Lumbar epidural steroid injections for lumbosacral radicular pain in patients with normal imaging: A propensity-matched study

Abstract

Background

Epidural steroid injections are indicated for radicular pain, with a pre-injection MRI not mandated in guidelines. There is conflicting evidence that MRI findings correlate with outcomes.

Methods

Fourteen patients with near-normal imaging (i.e., no evidence of nerve root compression and minor degenerative changes if present) who underwent ESI for lumbosacral radicular pain and were followed for up to 12 weeks were propensity matched against 14 patients with radicular pain secondary to concordant MRI pathology. The primary outcome measure was mean reduction in average leg pain relief at 4 and 12 weeks. Secondary outcome measures included average back pain, function, analgesic reduction, satisfaction, and a categorical measure of success predesignated as a ≥2-point decrease in average leg pain score coupled with a positive global perceived effect and not requiring any additional intervening interventions.

Results

For mean reduction in average leg pain at 4 weeks, there were no significant differences between those with near-normal MRIs and those with abnormal imaging (2.36 (SD 2.55) vs. 2.61 (SD 2.15); P = 0.72). For average back pain reduction at the same time point, the mean reduction was 0.75 (1.73) among cases vs. 1.07 (2.01) in control patients (P = 0.57). There were also no differences observed in pain reduction outcomes at 12 weeks. The average reduction in Oswestry Disability Index at 12 weeks favored the near-normal imaging group (8.64 % (SD 11.36) vs. 0 % (7.69); P = 0.047). A trend was noted wherein more patients with abnormal imaging experienced a positive outcome at 4 weeks (50 % vs. 28.57 %; P = 0.22) but not 12 weeks (28.57 % in both groups).

Conclusions

There were no significant differences in pain outcomes compared to control patients, though patients with near-normal imaging fared worse than historical controls and the larger cohort from which propensity-matched patients were selected. The possibility of poorer outcomes should be considered when selecting patients with normal imaging and radiculopathy for ESI.

1. Introduction

Epidural steroid injection (ESI) is the most commonly performed procedure in pain clinics worldwide, with the main indication being radicular pain [1]. Radicular pain can arise from several causes including infectious (e.g., herpes zoster), nerve root ischemia (e.g., neurogenic claudication), frank compression (e.g., from a herniated disc), or from chemical irritation, which is usually the result from inflammatory cytokines extravasating from degenerated discs and sensitizing nerve roots [2,3]. The clinical relevance of this latter mechanism is subject to debate, but without significant degeneration (which results in elevated intradiscal cytokine levels) accompanied by a full-thickness annular tear(s), there is no avenue by which chemical irritation of nerve roots could feasibly occur.

Radicular pain is defined as pain arising from activation of nociceptive fibers in a nerve root and dorsal root ganglion (DRG), while radiculopathy-which technically derives from the Latin terms for “disease” and “nerve root(s)”- often refers to nerve root pathology resulting in objective findings such as neurological weakness, dermatomal sensory loss, loss of reflexes, or abnormalities on electrodiagnostic testing [4,5]. Yet, numerous national guidelines from the American College of Physicians, World Institute of Pain, Multispecialty Working Group under the auspices of the U.S. Food and Drug Administration, as well as review articles and guidelines from major insurance companies assert that radicular pain, without a requirement for abnormal imaging or other objective evidence of nerve root involvement, is the principal indication for an ESI [[6], [7], [8], [9], [10], [11], [12], [13]]. This is in part based on a randomized trial finding that an MRI did not significantly improve outcomes or affect decision making in patients with a clinical presentation of lumbosacral radicular pain [14].

It is well-known that there is a high prevalence rate of MRI abnormalities in asymptomatic individuals, and that unnecessary MRIs lead to unnecessary procedures [15]. In light of the low specificity of MRI to detect pain-generating pathology, studies evaluating surgical decompression have found an inverse correlation between outcomes and the size of a disc herniation, presumably because very small herniations are more likely to be incidental findings in patients who present with otherwise unexplained sciatica [16,17]. For lumbar facet joint blocks and radiofrequency ablation, the correlation between response and imaging findings is mixed at best [18]. Regarding ESI, there are no studies that have evaluated outcomes in patients with normal imaging. Although several studies found that patients with non-compressive disc degeneration fared worse than individuals with a compressive etiology stemming from a herniated disc or spinal stenosis, a more recent systematic review found no consistent relationship between imaging findings and ESI outcomes [[19], [20], [21]]. Given recommendations from guidelines and authorization requirements, it is a near-certainty that there are many individuals who present with radicular pain accompanied by completely normal imaging and receive ESI. Yet, the selection criteria for clinical trials that require imaging exclude these patients, leading to a glaring literature gap on how these patients fare clinically. The objective of this study is thus to determine pain and functional outcomes in patients who present with clinical evidence of radicular pain with normal or near-normal MRIs.

