Lumbar zygapophysial joint arthropathy is a challenging condition affecting up to 15% of patients with chronic low back pain. The onset of lumbar facet joint pain is usually insidious, with predisposing factors including spondylolisthesis, degenerative disc pathology, and old age. Despite previous reports of a "facet syndrome," the existing literature does not support the use of historic or physical examination findings to diagnose lumbar zygapophysial joint pain. The most accepted method for diagnosing pain arising from the lumbar facet joints is with low-volume intraarticular or medial branch blocks, both of which are associated with high false-positive rates. Standard treatment modalities for lumbar zygapophysial joint pain include intraarticular steroid injections and radiofrequency denervation of the medial branches innervating the joints, but the evidence supporting both of these is conflicting. In this article, the authors provide a comprehensive review of the anatomy, biomechanics, and function of the lumbar zygapophysial joints, along with a systematic analysis of the diagnosis and treatment of facet joint pain.
SINCE its original description almost 100 yr ago, thousands of scientific articles have been published on lumbar zygapophysial (l-z) joint pain, and facet interventions represent the second most common type of procedure performed in pain management centers throughout the United States.1But despite the plethora of research and clinical emphasis on this disorder, almost every aspect of l-z joint pain, from diagnosis to treatment, remains mired in controversy. Even among pain specialists, lumbar facet joint pain remains a misunderstood, misdiagnosed, and improperly treated medical condition. In this article, l-z joint pain is defined in a functional capacity as pain originating from any structure integral to both the function and configuration of the lumbar facet joints, including the fibrous capsule, synovial membrane, hyaline cartilage surfaces, and bony articulations.
A critical issue that must be addressed before embarking on any review of l-z joint pain is whether pain can be definitively attributed to these joints. In the 1960s and 1970s, this question posed a legitimate controversy that was vigorously debated in the medical literature.2,3However, in the past 20 yr, the scales of this controversy have resolutely tipped toward the conviction that l-z joints can be and often are a primary source of low back pain (LBP).4
Compelling evidence underlies this paradigmatic shift in thinking. The facet joint capsule and surrounding structures are richly imbued with nociceptors that fire when the capsule is stretched or subjected to local compressive forces.5,6In both pain patients and volunteers, chemical or mechanical stimulation of the facet joints and their nerve supply has been shown to elicit back and/or leg pain.7–10During spine surgery performed under local anesthetic (LA), lumbar facet capsule stimulation elicits significant pain in approximately 20% of patients.11Last and most significantly, LA blocks of either the facet joints themselves or the medial branches innervating them have been shown to relieve pain in a substantial percentage of patients with chronic LBP.12Therefore, like other synovial joints in the human body, the l-z joints represent a potential pain generator in patients with chronic LBP.
In light of the prevalence of lumbar facet joint pain and the frequency with which facet blocks are performed, several reviews have been undertaken on this phenomenon in the past 15 yr. Although some of these articles provided keen insight on various aspects of the condition, most were limited by their confined scope and specialized target audiences. The purpose of this review is therefore to provide a comprehensive, evidence-based framework on the anatomy, pathophysiology, prevalence, diagnosis, and treatment of lumbar facet pain. Articles reviewed were obtained via MEDLINE and Ovid search engines, books and book chapters, and bibliographic references dating to the early 1900s.
Historic Review
The l-z joints, often referred to as facet joints, have long been recognized as a potential source of LBP. In 1911, Goldthwaite13first noted that the peculiarities of the facet joints could be a significant source of back pain and instability. Sixteen years later, after anatomical dissections of 75 cadavers, Putti14suggested that local inflammation and degenerative changes in lumbar facet joints could result in sciatica from irritation of nerve roots. In 1933, Ghormley15coined the term “facet syndrome,” which he defined as lumbosacral pain, with or without sciatica, that was likely to occur after a sudden rotatory strain. Shortly thereafter, interest in the l-z joints as potential sources of back pain waned after the landmark article by Mixter and Barr16implicating lumbar disc rupture as the major cause of low back and leg pain. The 1940s saw a resurgence in the interest of l-z joints as pain generators when Badgley17suggested that up to 80% of cases of LBP and sciatica are due to referred pain from l-z joint pathology, rather than direct nerve root compression. In 1963, Hirsch et al. 7published the first account whereby the injection of l-z joints reproduced patients’ back pain. Rees,18who reported a success rate of 99.8%, is generally credited with promoting percutaneous “facet rhizolysis” with his ground-breaking report in the early 1970s. Later, it was shown that the technique advocated by Rees may not have been sufficient to achieve rhizotomy in most patients.3Buoyed by a high incidence of hemorrhagic complications, Shealy19,20pioneered the use of fluoroscopically guided radiofrequency facet denervation to treat l-z joint pain in the mid-1970s.
Anatomy and Innervation
The lumbar facet joints form the posterolateral articulations connecting the vertebral arch of one vertebra to the arch of the adjacent vertebra. As true synovial joints, each facet joint contains a distinct joint space capable of accommodating between 1 and 1.5 ml of fluid, a synovial membrane, hyaline cartilage surfaces, and a fibrous capsule.21The fibrous capsule of the lumbar facet joint is approximately 1 mm thick and composed mostly of collagenous tissue arranged in a more or less transverse fashion to provide maximum resistance to flexion.22,23The joint capsule is thick posteriorly, supported by fibers arising from the multifidus muscle. Superiorly and inferiorly, the capsule attaches further away from the osteochondral margins, forming subcapsular recesses that in the normal joint are filled with fibroadipose menisci. Anteriorly, the fibrous capsule is replaced by the ligamentum flavum.24
Each facet joint receives dual innervation from medial branches arising from posterior primary rami at the same level and one level above the z-joint.25,26For example, the inferior pole of the L4–L5 facet joint receives innervation from the L4 medial branch and its superior pole is innervated by the L3 medial branch, which are typically blocked on the transverse processes of L5 and L4, respectively. The medial branches of L1–L4 dorsal rami course across the top of their respective transverse processes one level below the named spinal nerve (e.g. , L4 crosses the transverse process of L5), traversing the dorsal leaf of the intertransverse ligament at the base of the transverse process. Each nerve then runs downward along the junction of the transverse and superior articular processes, passing beneath the mamilloaccessory ligament and dividing into multiple branches as it crosses the vertebral lamina (fig. 1). In some cases, the mamilloaccessory ligament becomes calcified, which may lead to nerve entrapment.27This is most common at L5 (approximately 20%) but also occurs at L4 (10%) and L3 (4%). In addition to two l-z joints, the medial branches also innervate the multifidus muscle, the interspinous muscle and ligament, and the periosteum of the neural arch.7,28–30The L5 nerve differs in that it is the dorsal ramus itself that runs along the junction of the sacral ala and superior articular process of the sacrum.29,31Its medial branch arises opposite the inferolateral corner of the base of the lowest facet joint. At this level, it is the dorsal ramus rather than its medial branch that is amenable to blockade.
Fig. 1. Right lateral oblique view of the lumbar vertebral bodies and the dorsal rami medial branches. Drawing by Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland).
Some people may have aberrant or additional innervation of the facet joints. In a study conducted in asymptomatic volunteers, Kaplan et al. 32found that 1 in 9 subjects who underwent L4–L5 (n = 5) and L5–S1 (n = 4) facet joint medial branch blocks (MBB) felt pain during repeated capsular distension concordant with pain experienced during previous l-z joint provocation without MBB. Assuming the blocks were technically successful, one explanation for this finding is that in a small subset of individuals, aberrant innervation may account for false-negative diagnostic MBB. Some authors have suggested that l-z joints may also receive innervation from the medial branch below the joint (i.e. , S1 for the L5–S1 joint),33,34the dorsal root ganglion,35and the paravertebral sympathetic ganglia,36,37but these assertions remain unproven (fig. 2).
