duration of stimulation was lengthened to 6 minutes, the volume of coagulation achieved 64% and 73% of the maximal radius at 1 and 2 minutes time, respectively. The conventional stimulation period of 90 seconds would be expected to create an area of coagulation somewhere between these two values. The radius of tissue coagulation produced at 80±C and 90±C were similar.

Discussion

Significant dissonance exists between the primary RF literature and the procedure that was actually performed in the three RCTs [1–3]. This compromises any conclusions that can be drawn from the results of these trials. Furthermore, these disparate findings were further confounded by the conclusions of two systematic reviews which focused on study design rather than the fundamental clinical and anatomical aspects of the procedure [4,5]. The procedural and technical methods used in the open prospective trial by Dreyfuss et al. [13] were exemplary and the study provides a benchmark of optimal outcome.

A summary of the procedural guidelines is outlined in Table 3. Patients should be selected to undergo blind comparative diagnostic medial branch blocks using the identified clinical criteria which could increase the pretest probability of zygapophysial joint pain. The unacceptable false-positive response rate of single diagnostic blocks requires the use of comparative blocks, which should be performed with contrast medium to detect inadvertent vascular uptake before the injection of local anesthetic. The needle and electrode positions for the diagnostic blocks and RF neurotomy should be fluoroscopically confirmed.

View this table:
Table 3

Radiofrequency (RF) neurotomy procedural guidelines for treatment of low back pain (LBP)

Patient selection criteria for a diagnostic medial branch blockLBP relieved by recumbency plus any four of the following including age > 65 years or LBP not exacerbated by either cough, hyperextension, forward flexion, rising from flexion, or extension-rotation
Diagnostic medial branch block
  Needle positionBetween upper border transverse process and mamillo-accessory ligament, confirmed with fluoroscopy and contrast medium
  ComparativeLidocaine 2% 0.5 mL, bupivacaine 0.5% 0.5 mL, subjects blind to injectate
  Assessment of pain responsePain assessment 30 minutes following block and hourly for 6 hours, ≥ 80% pain relief for at least 1 hour following lidocaine and 3 hours following bupivacaine block
RF parameters
Needle positionL1–L4; electrode tip at superior aspect of groove formed by junction of transverse process and superior articular process
L5; electrode tip at groove between ala of sacrum and superior articular process. Positions confirmed with fluoroscopy
  Electrode size18-gauge, 1-cm exposed tip
  Lesioning parameters120-second duration at 80°C

The lesioning data presented herein, in conjunction with the anatomical findings, show that the shorter distal compared with the circumferential radius of the RF lesion necessitates placement of the electrode parallel to the course of the nerve along the base of the superior articular process. An 18-gauge RF electrode with a 1-cm exposed tip stimulated for 120 seconds at 80°C should create an area of tissue coagulation with a sufficient circumferential and distal radius to fully encompass the targeted nerve.

As a result of procedural limitations, the efficacy of RF neurotomy for lumbar zygapophysial joint pain remains undetermined despite three RCTs. The evidence-based procedural guidelines developed in this review provide consistent criteria for multisite studies that could enroll a sufficiently large homogenous study cohort. These guidelines would enable the screening of fewer patients in order to maximize the yield of diagnostic blocks and would increase the technical accuracy of the RF procedure. This, in turn, would ensure that patients enrolled in a RCT actually have zygapophysial joint pain and that subjects in the treatment arm would actually receive the procedure.

References

View Abstract