2. Patients and methods

Approval to conduct this case-control study was granted by the Johns Hopkins and Northwestern Institutional Review Boards, and patients who provided informed consent. All patients were treated over a 16-year period from January 5, 2009 until November 1, 2024 owing to the “rarity” of the event. Seven of 14 “cases” were recruited from a prior study evaluating the effect of MRI on ESI outcomes (i.e., the group in which the doctor did not look at the MRI before the ESI), with the others referred from surgeons or neurologists for diagnosis and treatment [14]. Control patients were garnered from the same study and several other prospective studies in which the same data were recorded [14,19,[22], [23], [24]]. Inclusion criteria were normal or near-normal MRI, defined as no greater than a modified Pfirmann grade 3 (out of 8) degeneration at a single disc without evidence of a full thickness annular tear; radicular-type pain radiating into the lower leg; symptom duration >3 months; pain described in terms such as “shooting”, “stabbing”, “burning” or “lancinating”; and absence of unequivocal neurological signs such as sensory deficit in a single dermatomal distribution; neurological muscle wasting or loss of reflex concordant with single disc minor degeneration (Fig. 1) [25]. Excluded from participation were individuals without an MRI, an MRI showing pathology concordant with symptoms, or with diffuse (>1 disc involved) or extensive (protruding or herniated disc or moderate degeneration) disc degeneration based on the modified Pfirmann grading system; diffuse pain, fibromyalgia or > 3 active pain conditions; ESI within 2 years; provocative tests consistent with piriformis syndrome; severe psychiatric condition; and individuals with ongoing secondary gain.

Fig. 1.

T2-weighted sagittal magnetic resonance imaging (MRI) scan in a study subject with a near-normal MRI (with permission).

2.1. Epidural steroid injections

ESI were performed under the supervision of or by a pain medicine specialist using image guidance, with local anesthesia provided with lidocaine 1 %. The spinal level targeted was based on clinical presentation. In all cases, interlaminar ESI were performed for bilateral pain while transforaminal injections were done for those with unilateral symptomatology. For interlaminar ESI, a 20-gauge Tuohy needle was inserted midline for symmetrical symptoms or parasagittally towards the more painful side. The needle was then advanced into the epidural space using fluoroscopic imaging in multiple views and the loss of resistance technique. For transforaminal ESI, a 22-gauge spinal needle was inserted in a coaxial view into the targeted foramen with the image intensifier angled obliquely >30°. With transforaminal injections, we generally employed a an infra-pedicular, “preganglionic approach” for paracentral herniations affecting the traversing nerve root, i.e., for those with suspected L5 pathology, the L4-5 (rather than L5-S1) level was chosen for transforaminal injections. Proper positioning was confirmed by real-time contrast injection. After confirmation of correct placement, a 4 mL solution containing 1–1.5 mL of 40 mg/ml of depo-methylprednisolone, 1 mL of bupivacaine 0.25 % and 1.5–2 mL of saline was injected for interlaminar ESI. For transforaminal injections, a 3 mL solution containing 1 mL of bupivacaine 0.25 %, 0.5–1 mL of normal saline, and either 1–1.5 mL of depo-methylprednisolone 40 mg/mL or 1 mL or dexamethasone 10 mg/mL, was administered.

2.2. Data collection

Between the injection and first follow-up, no co-interventions were permitted. Baseline data collected on the date of the procedure included demographic information, clinical information including duration of pain, laterality, average back and leg pain scores on a 0–10 numerical rating scale (NRS) over the past week, Oswestry Disability Index version 2.1 (ODI) score expressed as a percentage, level and type of ESI, and opioid and non-opioid analgesic usage [14,26]. The first follow-up visit occurred 4-weeks post-injection during which back and leg pain scores, ODI scores, analgesic reduction defined as cessation of a non-opioid analgesic or 20 % reduction in opioid use, and a binary global perceived effect (GPE) outcome based on the question “How do you feel your procedure helped you [23]:

• 1)Not at all or I feel worse;
• 2)a little bit but not appreciably better;
• 3)Somewhat better, a definitely noticeable improvement; or
• 4)Dramatically better, a marked improvement?