Fig. 2. Schematic drawing of the spinal cord and segmental spinal innervation. Drawing by Specialist Frank and Angela Dill, US Army, and Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland).
Histologic studies have demonstrated that the lumbar facet joints are richly innervated with encapsulated (Ruffini-type endings, pacinian corpuscles), unencapsulated, and free nerve endings.5The presence of low-threshold, rapidly adapting mechanosensitive neurons suggests that in addition to transmitting nociceptive information, the l-z facet capsule also serves a proprioceptive function. Besides substance P and calcitonin gene–related peptide, a substantial percentage of nerve endings in facet capsules have also been found containing neuropeptide Y, indicating the presence of sympathetic efferent fibers.38,39Nerve fibers have also been found in subchondral bone and intraarticular inclusions of l-z joints, signifying that facet-mediated pain may originate in structures besides the joint capsule.40–42In degenerative lumbar spinal disorders, inflammatory mediators such as prostaglandins43and the inflammatory cytokines interleukin 1β, interleukin 6, and tumor necrosis factor α44have been found in facet joint cartilage and synovial tissue.
Function and Biomechanics
The basic anatomical unit of the spine, often referred to as the three-joint complex, consists of the paired zygapophysial joints and the intervertebral disc. Together, these joints function to support and stabilize the spine, and prevent injury by limiting motion in all planes of movement. Macroscopically, each facet joint is composed of a posteromedially facing concave superior articular process from the inferior vertebral body, and a smaller anterolaterally facing inferior articular process from the superior spinal level. The shape and orientation of the l-z joints determine the role each plays in protecting the spine against excessive motion. Facet joints oriented parallel to the sagittal plane provide substantial resistance to axial rotation but minimal resistance to shearing forces (backward and forward sliding), whereas joints oriented more in a coronal plane tend to protect against flexion and shearing forces but provide minimal protection against rotation (fig. 3).24In an anatomical study published in 1940 by Horwitz and Smith,45the authors found that the L4–L5 z-joints tended to be more coronally positioned (almost 70° with respect to the sagittal plane), whereas the L2–L3 and L3–L4 joints were likely to be oriented more parallel (< 40°) to the sagittal plane. In more recent studies by Masharawi et al. 46and Punjabi et al. ,47the investigators found that the upper lumbar facet joints (T12–L2) were oriented closer to the midsagittal plane of the vertebral body (mean range, 26°–34°), whereas the lower facet joints tended to be oriented away from that plane (40°–56°). In the upper lumbar spine, approximately 80% of the facet joints are curved, and 20% are flat. In the lower lumbar spine, these numbers are reversed.45Studies by Grobler et al. 48and Boden et al. 49found a positive association between degenerative spondylolisthesis and more sagittally oriented lower lumbar facet joints.
Fig. 3. Segmental variation in lumbar zygapophysial facet joint orientation in the transverse plane. Adapted from Taylor and Twomey 277,and Masharawi et al. 46 Drawings by Specialist Frank and Angela Dill, US Army, and Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland).
Fig. 3. Segmental variation in lumbar zygapophysial facet joint orientation in the transverse plane. Adapted from Taylor and Twomey 277,and Masharawi et al. 46 Drawings by Specialist Frank and Angela Dill, US Army, and Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland).
In young people, the l-z joints are quite strong, capable of supporting almost twice body weight.23As aging occurs, the joints become weaker and more biplanar, transitioning from a largely coronal orientation to a more prominent sagittal positioning. The orientation of each joint to the sagittal plane can also differ between the paired facet joints at the same spinal level. This phenomenon is known as tropism and has a 20–40% incidence among the general population.24,50Most50–55but not all49,56studies have demonstrated a positive relation between facet joint tropism and intervertebral disc degeneration and herniation.
Although most of the axial load is borne by the intervertebral discs, the two other components of the three-joint complex, the l-z joints, also play a role in weight bearing, with the proportion dependent on several factors. In a study by Yang and King,57the authors determined that normal facet joints typically carry 3–25% of the axial burden, which can increase even higher in patients with degenerative disc disease facet arthritis. During this seminal experiment, it was observed that with significant facet loading the inferior tip of the facets bottomed against the laminae below, functioning as pivots for the entire vertebral body to rotate backward. This resulted in considerable stretching of the superior facet capsules, manifesting as blood extravasating onto the outer surface of the capsule. In another cadaveric study, Adams and Hutton58found that the lumbar facet joints resist approximately 16% of the intervertebral compressive force when standing erect versus near 0% when sitting, which helps explain the high intradiscal pressures during unsupported sitting.59Finally, after conducting in vitro experiments measuring induced loads on cadaveric lumbar facets, Lorenz et al. 60concluded that with increasing compressive loads, the absolute facet loads increases only slightly, so that the proportion of the axial burden borne by the l-z joints actually decreases with increasing stress. During extension, the normal load on the facet joints is always higher when compared with the neutral mode. After facetectomy, while the remaining load on the vestigial facet is substantially reduced, the peak pressure increases.60In summary, the l-z joints serve primarily a protective role, functioning to limit movement in all planes of motion. Whereas they do play a part in weight bearing, the proportion of the axial load they bear is inversely proportional to the amount of stress.
Mechanisms of Injury
Cadaveric and Animal Studies
Although in rare instances the development of facet joint arthropathy can be traced to a specific inciting event,61the overwhelming majority of cases of l-z joint pain are the result of repetitive strain and/or low-grade trauma accumulated over the course of a lifetime. Khalsa et al. conducted a series of cadaveric experiments designed to determine which physiologic movements are associated with the greatest degree of facet joint capsule strains. They found that the joint moments measured at any given motion increased with greater magnitudes of joint displacement, and were significantly larger in the two most caudad facet joints (L4–L5 and L5–S1).62With lateral bending, strains of the joint capsule tended to be larger in magnitude in the three most caudad joints (L3–S1) during contralateral flexion (i.e. , the left facet joints are most strained during right lateral flexion), whereas the two most cephalad joints (L1–L2 and L2–L3) bore the greatest strain during bending to the ipsilateral side. For the upper three facet joints, the maximum joint displacement and greatest strain was associated with lateral bending, usually to the right. For the two lowest joints, the greatest degree of strain occurred during forward flexion (table 1).
