Finding an Answer: Comments on a Randomized Trial of Epidural Glucocorticoid Injections for Lumbar Spinal Stenosis

Abstract

Recently, Friedly et al. published a study on the effectiveness of epidural injections of steroids (ESIs) for lumbar spinal stenosis (LSS) [1]. Because of its apparent implications for clinical practice, this study attracted comments both in the professional [2] and lay press [3]. However, the study prompts reflection on a number of issues that extend beyond the conventional appraisal of a study. Those issues pertain to how physicians think about a condition, how it should be treated, and how to best assess the effectiveness of that treatment. The following article addresses those issues in the interest of informing physicians about how contemporary practices are confounded by lack of diagnostic discipline, and how this makes the conduct of controlled trials complicated and, therefore, makes the results difficult to assess. The cardinal issues are the symptoms of LSS; the diagnosis of LSS; the rationale for treatment; the consequent appropriate selection of patients; the outcomes assessment, either in practice or in a study including the reporting of those outcomes; and the need for rigorous control of the technical performance of the therapeutic intervention. Each of these issues has a critical bearing on appraising not only the literature on LSS, but also on how physicians treat this condition in conventional practice. The literature is profuse with heterogeneity in symptomology related to LSS. While there is no consensus in the current literature on acceptable LSS symptoms, the most commonly quoted symptoms are neurogenic claudication, radicular pain, and low back pain (LBP). Additional symptoms commonly mentioned in literature were fatigue and loss of power in the legs, anesthesia, and a feeling of numbness in the sacral dermatomes. In an attempt to achieve consensus, Katz et al. in 1995 published symptoms most strongly associated with the diagnosis of LSS (likelihood ratio ≥ 2), which included advanced age, severe lower-extremity pain, and absence of pain when seated [4]. In a recent, randomized trial by Weinstein et al., Spine Patient Outcomes Research Trial (SPORT), enrollment criteria included LSS on cross-sectional imaging and symptoms of either neurogenic claudication or radicular pain [5]. Patients were randomized to surgery and non-operative treatment, with short-term results favoring surgery. However, similar benefits from surgery were noted in patients with and without neurogenic claudication. A subgroup analysis demonstrated that patients with predominant leg pain improved significantly more with surgery than predominant LBP patients [6]. Overall, neurogenic claudication is the most commonly mentioned symptom of LSS; although the North American Spine Society guidelines list back pain as the most common complaint with a prevalence of 95%, followed by claudication (91%), leg pain (71%), weakness (33%), and voiding disturbances (12%) [7]. Claudication was first described by Dejerine in association with syphilitic arteritis of the spinal cord [8]. Blau and Louge described six cases of intermittent claudication caused by protrusion of a lumbar disc [9]. A 1964 report of two cases of “intermittent claudication from compression of cauda equina” resolved by lumbar laminectomy by Brish et al. was published 3 years later [10]. The contemporary definition of neurogenic claudication includes decreased ability to walk and stand with discomfort that radiates beyond the spinal area into the buttocks and frequently into the thigh and lower leg; it is exacerbated by lumbar extension and improves with lumbar flexion [17]. If neurogenic claudication is considered a hallmark symptom of LSS, a recent review of seven randomized controlled trials (RCT) revealed that the actual presence of neurogenic claudication was used as eligibility criteria in only 71% of LSS studies. When neurogenic claudication was necessary for enrollment eligibility, its definition varied considerably across studies, including symptoms such as “fatigue or loss of sensation in the lower limbs aggravated by walking” or “sitting as a better position for symptom severity than standing or walking” [24]. Similar to the variability in defining neurogenic claudication, all studies in this review presented imaging findings consistent with LSS, but a detailed radiological definition of LSS was provided only 57% of the time. The emphasis on neurologic features in the original literature was consistent with the proposed pathology [17,19,11–13]. By definition LSS is a narrowing of the spinal canal; therefore, it is logical that pain and neurologic symptoms would be a result of compromise of the nerve roots contained within the canal. The same link is missing for back pain: The pathophysiology of such a mechanism has neither been explained, nor supported in the literature. All the experimental and clinical evidence points to back pain arising from sources other than the nerve roots, such as the intervertebral discs, the zygapophysial joints, the sacroiliac joint, or perhaps the back muscles. Until proven otherwise, it becomes more rational to infer that the back pain reported by patients with LSS arises from one or another of these structures, rather than being caused by compression or ischemia of the cauda