For the GPE to be designated as positive, a patient had to respond with answers 3 or 4 (i.e., at least somewhat better). For the one propensity-matched patient whose data was obtained from a later study in which satisfaction was measured on a 5-point Likert scale, a positive GPE was defined as a score of ≥4/5 (i.e., satisfied or very satisfied) [19]. In addition to the above-noted outcomes, a positive categorical outcome was pre-designated to be a 2-point or greater reduction in average leg pain coupled with a positive GPE and not requiring any additional procedural or pharmacological therapies for pain. For those who experienced a positive categorical outcome at 4 weeks, the final follow-up occurred at 12-weeks post-procedure at which time the same outcome measures were collected, while those who experienced a negative 4-week outcome exited the study to receive alternative care. The co-primary outcome measures were mean reduction in average leg pain score and the proportion of patients with a positive binary outcome.

2.3. Statistical analysis

RStudio was used to evaluate primary and secondary outcomes. Categorical variables such as global perceived effect and successful outcome were analyzed using chi-square or Fisher's exact test as appropriate, while continuous variables such as pain and disability scores were compared using paired (normal MRI within group differences) or independent t-tests. A P-value <0.05 was used to define statistical significance.

Propensity matching was performed with the SPSS R Essentials plug-in. Matched data for patients with concordant imaging were obtained from 5 prospective trials containing 496 patients who received ESI and had the same follow-up periods, with the same methodology for dealing with unsuccessful outcomes (last failed observation carried forward) [14,19,[22], [23], [24]]. For one study in which patient global impression of change was used instead of the binary GPE, a score of 5 out of 7 was considered ‘positive’. Data were paired using age, duration of pain, baseline leg pain, and having a co-existing psychiatric disorder as matching variables. Matching was performed using the “nearest neighbor” method based on a tolerance level of 0.2, prioritizing baseline leg pain score and duration of pain.

Since there is no biological basis for radicular or axial LBP to improve from ESI after 4 weeks and patients were eligible to receive other treatments, we employed the conservative strategy of carrying forward (last observation carried forward) leg, back and ODI scores at 4 weeks–12 weeks for those with a negative categorical outcome. We did the same with negative binary outcomes for medication reduction and GPE, though for individuals with a negative categorical outcome and positive medication reduction or GPE, these variables were dropped at 12 weeks.

Because of the rarity of performing an ESI in individuals with a normal MRI, we conducted a pre-study power analysis to determine the likelihood of finding a large significant effect for mean reduction in average leg pain score between groups. Assuming an average decrease of 1-point in the normal MRI group and a decrease of 2.8-points (SD 2.45) at 12 weeks, a propensity-matched study with 14 people in each group would have 50 % power to detect a difference in mean reduction in average leg pain score with an alpha set at 0.05 for a 2-tailed independent t-test.

3. Results

3.1. Baseline demographics

Fig. 2 shows the flow of subjects and the proportion with successful outcomes at 4- and 12-week follow-up visits stratified by treatment group. There were 14 patients in each group, with an estimated 4 additional patients with near-normal MRIs and focal radicular pain who did not receive payor authorization to receive a procedure. Patients with near-normal MRIs were clinically and demographically similar to their propensity-matched controls. The average age in the treatment group was 33.93 (9.50) vs. 37.14 (14.83) in the control group with concordant imaging. The mean duration of pain was 4.30 years (3.03) among those with normal MRIs vs. 2.50 (1.61) in those with abnormal imaging. 57.14 % vs. 50 % in the case and control groups identified as females, respectively. 35.71 % of cases identified as White vs. half of the control group. One patient in the normal MRI group and 2 in the control group received transforaminal injections with dexamethasone while the rest received depo-methylprednisolone, with no differences in any outcomes stratified by type or dose of steroid (Table 1).

Fig. 2.

Flow chart demonstrating the progression of study subjects through 12-week follow-up.

Table 1.

Baseline demographic and clinical characteristics by study group.