Table 1. Motions Associated with the Largest Intervertebral Angulation and Strain for the Lumbar Facet Joints

In a follow-up study, the authors fixated human lumbar spine specimens with a single anterior thoracolumbar plate on L4–L5, and then measured capsular displacement and strains for a wide range of physiologic motions.63For all motions, increased joint moments occurred at the level of fixation and both adjacent levels. There was also an increase in intervertebral angulation at L3–L4 and L5–S1, and decreased motion at L4–L5. Increased strain was noted bilaterally at L3–L4 and L5–S1, and contralaterally at L4–L5. On the side ipsilateral to the fixation, decreased strain was found at L4–L5. These findings support the work of other investigators demonstrating that fusing two vertebrae results in a transfer of motion previously occurring at the operated level to adjacent segments, particularly L5–S1, which in turn can lead to accelerated degeneration.64–68
In a subsequent study, Little and Khalsa69found that both sustained and repetitive lumbar flexion increased capsular motion and joint strain from L3–L4 to L5–S1, with creep developing more rapidly during sustained flexion than with repetitive movements. Interestingly, these parameters remained elevated even after a 20-min recovery period. Although these studies provide a theoretical framework for the development of microinjury associated with repetitive spine movements, cadavers cannot experience pain, and the viscoelastic material properties factored into computational cadaveric spine models may differ from those found in live patients.70
In response to repetitive strain and inflammation, the synovial l-z joints can fill with fluid and distend, resulting in pain from stretching the joint capsule.71Distension of the articular recesses can also compress the exiting nerve root in the neural foramen or spinal canal, especially when the foramen is already narrowed by joint hypertrophy and/or osteophytes. 72–75This can lead to concomitant sciatica and mask the facet pathology underlying the radiculopathy. Capsular irritation may also result in reflex spasm of the erector spinae, multifidus, and other paraspinal muscles.72,76,77
In in vivo and in vitro experiments designed to reproduce facet capsular stretch, nerves typically cease firing shortly after the stimulus is removed.78–80Although these findings indicate that capsular strain could cause acute facetogenic pain, they do not explain its persistence. The pathophysiologic basis for persistent lumbar facet pain was established in a series of elegant experiments conducted by Cavanaugh, Yamashita, Ozaktay, et al. in New Zealand white rabbits. In these studies, the application of inflammatory chemicals or algesic mediators such as substance P and phospholipase A2, were found to result in inflammatory changes consisting of vasodilation, venous congestion, and the accumulation of polymorphonuclear leukocytes. Neuronal sensitization occurred in both nociceptive and proprioceptive nerve endings, being manifested by reduced mechanical thresholds, increases in multiunit discharge rate and recruitment of previously silent units.81–84Persistent nociceptive input invariably leads to peripheral sensitization, and if the underlying stimulus is not removed, central sensitization and neuroplasticity can develop.85Whereas the spinal structures in humans are subject to greater stress than other mammals, because of the array of anatomical and functional differences that exist between the l-z joints in various mammalian species,86caution should be exercised when extrapolating the results of animal studies to humans.
One inferential interpretation of these preclinical findings is that chronic l-z joint pain is likely to occur with repetitive, chronic strains as might be seen in the elderly or, less frequently, after an acute event such as tearing the joint capsule by stretching it beyond its physiologic limits. This hypothesis is supported by clinical studies indicating a higher prevalence of facet arthropathy in elderly patients87–89and numerous cases of lumbar facet arthropathy after high-energy trauma.61
Human Studies
Clinically, several conditions may predispose individuals to chronic facet joint strain. Radiologic studies conducted in LBP patients and asymptomatic controls have shown a positive correlation between sagittally oriented facet joints and degenerative spondylolisthesis.48,49In these patients, recurrent rotational strains result in myriad changes to the discs and paired l-z joints, including loss of disc height, osteophyte formation, and degenerative hypertrophy of the facets.90,91
Changes in any component of the three-joint spinal unit lead to predictable changes in the other components. Degeneration and loss of structural integrity of the intervertebral discs have been shown to result in concomitant degenerative changes in the l-z joints.92–94The reverse is also true. Degeneration and motion abnormalities at the l-z joints can induce and accelerate degeneration of the intervertebral discs.58,95,96In a magnetic resonance imaging (MRI) study evaluating the relation between facet joint osteoarthritis and degenerative disc disease (DDD), Fujiwara et al. 97found that facet joint osteoarthritis was rarely found in the absence of disc degeneration but tended to be most pronounced at spinal levels associated with advanced DDD. The authors concluded that disc degeneration is a more reliable indicator of aging than facet joint osteoarthritis, and in most people, DDD precedes facet osteoarthritis.
Paradoxically, in the only clinical study evaluating the relative contributions of DDD and facet arthropathy to chronic LBP, Schwarzer et al. 98found the combination of discogenic and l-z joint pain to be a relatively rare occurrence. In 92 patients who underwent both discography and confirmatory l-z joint blocks with lidocaine and bupivacaine, 39% had at least one positive discogram with a negative control disc, and 9% obtained concordant pain relief after the series of analgesic facet joint blocks. But only 3% of patients had both positive discography and a symptomatic l-z joint. The discrepancy between the handful of basic science studies demonstrating a correlation between DDD and facet joint degeneration and the lone clinical study finding minimal overlap between the two pain generators indicates that more research is needed on this topic.
Aside from osteoarthritis, several other conditions may affect the facet joints. These include inflammatory arthritides such as rheumatoid arthritis, ankylosing spondylitis and reactive arthritis,99–101synovial impingement, meniscoid entrapment, chondromalacia facetae, pseudogout, synovial inflammation, villonodular synovitis, and acute and chronic infection.102–106Intrafacetal synovial cysts can be a source of pain because of distension and pressure on adjacent pain-generating structures, calcification, and asymmetrical facet hypertrophy.107–110In a retrospective review of MRI scans in 303 consecutive patients with LBP, Doyle and Merrilees111found that 9.5% had facet joint synovial cysts, the large majority of which were located posteriorly. Trauma may also cause lumbar facetogenic pain. There are more than two dozen reported cases of lumbar facet dislocation after rapid deceleration injuries (e.g. , traffic accidents), most involving L5–S1.61,112–115The mechanism of injury in these cases is purported to be a combination of hyperflexion, distraction, and rotation.61,112,116In a posthumous study conducted in 31 lumbar spines of subjects who died of traumatic injuries (mostly motor vehicle accidents), Twomey et al. 117found occult bony fractures in the superior articular process or subchondral bone plate in 35% of victims, and z-joint capsular and/or articular-cartilage damage in 77% of cases. The authors concluded that occult bony and soft tissue injuries to the l-z joints may be a common cause of LBP after trauma.
Prevalence
The prevalence rate of l-z joint pain varies widely in the literature, ranging from less than 5% to upward of 90%.118–125To a large extent, the wide discrepancy in prevalence rates is a function of the diagnostic methodology used and the perspective and conviction of the investigator. Numerous reviews have outlined the inherent flaws in diagnosing l-z joint pain using historic, physical examination, and radiologic findings and concluded that an analgesic response to image-guided intraarticular or MBB is the only reliable and valid method to identify a facet joint(s) as the primary pain generator.126–128Furthermore, the false-positive rate of uncontrolled facet blocks has been found to range between 25% and 41% using comparative LA injections or saline controls,129–132leading some experts to conclude that the use of controlled blocks is the only reliable means to diagnose lumbar facetogenic pain.133
Using single LA blocks, the prevalence of l-z joint pain has been reported to range from 8% to 94%.89,119When placebo-controlled and comparative LA facet blocks are used, the reported prevalence rates decline significantly, ranging from 9% to 42%.98,131The estimated prevalence rates increase in conjunction with age in the populations studied. In a comprehensive epidemiologic study on LBP, spine surgeons from eight academic medical centers in the United States collected demographic and clinical information on more than 4,000 patients during a 5-yr period.118Final diagnoses were rendered based on historic and physical examination findings, radiologic and other diagnostic studies, and response to treatment and/or diagnostic injections. Among the 2,374 patients who remained in the study, “facet joint arthritis” was the final diagnosis in 4.8% of cases. In an epidemiologic study conducted in a primary care setting, l-z joint pain was estimated to account for approximately 6% of patients with chronic LBP.134Based on studies using comparative or controlled blocks, in descending order, the L5–S1, L4–L5, and L3–L4 facet joints are most frequently implicated in l-z joint pain.98,130,132
One problem that emerges when synthesizing data from published prevalence studies is that almost all excluded patients with neurologic signs or symptoms secondary to a herniated disc, the most common cause of chronic LBP, and many excluded patients with previous back surgery. In patients with facet hypertrophy, foraminal narrowing can actually cause radicular symptoms.72–74,135A second confounding factor is that the best prevalence studies used comparative MBB to estimate l-z joint pain. The primary dorsal rami divide into three nerves as they approach their respective transverse processes, the largest of which is the medial branch. In addition to supplying two facet joints, the medial branch also innervates the multifidus, interspinales muscle and ligament, and the periosteum of the neural arch. The two other main branches of the dorsal ramus are the intermediate branch, which sends fibers into the longissimus muscle, and the lateral branch, which innervates the iliocostalis muscle, the thoracolumbar fascia, the skin of the lower back and buttock, and the sacroiliac joint.29,136,137At the superomedial border of the transverse processes where the lumbar medial branches are most amenable to blockade, these nerves are in such close proximity that anesthetizing any one of them will almost invariably lead to blockade of the others branches of the primary dorsal rami. Therefore, MBB may not only block nociceptive signals arising from the l-z joints, but also from several other potential pain-generating structures around the lumbar spine. This makes the true prevalence rate of l-z joint pain exceedingly difficult to estimate. Based on the evidence that does exist, the lumbar facet joints seem to be the primary pain generator in approximately 10–15% of chronic LBP patients (table 2).