equina. This has been demonstrated in a study of patients with radiographic LSS, where the actual source of the patient’s back pain was traced to the zygapophysial joints, which was then treated by medial branch radiofrequency neurotomy [14]. In contrast to low back pain, it seems theoretically plausible that patients with LSS could present with radicular pain. However, it is important to note the differences between radicular pain caused by disc protrusion and neurogenic claudication caused by spinal stenosis. Although both are similar in clinical presentation, they have significant differences in pathophysiology and natural history analogous to the example comparing exercise induced angina and myocardial infarction (although both present with pain and are due to the same underlying principles). These differences can affect treatment outcomes and for that reason it is important to exactly identify the symptomatology of LSS before selecting a treatment modality or designing an outcome study. Despite the fact that LSS has been recognized for over 50 years, there is lack of consistency in use of this nomenclature: whether LSS is a clinical entity, a radiologic observation, or a term describing an anatomical state of the lumbar spine [15]. Historically, the first comprehensive report of LSS by Verbiest in 1954 described seven cases of a clinical condition in which there were symptoms of compression of the caudal nerve roots on standing or walking, but not at rest [16]. In all cases myelography showed a block in the lumbar region and narrowing of the spinal canal was found during surgery. The author suggested that the narrowing was due to encroachment on the spinal canal by the articular processes and that decompression of the dural sheath may be followed by complete relief. An updated and more accurate anatomical definition of LSS proposes that central LSS may result from a decrease in the anteroposterior, transversal, or combined canal diameter secondary to loss of disc height with or without bulging of the intervertebral disc; and hypertrophy of the facet joints and the ligamentum flavum [5]. The same process can also lead to lateral recess and foraminal spinal stenosis. The pathophysiology of LSS has been attributed to mechanical compression, ischemia, or both, of the lumbosacral nerve roots due to narrowing of the lateral and central vertebral canals [17,18]. The likely mechanism of ischemia is mechanical compression leading to further pressure on the venules surrounding the nerve roots causing its engorgement leading to ischemic and inflammatory nerve impairment [6,19–21]. Despite the anatomical description of LSS, the fact that up to 21% of asymptomatic subjects have significant radiographic findings of LSS poses an obstacle in reaching a consensus on diagnostic criteria for LSS, and also raises questions regarding the pathophysiologic mechanisms [22–24]. These findings imply that even detailed radiographic criteria alone (as summarized by Steurer et al. [25]) do not establish that LSS is the cause of a patient’s symptoms. The actual cause would be additional pathologic change leading to symptoms. A similar mechanism exists in coronary artery disease. Patients with atherosclerotic plaques causing coronary artery narrowing can be asymptomatic, only have exercised-induced angina, only have non-pain symptoms such as shortness of breath, or they could even suffer from an acute myocardial infarction. If an author presented a study to the New England Journal of Medicine for a treatment of coronary artery disease, with inclusion criteria limited to the presence of atherosclerosis and shortness of breath, it is probable that the article would not be published. This is because the authors would have failed to exclude other common causes of shortness of breath such as chronic obstructive pulmonary disease, asthma, anemia, and even a pulmonary embolism. Even if the proposed treatment were effective for atherosclerosis, the study would only demonstrate a positive effect if the authors were lucky enough to enroll a large percentage of subjects with this disease versus those with another. Unfortunately this exact problem is endemic in spine research. Due to the lack of a clear diagnostic algorithm for patients with spine pathology, both clinical care and even highly publicized research studies suffer from the critical flaw of heterogeneity [26]. It seems that LSS might be promoted to a diagnosis only if the attributable symptoms, and the cause of those symptoms, are stipulated; all too often this is not done in the literature. Perhaps the reason is that the possible combinations are too diverse both for anatomical variations (central, lateral recess, foraminal, size of stenosis) and symptoms of LSS. It would be very cumbersome to stipulate all the variations. However, lack of discipline in this regard is not without significance for studying the outcomes of ESIs since certain combinations may provide evidence for the use of such treatment, but for others the evidence may be lacking. In order to establish a diagnosis, a consensus on diagnostic criteria, including a combination of objective radiographic anatomical findings and patient symptoms, needs to exist. An explicit rationale for the use of epidural steroids for LSS has rarely been stated. The analgesic mechanism of action of steroids