Pain Referral Patterns
In an attempt to better understand l-z joint pain, many investigators have attempted to identify pain referral patterns using a variety of different methods. These include pain provocation via stimulation of the facet joint capsules and medial branches in pain patients and asymptomatic volunteers, and mapping out pain diagrams in subjects in whom pain was relieved by the injection of LA, with or without steroid. For the most part, these studies have not demonstrated any reliable pain referral pattern stemming from any of the l-z joints. Discrepancies between pain provocation and pain patterns/histologic findings have been found not only for l-z joint stimulation,138,139but also during sacroiliac joint and selective nerve root blocks.140,141Part of the problem with using provocative tests to delineate pain referral patterns is that artificial stimulation of the l-z joints and/or their nerve supply may not simulate physiologic conditions.
Nevertheless, certain patterns do emerge when synthesizing the existing data. To summarize these findings, the joint capsule seems to be more likely to generate pain than the synovium or articular cartilage. There is also considerable overlap between all lumbar facet joints, with the referral pattern being more widespread and variable in patients with chronic pain than in asymptomatic volunteers. All of the lumbar facet joints are capable of producing pain that can be referred into the groin, although this is more common with lower facet joint pathology. Pain emanating from upper facet joints tends to extend into the flank, hip, and upper lateral thigh, whereas pain from the lower facet joints is likely to penetrate deeper into the thigh, usually laterally and/or posteriorly. Infrequently, the L4–L5 and L5–S1 facet joints can provoke pain extending into the lower lateral leg and, in rare instances, even the foot. In patients with osteophytes, synovial cysts, or facet hypertrophy, the presence of radicular symptoms may also accompany sclerotomal referral patterns (table 3and fig. 4).
Fig. 4. Pain referral patterns from the lumbar facet joints. In descending order, the most common referral patterns extend from the darkest (low back) to the lightest regions (flank and foot). The key at the bottom of the figure legend is listed in order of affected frequency ( i.e. , low back to foot). The facet levels next to each location represent the zygapophysial joints associated with pain in each region. Data adapted from McCall et al. ,10,Marks, 269,and Fukui et al. 270 Drawings by Specialist Frank and Angela Dill, US Army, and Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland). Low back: L5–S1, L4–L5, L3–L4 Buttock: L5–S1, L4–L5, L3–L4 Lateral thigh: L5–S1, L4–L5, L3–L4, L2–L3 Posterior thigh: L5–S1, L4–L5, L3–L4 Greater trochanter: L5–S1, L4–L5, L3–L4, L2–L3 Groin: L5–S1, L4–L5, L3–L4, L2–L3, L1–L2 Anterior thigh: L5–S1, L4–L5, L3–L4 Lateral lower leg: L5–S1, L4–L5, L3–L4 Upper back: L3–L4, L2–L3, L1–L2 Flank: L1–L2, L2–L3 Foot: L5–S1, L4–L5
Fig. 4. Pain referral patterns from the lumbar facet joints. In descending order, the most common referral patterns extend from the darkest (low back) to the lightest regions (flank and foot). The key at the bottom of the figure legend is listed in order of affected frequency ( i.e. , low back to foot). The facet levels next to each location represent the zygapophysial joints associated with pain in each region. Data adapted from McCall et al. ,10,Marks, 269,and Fukui et al. 270 Drawings by Specialist Frank and Angela Dill, US Army, and Frank M. Cort, M.S. (Research Associate, Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland). Low back: L5–S1, L4–L5, L3–L4 Buttock: L5–S1, L4–L5, L3–L4 Lateral thigh: L5–S1, L4–L5, L3–L4, L2–L3 Posterior thigh: L5–S1, L4–L5, L3–L4 Greater trochanter: L5–S1, L4–L5, L3–L4, L2–L3 Groin: L5–S1, L4–L5, L3–L4, L2–L3, L1–L2 Anterior thigh: L5–S1, L4–L5, L3–L4 Lateral lower leg: L5–S1, L4–L5, L3–L4 Upper back: L3–L4, L2–L3, L1–L2 Flank: L1–L2, L2–L3 Foot: L5–S1, L4–L5
Diagnosis
History and Physical Examination
Numerous studies have attempted to delineate a discrete set of historic and physical findings pathognomonic or at least suggestive of lumbar facet arthropathy. Fairbank et al. 142conducted a prospective study in 41 patients with acute LBP whereby two-level intraarticular facet blocks were performed with low-volume bupivacaine. Among the 25 patients who completed the study, 14 obtained at least temporary relief after the injections. Compared with nonresponders, responders tended to have pain localized to the back and thigh, and to report pain during forward flexion.
In 1988, Helbig and Lee143designated a “lumbar facet syndrome” based on a retrospective study conducted in 22 patients. The authors found that patients who responded to intraarticular facet injections (injection parameters not noted) were more likely to have back pain associated with groin or thigh pain, paraspinal tenderness, and reproduction of pain during extension–rotation maneuvers. Pain radiating below the knee was negatively associated with a positive response to facet blocks.
Despite the widespread acceptance of the “lumbar facet syndrome,” a multitude of larger and more methodologically sound studies have failed to duplicate the findings of Helbig and Lee. In a large study conducted in 390 patients with chronic LBP, Jackson et al. 89were unable to identify any historic or physical examination variables associated with analgesic response to facet injections. Schwarzer et al. 12conducted a prospective study attempting to identify clinical features in 176 patients with chronic LBP undergoing double, confirmatory blocks. In the 15% of patients who achieved concordant pain relief with lidocaine and bupivacaine, no clinical feature was associated with a positive response. In a randomized, placebo-controlled study performed on 80 patients with chronic LBP, Revel et al. 88identified seven variables associated with a positive response to facet joint anesthesia: age greater than 65 yr and pain not exacerbated by coughing, not worsened by hyperextension, not worsened by forward flexion, not worsened when rising from forward flexion, not worsened by extension–rotation, and well-relieved by recumbency. However, subsequent investigations have also failed to corroborate the findings of Revel et al. (table 4). In summary, no historic or physical examination findings can reliably predict response to diagnostic facet joint blocks.
Radiologic Findings
The prevalence of abnormal l-z joint changes on radiologic imaging depends on the age and presence of symptoms in the study population, the imaging modality used, and the threshold use for rendering a diagnosis of “abnormal” (table 5). In studies conducted in patients with LBP, the incidence of degenerative facet disease on computed tomographic (CT) scanning ranges from around 40% in some studies125,144to upwards of 85% in others.119MRI is considered to be somewhat less sensitive than CT imaging for detecting degenerative facet changes,119,145,146although several studies conducted in chronic LBP patients found both the sensitivity and specificity of MRI to be more than 90% compared with CT.97,145In a study by Fujiwara et al. 97conducted in 14 subjects with DDD, the extent of lumbar facet osteoarthritic changes was minimal in patients younger than 40 yr. In patients older than 60 yr, the prevalence of degenerative changes increased significantly but was by no means universal. No facet osteoarthritis was observed in any patient in the absence of disc degeneration. In CT and MRI studies conducted in asymptomatic volunteers, the prevalence of facet degeneration ranges from 8% to 14%.147–149In one study by Weishaupt et al. 147conducted in 60 asymptomatic volunteers aged 20–50 yr, disc bulging or protrusion was found in 37 and 40 subjects, but severe osteoarthritis of the facet joints was absent in all 60 subjects. The authors suggested that severe osteoarthritis of the l-z joints may play a prominent role in LBP because of its absence in asymptomatic individuals.
The ability of radiologic imaging to predict response to diagnostic l-z joint blocks has been conflicting at best. Whereas some studies have found a positive correlation between CT, MRI, or other imaging studies and response to l-z joint blocks,124,125,143,144,150an equal number have not.87,89,119,142,151,152Results from the three largest studies have also been mixed. The largest study, by Jackson et al. ,89found no relation between radiographic evidence of l-z joint degeneration and response to single, intraarticular facet injections in 390 patients. This is in contrast to Carrera and Williams,144who found that 73% of chronic LBP patients (n = 63) demonstrating CT evidence of lumbar facet disease experienced pain relief after large-volume (2–4 ml) facet blocks versus only 13% in whom CT scans showed no pathology. In the only study using placebo-controlled blocks to confirm a diagnosis of l-z joint pain, Schwarzer et al. 152found no correlation between CT findings and a positive response to LA but not saline blocks in 63 patients. The results of Jackson et al. and Schwarzer et al. support clinical studies showing no correlation between MRI findings and results of medial branch radiofrequency denervation. In a study by Cohen et al. 153conducted in 192 patients who underwent radiofrequency denervation based on a positive response to single MBB, the authors found no association between MRI evidence of facet hypertrophy or degeneration and 6-month outcomes. Finally, Kawaguchi et al. 154found no significant association between the degree of radiographic lumbar facet joint abnormalities and LBP symptoms in a study conducted in 106 patients with rheumatoid arthritis. In summary, the evidence in the literature does not support the routine use of radiologic imaging to diagnose l-z joint pain.
Diagnostic Blocks
It is generally accepted in clinical practice that diagnostic blocks are the most reliable means for diagnosing l-z joints as pain generators. Numerous guidelines and reviews have asserted that intraarticular injections and MBB are equally effective in diagnosing l-z joint pain.105,126,127,133Although this statement may seem to have face validity, several factors may undermine the utility of diagnostic blocks, especially MBB. In a cadaveric study, Kellegren155showed that 0.5 ml injectate spread into an area encompassing 6 cm2of tissue. In view of the close proximity of the medial branch nerves to the lateral and intermediate branches, even the injection of a low-volume of anesthetic is likely to block these nerves. Because these nerves, along with the medial branches themselves, contribute heavily to the innervation of the paraspinal muscles and fascia, ligaments, sacroiliac joints, and skin, MBB can relieve LBP even in the presence of normal l-z joints.
Whereas properly performed intraarticular facet injections may be inherently more accurate in diagnosing l-z joint pain, these blocks can be technically challenging and fraught with their own limitations. After injecting somewhere between 1 and 2 ml of fluid, the joint capsule is likely to rupture, with the excess injectate extravasating into several possible pain-generating structures. Depending on the point of rupture, these structures may include the epidural space, intervertebral foramen, ligamentum flavum, and paraspinal musculature.8,71,122,123
There are no crossover studies comparing the validity of MBB to intraarticular l-z joint injections, and only two studies comparing them at all. Nash156conducted a prospective study in 67 patients with axial LBP who were randomly assigned in pairs to receive either MBB with 2 ml LA or intraarticular injections with 1.5 ml LA and steroid. In the 26 pairs who completed the study, 12 reported MBB to be more beneficial at their 1-month follow-up, 11 reported the intraarticular injection to be better, and 3 reported no difference between the two. In the second study, Marks et al. 8randomly assigned 86 axial LBP patients to receive either intraarticular injections or MBB using 2 ml LA and steroid. The authors found no difference in the immediate response between the two groups, although the intraarticular group experienced better pain relief at their 1-month but not 3-month follow-up. There are numerous flaws with these studies that limit the conclusions one can draw on the comparative validity of these two procedures, with the main one being the lack of a definitive diagnosis in the study subjects. Based on prevalence and false-positive rates in chronic LBP patients (table 2), MBB and intraarticular seem to provide comparable diagnostic value. In the four placebo-controlled studies evaluating radiofrequency denervation outcomes in patients with “confirmed” l-z joint pain, the only study that screened patients with MBB demonstrated positive outcomes,157,versus only one of three that used diagnostic intraarticular injections.158–160Ultrasound-guided medial branch and intraarticular blocks have also been demonstrated to provide comparable accuracy to fluoroscopically guided injections, although they may be less likely to detect low-volume intravascular uptake and are less accurate in obese patients.161–164Because MBB are technically easier to perform than intraarticular injections and involve anesthetization of the nerves to be lesioned, it seems more logical to use these blocks as a prognostic tool before radiofrequency denervation.
False-positive Blocks
Numerous studies have documented a high false-positive rate for lumbar facet blocks, ranging from 25% to 40% using comparative blocks or saline controls (table 2).128–130This rate seems to be unaffected by the type of block used (i.e. , intraarticular or MBB). In a study evaluating the utility of comparative LA MBB versus placebo-controlled MBB to diagnose cervical z-joint, Lord et al. 165found that the use of serial blocks with two different LA had a high degree of specificity (88%) but only marginal sensitivity (54%). In addition to the anatomical differences between cervical and lumbar facet joints, this study used the stringent criterion of “complete or profound pain relief” as the benchmark for a positive block, leaving the question of extrapolation subject to debate. However, one interpretation of these findings is that comparative, LA blocks predispose patients to false-negative diagnoses.
The reasons for false-positive facet blocks are multifactorial and include placebo response (18–32%) to diagnostic facet interventions, use of sedation, the liberal use of superficial LA, and the spread of injectate to pain-generating structures other than those targeted.166Although some investigators have disputed this assertion,167,168it is our belief that not only opioids, but sedatives such as midazolam, can lead to false-positive blocks by interfering with the interpretation of analgesic response (i.e. , preventing a patient from engaging in normal activities) and virtue of their muscle relaxant properties.169In a recent survey conducted in 500 patients receiving facet blocks or epidurals at an outpatient spine center, only 17% requested sedation.170
Even in patients with symptomatology concordant with unambiguous pathology, diagnostic blocks may lack specificity. North et al. 171conducted a prospective study in 33 patients with L5 or S1 radiculopathy and radiologic evidence of ongoing nerve root compression. All patients underwent a battery of LA blocks that included selective nerve root block, sciatic nerve block, MBB, and subcutaneous control injections. The authors found that approximately 90% of patients obtained almost complete pain relief after the selective nerve root block, 70% obtained almost complete relief after the sciatic block, and a majority received at least 50% pain relief after the MBB. In contrast, the median degree of pain relief after the subcutaneous injection was around 30%. The authors concluded that uncontrolled LA blocks lack specificity in the diagnostic evaluation of referred pain syndromes.