has not been clearly explained and accepted. The most accepted theory is the role of steroids in suppressing inflammation. In the lumbar spine, the inflammatory mediators can be present in elevated concentrations in degenerative discs [27] and may contribute to sensitization of peripheral nociceptors via prostaglandins [28]. Alternatively, steroids have been shown to decrease the ectopic discharge in injured nerves and decrease nerve conduction in small fibers [29,30]. These theories form a foundation for the role of steroids in treatment of radicular pain, with or without radiculopathy, caused by acute lumbar disc herniation even without full scientific proof. As opposed to acute disc herniation, the evidence for the role of inflammation in LSS is sparse, and therefore provides a weaker rationale for use of epidural steroids. In vitro evidence exists that inflammation can increase ligamentum flavum hypertrophy [31–33], but no evidence exists that steroids or oral anti-inflammatory medications can address that inflammation. There is some evidence that inflammation is involved to some degree in nerve root pathology in LSS [10,34]. It is possible that some unknown mechanism of epidural steroids (or other substances injected into epidural space along with steroids) may have beneficial effect in patients suffering from spinal stenosis related pain. Future research may shed light on this. Until then, a sound scientific rationale for the use of epidural steroids is not robust. It is important to have some knowledge of the expected outcomes of the disease in the absence of treatment. In the case of Friedly et al., the authors demonstrate pain relief nearing 40% in pooled subjects at 6 weeks post-procedure. This improvement is substantially greater than the typically predicted placebo response [35–37]. Additionally, there are other studies showing responses following epidural injections in the treatment of lumbar stenosis [38–41]. In some spine conditions, such as herniated discs, patient improvements at 6 weeks following a procedure may be attributed to either the natural history of the disease or the intervention provided to the patient. In the case of lumbar stenosis, the known natural history of the disease is typically not improvement over time [42]. When designing a prospective trial, all variables that may affect the outcome should be accounted for in determining the appropriate study size. In the case of spinal stenosis, this should include procedural variables such as: medication utilized, dosage, and route of injection. It would also include patient variables that may affect the outcome including: degree of stenosis, severity of symptoms, and duration of symptoms. Although Friedly et al. should be commended for performing a large RCT on 400 subjects, it is unfortunate that these multiple independent variables were grouped and analyzed together. Patients with a variety of diagnoses and symptom severities were injected in an assortment of ways with a myriad of medications. It is highly likely that statisticians and researchers will utilize this dataset for future subgroup analyses of these variables. However, this dataset should be scrutinized closely as each of these subgroups may not have sufficient numbers to facilitate an appropriate independent analysis. Choosing a proper study design is the best way to answer the proposed clinical question. In general, RCTs are felt to provide the highest level of evidence. There are two types of RCTs available from which to choose. An explanatory study will test a treatment against a placebo or “sham” to determine efficacy of the treatment. A pragmatic RCT will test one treatment against another to determine which treatment is superior. Friedly et al. chose the question “Are epidural injections of steroids plus lidocaine better than lidocaine alone for reducing pain and improving function in patients with lumbar central canal spinal stenosis?,” thereby selecting a pragmatic RCT study design. While the use of high quality observational studies and pragmatic RCTs has been supported by the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) system [43], an explanatory RCT would be a preferred study design choice for large, multicenter trials. It would provide the highest level of evidence and better address the question of whether epidural steroids are an effective treatment for spinal stenosis. Some physicians believe that injecting lidocaine, as it relates to intermediate-term, or 6-week relief, is a placebo treatment. This argument is founded on the action of the medication and known duration of local anesthetic effects. However, there is certainly reason to question whether administration of any substance into the epidural space is a placebo treatment [44]. Therefore, in order to conduct a valid explanatory RCT a true placebo group is necessary. In study design, an investigator determines the desired patient population early on. Diagnostic inclusion and exclusion criteria must be clearly defined in order for the reader of the article to determine if the study population is relevant to his or her own practice. In determining whether or not a treatment is effective, the investigator must choose outcome measures to evaluate the effectiveness of the treatment. The ideal outcome measure will evaluate the direct impact the treatment has on the