Dreyfuss et al. 31attempted to measure the specificity of lumbar MBB by performing 120 fluoroscopically guided injections in healthy volunteers using 0.5 ml injectate and identifying the contrast spread using CT scan. Two target points were chosen, one at the superomedial border of the transverse process and a second lower site midway between the upper border of the transverse process and the mamilloaccessory ligament. Aberrant contrast was noted to spread into the intervertebral foramen or epidural space 16% of the time, being more common at cephalad spinal levels. When the lower target points were used, spread into adjacent neural structures only occurred when a needle was inadvertently placed too high. In all cases, distal spread was noted into the cleavage plane between the multifidus and longissimus muscles. In no instance did the 0.5 ml of contrast fail to bathe the target. The investigators concluded that lower volumes may be adequate for MBB and that using the lower target point may increase the specificity of lumbar MBB.
Following up on the Dreyfuss study, Cohen et al. 135sought to determine whether spread into the epidural space or intervertebral foramina could account for false-positive MBB by examining the relation between clinical signs of radiculopathy, discographic findings, and radiofrequency outcomes in 78 patients with positive MBB who went on to fail radiofrequency denervation. The authors found a negative correlation between discogenic pain and failed radiofrequency denervation and no association between radicular pain and radiofrequency treatment outcomes. In contrast, there was a trend toward patients with failed back surgery syndrome to have a negative outcome after radiofrequency lesioning. The authors concluded that myofascial pain might be a significant cause of false-positive MBB.
The evidence that the inadvertent treatment of myofascial pain may be a significant cause of false-positive MBB is circumstantial, but multifaceted. In their large, multicenter epidemiologic study involving more than 2,000 patients, Long et al. 118found myofascial pain to be the second most common cause of chronic LBP after herniated disc. Controlled studies conducted in chronic LBP patients have shown efficacy for both muscle relaxants and low-volume botulinum toxin injections, and electromyographic evidence of increased activity compared with matched controls.172–175Finally, Ackerman et al. 176tested the hypothesis as to whether myofascial pain could account for the high rate of false-positive facet blocks in a double-blind study conducted in 75 men with chronic LBP. Subjects received either intraarticular facet injections or MBB using two techniques: one in which LA was used to provide superficial anesthesia down to the target point, and a second in which saline was injected as the needle was advanced. The authors found that the incidence of postprocedure pain relief was significantly higher in patients who had LA injected into their musculature than in those who received saline injected superficially. The injection of LA into the skin and soft tissues may also reduce LBP by means other than the inadvertent treatment of myofascial pain. In studies by Woolf et al. ,177,178the authors found that the superficial injection of even very small amounts of lidocaine reduced nociceptive behavior in animal models of neuropathic pain, a finding attributed to the systemic absorption of the sodium channel blocker (table 6).
To reduce the amount of superficial anesthesia used for MBB, Stojanovic et al. 179introduced the single-needle technique whereby multiple medial branches are blocked using a single skin entry point. In a prospective, crossover study comparing the single-needle and conventional multiple-needle techniques, the authors found the single-needle technique required significantly less superficial LA, resulted in less procedure-related pain, and was quicker to perform than the multiple-needle approach.180With regard to final needle position, contrast spread, and postprocedure pain relief, no differences were noted between the two techniques.
False-negative Blocks
False-negative blocks may also occur during diagnostic facet injections. In a study by Kaplan et al. 32conducted in 18 asymptomatic volunteers, the authors found that properly performed MBB result in failure to anesthetize the corresponding facet joint 11% of the time, even with the avoidance of venous uptake. Although this may have occurred because the LA did not engulf the target nerve, a more likely cause for the false-negative rate is the presence of aberrant or additional innervation to facet joints aside from medial branches. In the same study, the authors found inadvertent venous uptake occurred during 33% of nerve blocks. When the needle was repositioned to avoid venous uptake, analgesia was achieved only 50% of the time. The authors concluded that when venous uptake occurs, it may be advantageous to repeat the procedure on a separate occasion rather than redirecting the needle to avoid false-negative results. In a CT study by Dreyfuss et al. ,31the authors found inadvertent venous uptake occurred only 8% of the time.
Should Single or Double Diagnostic Blocks Be Used?
The high rate of false-positive facet blocks has led numerous experts to advocate performing double blocks, using either saline controls or two different LA, before proceeding to definitive therapy.12,128,130–133,181However, this is rarely done in clinical practice or controlled outcome studies, and there are currently no studies comparing outcomes using single and double blocks. In the only uncontrolled study assessing outcomes for medial branch radiofrequency denervation after comparative LA blocks, Dreyfuss et al. 182found that 60% of the 15 patients who proceeded to radiofrequency lesioning achieved at least 90% pain relief at 12 months, and 87% achieved at least 60% relief. The basis for undergoing treatment was at least 80% concordant pain relief after lidocaine and bupivacaine MBB. However, 460 patients were interviewed for this study, and after history, physical examination, and record review, only 41 patients were chosen to participate. Had single blocks or less stringent criteria been used to designate radiofrequency candidates, the success rate would have indubitably been lower in this study, but the total number of successful treatments would have almost certainly been higher.
When considering whether to perform double blocks, several factors must be considered, including the patient’s relative risk for a false-positive block, the complication rate of the diagnostic and definitive procedure (i.e. , radiofrequency denervation), the false-negative rate of diagnostic injections, and the cost effectiveness, including dropout rate, of performing an additional diagnostic injection. With the exception of a very low (< 5%) incidence of neuritis183that can be reduced even further with the preemptive use of steroid or pentoxifylline,184the complication rate is similar for diagnostic facet blocks and radiofrequency denervation. In a systematic study of the cost-effectiveness of using controlled facet blocks, Bogduk and Holmes185determined that the use of placebo-controlled injections cannot be justified in the United States based on financial considerations. Considering that a substantial percentage of patients will respond with long-term pain relief even to sham denervation,159,160it is unlikely that the routine use of confirmatory facet blocks will become standard of care anytime soon.
Treatment
Conservative Treatment and Pharmacotherapy
The treatment of l-z joint pain ideally consists of a multimodal approach comprising conservative therapy, medical management, procedural interventions, and if indicated, psychotherapy. There are no clinical studies specifically assessing pharmacotherapy or noninterventional treatment for lumbar facet arthropathy, although there are several controlled studies evaluating conservative treatment for axial LBP. Tailored exercise programs and yoga have been shown to reduce pain and prevent relapses in patients with chronic LBP.186–190In randomized, clinical trials, osteopathic manipulation has been shown to provide moderate relief for LBP patients,191,192although one study showed no difference between the benefits afforded by true and sham manipulation.193Acupuncture has also been shown in randomized trials to provide significant benefit in patients with chronic LBP.194–196However, similar to manipulation, one of the largest and most methodologically sound studies found no difference between true and sham acupuncture.197
Nonsteroidal antiinflammatory drugs and acetaminophen are widely considered first line drugs for the treatment of LBP, with little evidence to support one particular drug over another.198–200Adjuvants have also been shown to be effective in relieving LBP. In a comprehensive review on published clinical trials evaluating pharmacotherapy in LBP, Schnitzer et al. 201found strong evidence to support the use of antidepressants for chronic LBP, and muscle relaxants in acute back pain. Numerous studies have also provided incontrovertible evidence that untreated psychopathology can adversely affect LBP treatment outcomes. In a study by Polatin et al. 202conducted in 200 chronic LBP patients, the authors found that 77% met lifetime criteria and 59% demonstrated current symptoms for at least one psychiatric diagnosis, with the most common being depression, substance abuse, and anxiety disorders. Most, but not all studies, have shown untreated psychopathology to negatively impact LBP treatment outcomes.203In the only study assessing the influence of psychopathology on l-z joint interventions, Lilius et al. 204found a strong correlation between a negative response to intraarticular and periarticular LA and steroid injections and inappropriate signs and symptoms. We believe that the optimal management of l-z joint pain should encompass both interventional and noninterventional treatment, although clinicians are encouraged to exercise caution when extrapolating the results of studies conducted in patients with nonspecific LBP to those with clear-cut facet pathology.
Intraarticular Steroid Injections
The use of intraarticular steroid injections to treat l-z joint pain is a controversial subject. In uncontrolled studies, the long-term relief of back pain after intraarticular steroid injection ranges from 18% to 63%, with most of these studies being conducted in patients who did not undergo previous diagnostic l-z blocks.118–120,123,124,144,205Studies have also reported intermediate-term pain relief after intraarticular LA alone,142normal saline,206,207and hyaluronic acid.208
In the controlled trials that do exist, the results are mixed (table 7). In the largest study, Lilius207reported no significant difference in outcomes between 109 patients who received large-volume (8 ml) LA and steroid injected into l-z joints or around l-z joints or intraarticular saline injections. In a randomized, controlled study comparing intraarticular steroid and saline in 97 patients with chronic LBP, Carette et al. found a statistically significant benefit favoring steroid only at 6 months after the procedure.206This study is the second largest and most methodologically sound study evaluating intraarticular steroids, and the only one that prescreened study patients based on diagnostic facet blocks. One flaw in these studies is that most used saline injections as the control group, and normal saline has been shown to provide better pain relief than that expected with a true placebo for a multitude of invasive procedures.209–211In five recent review articles, the authors were split as to whether intraarticular steroids constituted an effective treatment for l-z joint pain, with three concluding they did not.128,181,212–214Based on the existing evidence, including basic science studies demonstrating inflammatory mediators to be present in and around degenerated facet joints,43,44we believe that intraarticular steroid injections may provide intermediate-term relief to a small subset of patients with l-z joint pain accompanied by an active inflammatory process. Evidence to support this assertion is bolstered by several recent prospective and observational studies evaluating low- to intermediate-volume (1–3 ml) LA and steroid intraarticular l-z joint injections performed in more than 160 patients with axial LBP.150,215,216In these studies, patients with positive single photon emission computed tomography experienced dramatically better pain relief (> 75% success rate) compared with those with negative or no single photon emission computed tomography (< 40% success rate) up to 3 months after injection. In the two studies that followed patients for 6 months after injection, the beneficial effect wore off after the 3-month evaluation.150,215Radionuclide bone scintigraphy is capable of depicting synovial changes caused by inflammation, degenerative changes associated with bone remodeling, and increased metabolic function. In addition to radiologic evidence of joint inflammation and degeneration, intraarticular steroid injections may be more effective in those patients who obtain definitive pain relief after a diagnostic screening block, and when LA is added to the injectate.
Radiofrequency Denervation
In 1971, Rees18first described percutaneous denervation as an effective new treatment for discogenic back pain. Notwithstanding his greater than 99% reported success rate, it remains a subject of controversy as to whether his technique actually achieved “facet rhizolysis,” since the instrument he used may not have been long enough to accomplish anything more than a myofasciotomy.3The technique as practiced today, which entails using radiofrequency energy channeled through a small-diameter needle to create a controlled burn that severs the l-z joint nerve supply, is generally credited to Shealy, who was motivated by what he perceived to be an unacceptably high incidence of local hemorrhagic complications.19,20Subsequently, it has been used with varying degrees of success to treat different forms of spinal pain, including whiplash,217sacroiliac joint pain,218,219discogenic pain,220and intractable sciatica.221There are literally dozens of uncontrolled trials touting the benefits of radiofrequency denervation for l-z joint pain,222with most reporting sustained relief in 50–80% of subjects without previous back surgery19,223–228and 35–50% of patients with failed back surgery syndrome.153,229–232
Only five placebo-controlled studies have been conducted evaluating radiofrequency denervation for l-z joint pain (table 8). In the first study, King and Lagger3randomly assigned 60 patients with low back and leg pain to receive empirical (without stimulation) radiofrequency denervation of the dorsal rami, a radiofrequency lesion made in the muscle, or a sham lesion after electrical stimulation. At their 6-month follow-up, 27% of patients in the facet denervation group experienced satisfactory pain relief versus 53% in the myotomy group and 0% in the sham group. The main criticism in this study is that no diagnostic blocks were performed to screen people for l-z joint pain. More than 15 yr later, Gallagher et al. 158randomly assigned 41 patients based on their response to diagnostic intraarticular blocks (equivocal or good response) to either sham or true denervation. A statistically significant difference in outcomes was observed at 1 month only between sham and true radiofrequency denervation in those patients who obtained a definitive response to diagnostic blocks. This difference persisted for the duration of the 6-month follow-up. In the smallest but most methodologically sound233study among the five controlled trials, van Kleef et al. 157found a 46% pain reduction in the radiofrequency lesion group versus an 8% reduction in the placebo group. At 12-month follow-up, 7 of 15 patients in the radiofrequency group continued to have a successful outcome versus only 2 of 16 in the sham group. Leclaire et al. 159conducted a placebo-controlled study in 70 patients with a putative diagnosis of facet arthropathy. At their 4-week follow-up, the only outcome variable that favored the treatment group was an improvement in mean Roland-Morris disability score. At 12 weeks, no difference was noted between groups for pain levels or any measure of functional capacity. The key flaw in this study is that the authors used “significant pain relief lasting > 24 h” after an intraarticular injection of LA and steroid as their main inclusion criterion. In addition to being ambiguous, the 24-h threshold is inconsistent with the pharmacodynamics of lidocaine. In the largest controlled study evaluating radiofrequency denervation, van Wijk et al. 160found the only difference between the treatment and control group at 3 months was that more radiofrequency patients reported a 50% or greater diminution in back pain than sham patients (62% vs. 39%). For mean reduction in VAS pain scores, change in analgesic intake, and functional assessments, no differences were noted between groups.
Table 8. Outcomes for Randomized, Controlled Studies Assessing Medial Branch Radiofrequency Denervation for Facet Joint Pain

The chief criticisms of these studies (see table 8for more details) include suboptimal needle positioning and not selecting study patients based on placebo-controlled or comparative LA blocks. In the only prospective study to evaluate radiofrequency outcomes in patients selected based on concordant response to serial LA blocks, Dreyfuss et al. 182reported that 87% of 15 patients obtained at least 60% pain relief 12 months status post radiofrequency denervation, with 60% of patients achieving at least 90% relief. In addition to stringent inclusion criteria, the authors used 16-gauge electrodes and assessed the efficacy of radiofrequency denervation by performing electromyography of the multifidus muscles.