diagnosis. For example, when evaluating a medication for hyperglycemia, fasting blood sugar level is an ideal outcome measure. In the case of most spine procedures, outcome measures are typically related to the pain and dysfunction that the condition causes. To determine the effectiveness of treatment in this realm, we rely on patient-reported outcomes. These measures typically assess outcomes such as symptoms (e.g., pain, claudication), function (e.g., ability to walk, perform activities of daily living), health status, health-related quality of life, and satisfaction with treatment. The validity and reliability of outcome measures are an integral of an and selection is the degree to which a or other measure test and no change is or across or is the degree to which a measure it is to measure. measures need to be for the diagnosis being For the study of pain, the for pain and the for pain have been When studying back pain, other outcome have been and In the study of lumbar stenosis, the Claudication also to as the and have been The outcome measures have a significant impact on evaluating how effective a treatment For example, when evaluating a treatment for lumbar radicular pain, appropriate outcome measures would evaluate pain, for back and leg pain, and for However, in the case of lumbar stenosis back pain is not a defining of the condition, an investigator would typically choose an outcome measure to address the of this condition, neurogenic claudication. outcome measures are and include the no direct studies to on ESIs for lumbar spinal stenosis, including the Friedly et al. have used appropriate and outcome measures as the for this Friedly et al. a but not for spinal stenosis. Until outcomes measures are the between an effective treatment and may be difficult to In the it has been conventional to report group outcomes. This is a in the literature that can result in a of the true outcomes. (or rely on a pain to a A of is not frequently found in patients with Even if it is it is to be present a treatment. This is true if the treatment results in a of patients with complete or complete pain relief. on a treatment that results in only and is not only but also to an actual treatment response from any patient. The appropriate therefore, should be a The definition of may to the condition and can be defined by The most commonly used example is pain relief and of although studies very rigorous and definition of such as complete relief of pain with full of to and no use of other health care have been reported treatment pain with from efficacy of of steroids for the treatment of lumbar radicular by and by the American of The alone is not a certain percentage of patients may have in one that not that the same percentage of patients will have in another study or clinical practice. In that percentage can on how subjects were included in the study. In order to for these the should be The the and highest percentage of patients could In an observational if this the intervention likely has no In a pragmatic if the of the two treatment there is likely no in effectiveness between the two could be or of no to In an explanatory if the of the and likely caused the beneficial of the treatment. These of analysis need to be to the question of spinal stenosis. from ESIs in patients with spinal stenosis exist. et al. have demonstrated in an observational study that of patients can have complete pain relief with ESIs the not include patients from treatment. An observational study whether the improvement was caused by the by associated with the or if patients improved because of the natural While can be used as a secondary outcome analysis of on outcome measures should be the the studies of ESIs for spinal stenosis, including Friedly et al., not provide for the outcome It would also be to whether the treatment the need for other health such as surgery. While the in pain both as important a in surgery would be a more response to treatment. As with all studies, the full is ability to the outcomes to patients is limited and as these do not treatment for spinal stenosis. the current state of the evidence to the use of epidural injections in the treatment of patients with LSS, questions are and more research is even the most comprehensive study to clearly demonstrated that the of steroids in a population by a variety of provides no as to injecting local anesthetic alone into the epidural space for patients with LSS on imaging [1]. However, since this study was a study only the effectiveness of steroids over lidocaine, it be used to answer the question of whether are effective for LSS. This would a study. any trial that diagnostic treated patients with a variety of symptoms in a variety of and failed to report complete outcomes could not provide any As a multicenter trial, it provide an of as present clinical practice in the but it may not a test of the of epidural steroids for LSS. is that the outcomes from this randomized controlled trial, with its will be used to a treatment. with that one can be injections for a patient population with lumbar spinal stenosis on which to be how they are in clinical practice, do not have a treatment effect beyond that of The authors to for his and of the Spine for review and and and of the Spine for review and comments on the A also to and Spine Society for and with the of this