Whereas some may construe these findings as evidence that radiofrequency denervation is a fundamentally flawed treatment, a more plausible interpretation is that they indicate a strong need to optimize radiofrequency denervation techniques and better identify those candidates who are likely to obtain positive outcomes. Several investigators have determined that placing the electrode parallel rather than perpendicular to the target nerve substantially increases the size of the lesion, thereby reducing the likelihood the treatment will miss or only partially coagulate the target nerve.234,235After a literature review and cadaveric study, Lau et al. 234concluded the ideal electrode position is across the lateral neck of the superior articular process rather than the groove formed at the angle of the superior articular and transverse processes, as was used in most studies.159,160Other investigators have found the maximal lesion size to be reached within 60 s of lesion time,235–237independent of whether the system is temperature or voltage-controlled.238Studies conducted in human myocardium have determined that irrigation fluid has either no effect or a slightly beneficial effect on lesion size.239Hence, the use of LA to prevent procedure-related pain or steroid to reduce the incidence of neuritis184should theoretically have no adverse effects on the efficacy of radiofrequency denervation.
Another flaw that pervades most radiofrequency studies is that sensory stimulation (usually at ≤ 0.5 V) is used to corroborate proximity of the electrode to the targeted medial branch. Whereas sensory stimulation is almost certain to be perceived when the electrode is placed on or adjacent to a neural structure, it is our experience that many patients perceive concordant sensory stimulation at 0.5 V or less, even when the electrode is purposefully placed in muscle, as during sham procedures. An attractive alternative to sensory stimulation is to instead (or in addition) attempt to elicit multifidus muscle contraction, because the same medial branch that innervates the facet joint also innervates this paraspinal muscle. In the two studies in which the medial branch was identified by motor stimulation of the multifidus muscle, both reported positive outcomes.157,182
In a large, multicenter outcome study, Cohen et al. 153attempted to identify factors associated with successful radiofrequency treatment in 192 patients who underwent denervation at three teaching hospitals after a single, positive MBB. Among the 15 variables analyzed for their association with treatment outcome, only paraspinal tenderness was found to predict a successful treatment. Factors associated with failed treatment included increased pain with hyperextension and axial rotation (i.e. , facet loading), duration of pain and previous back surgery. The latter two variables have been associated with treatment failure not only for radiofrequency denervation, but a host of other LBP interventions as well, including epidural steroid injections and open surgery.240–242When pain returns after radiofrequency denervation, which typically occurs between 6 months and 1 yr, repeated neurotomy can be performed with no diminution in efficacy.243In addition to continuous, high-temperature radiofrequency medial branch ablation, pulsed radiofrequency (2–6 months of effective pain relief),244cryodenervation (3–6 months of pain relief),245–248and phenol neurolysis249,250have also been reported to provide intermediate to long-term pain relief in uncontrolled studies.
Complications after Intraarticular Injections and Radiofrequency Denervation
Serious complications and side effects are extremely uncommon after facet interventions. The metabolic and endocrine sequelae of intrafacetal depot steroids have not been studied, but extrapolating from epidural steroid injections, one would expect suppression of the hypothalamic–pituitary–adrenal axis lasting up to 4 weeks depending on the depot steroid used, and impaired insulin sensitivity manifesting as elevated glucose levels for less than a week.251,252Although rare, a host of infections have been reported after intraarticular injections including septic arthritis, epidural abscess, and meningitis.253–255Case reports of spinal anesthesia and postdural puncture headache have also been published.256,257
Numbness and/or dysesthesias have been reported after radiofrequency denervation but tend to be transient and self-limiting.219,258Burns are rare with radiofrequency procedures and may result from electrical faults, insulation breaks in the electrodes, and generator malfunction.19,226,259The most common complication after facet joint radiofrequency is neuritis, with a reported incidence of less than 5%.183In one study, the administration of corticosteroid or pentoxifylline was found to reduce the incidence of postprocedure pain after radiofrequency denervation.184There is also a theoretical risk of thermal injury to the ventral rami if an electrode slips ventrally over the transverse process.
Surgery
Surgery is occasionally performed to treat facet arthropathy despite a lack of evidence supporting fusion for degenerative spinal disorders.260,261Not surprisingly, the results of studies evaluating the use of l-z joint blocks to predict lumbar arthrodesis outcomes are discouraging (table 9). In the three studies that compared surgical outcomes between facet block responders and nonresponders, all three failed to show a difference between groups.262–264Bough et al. 138conducted a retrospective review of 127 facet joints surgically removed from 84 patients in an attempt to correlate histologic evidence of facet degeneration with provocative response to preoperative facet arthrography. Although the authors found the positive predictive value of concordant pain reproduction to be 85%, the negative predictive value was only 43%, leading them to conclude that provocative facet arthrography was of little value as a presurgical screening tool. In a prospective case series, Lovely and Rastogi265found that 83% of 23 patients who responded to bracing and three successive facet blocks achieved at least 90% pain relief after fusion surgery at the latest follow-up. However, the large volumes used per block, the failure to exclude placebo-responders, and the lack of any comparison group undermine the conclusions that can be drawn. One reason patients with l-z joint pain might respond to arthrodesis is because some surgeons, either purposefully or inadvertently, perform medial branch rhizotomies during pedicle screw placement. In summary, there is no convincing evidence to support any surgical intervention for l-z joint pain aside from that resulting from a traumatic dislocation.
Conclusions
Pain originating from the l-z joints has long been recognized as a potential source of LBP. Anatomical studies suggest that with aging, the facet joints become weaker and their orientation changes from coronal to sagittal positioning, predisposing them to injury from rotational stress. The three most caudal facet joints, L3–L4, L4–L5, and L5–S1, are exposed to the greatest strain during lateral bending and forward flexion and are thus more prone to repetitive strain, inflammation, joint hypertrophy, and osteophyte formation. Osteoarthritis of the facet joints is commonly found in association with degenerative disc disease. The exact prevalence of facet disease resulting in axial LBP is unclear but may be as high as 10–15% of patients.
There are no discrete historic and physical findings pathognomonic for lumbar facet arthropathy. The referral patterns for pain arising from the lumbar facet joints at different levels overlap considerably. In addition to axial LBP, pathology arising from the lower facet joints is associated with referred pain to the buttock, thigh, groin, and sometimes lower leg, whereas that referred from the upper lumbar facet joints extends into the flank, hip, groin, and lateral thigh. Reports on the correlation between CT and MRI evidence of facet arthropathy and the response to diagnostic lumbar facet blocks are conflicting. Because the facet joint is innervated by the medial branches arising from the posterior rami of the spinal nerve at the same level and a level above the joint, LA blocks of these nerves have been advocated for diagnostic and prognostic purposes. Intraarticular l-z joint injection with LA has also been proposed as a method for diagnosing facet joint pain, with both procedures appearing to provide comparable diagnostic value. As with other blocks, the potential for false positive and false negative responses must be considered, and steps should taken to reduce their incidence.
In addition to providing short and occasionally intermediate-term pain relief, diagnostic blocks are considered predictive of the potential usefulness of subsequent neurolytic procedures such as radiofrequency denervation. In carefully selected patients who fail conservative treatments such as physical and pharmacologic therapies, intraarticular steroid injections and radiofrequency denervation are treatment options. Studies evaluating the long-term outcomes from these procedures have thus far provided conflicting evidence. The results of surgical therapies including arthrodesis for facet arthropathy are discouraging.
The authors thank Paul Dreyfuss, M.D. (Professor of Physical Medicine and Rehabilitation, University of Washington School of Medicine, Seattle, Washington), for his review of the manuscript.