Variable	Normal MRI Cases	Concordant MRI Controls
Mean age in years (SD)	33.93 (9.50)	37.14 (14.83)
Sex (Female, %)	8 (57.14 %)	7 (50 %)
Race (White, African American, Mixed, Other)	5, 4, 0, 5	7, 5, 1, 1
Psychiatric Co-Morbidity (n, %)	6 (42.86 %)	6 (42.86 %)
Mean Duration of Pain in Years (SD)	4.30 (3.03)	2.50 (1.61)
Baseline ODI Score (SD)	42.36 (15.16)	42.21 (13.61)
Baseline Leg Pain Intensity (SD, NRS)	6.96 (2.48)	6.29 (1.63)
Baseline Back Pain Intensity (SD, NRS)	6.57 (2.74)	6.21 (1.57)
Opioid Use (n, %)	4 (28.6 %)	7 (50 %)
Epidural Injection Type (n, %)
Transforaminal	6 (42.86 %)	10 (71.43 %)
Interlaminar	8 (57.14 %)	4 (28.57 %)

NRS- 0–10 numerical rating scale; ODI- Oswestry disability index.

3.2. Pain and functional outcomes

Outcomes stratified by group are presented in Table 2. There were no significant differences in any pain outcomes. For average leg pain at 12 weeks, the mean reduction in pain scores were 2.50 (SD 2.67) in those with normal MRIs vs. 1.82 (2.14) in controls whose imaging showed concordant pathology (P = 0.52). For average back pain at 12 weeks, the mean reductions were 0.93 (2.14) and 0.86 (2.24) in case and control patients, respectively. One patient in each group experienced complete eradication of both back and leg pain (0/10 pain) at 4 weeks, with the absence of pain persisting at 12 weeks for the person in the normal MRI group. For ODI scores, the average reduction was 9.21 (SD 12.05) in cases at 4 weeks vs. 3.50 (9.36) in control patients (P = 0.18). At 12 weeks, the mean difference in improvement became statistically significant favoring the normal-MRI patients (8.64 (11.36) vs. 0 (7.69), respectively; P = 0.047).

Table 2.

Outcomes stratified by primary treatment group.

Variable	Normal MRI	Concordant MRI Controls	P-Value
Average Leg Pain at 4 Weeks (SD)	4.61 (3.11)	3.68 (3.01)	0.43
Mean Reduction in Leg Pain (SD)	2.36 (2.55)	2.61 (2.15)	0.72
Average Back Pain at 4 Weeks (SD)	5.82 (2.74)	5.14 (2.82)	0.57
Mean Reduction in Back Pain (SD)	0.75 (1.73)	1.07 (2.01)	0.57
Average Leg Pain at 12 Weeks (SD)	4.46 (3.12)	4.46 (2.73)	1.0
Mean Reduction in Leg Pain (SD)	2.50 (2.67)	1.82 (2.14)	0.52
Average Back Pain at 12 Weeks (SD)	5.64 (3.11)	5.36 (2.84)	0.85
Mean Reduction in Back Pain (SD)	0.93 (2.14)	0.86 (2.24)	0.93
ODI at 4 Weeks (mean %, SD)	33.14 (19.24)	38.71 (16.27)	0.48
Mean Improvement in Function (SD)	9.21 (12.05)	3.50 (9.36)	0.18
ODI at 12 Weeks (mean %, SD)	33.71 (19.00)	42.21 (13.40)	0.30
Mean Improvement in Function (SD)	8.64 (11.36)	0 (7.69)	0.047
Positive Global Perceived Effect at 4 Weeks (n, %)a	8 (57.14 %)	10 (71.43 %)	0.69
Positive Global Perceived Effect at 12 Weeks (n, %)a,b	4 (40 %)	6 (54.54 %)	0.67
Medication Reduction at 4 Weeks (n, %)c	4 (36.36 %)	3 (30 %)	1
Medication Reduction at 12 Weeks (n, %, 95 % CI)b,c	(1, 11.11 %)	(4, 44.44 %)	0.29
Successful Outcome (N, %)d			0.29
None	10 (71.43 %)	7 (50.00 %)
At 4 Weeks Only	0 (0.00 %)	3 (21.42 %)
Through 12 Weeksb	4 (28.57 %)	4 (28.57 %)

NRS- numerical rating scale; ODI- Oswestry disability index; GPE; global perceived effect.

^aPositive GPE defined as a score ≥3/4 on a Likert scale corresponding to “Somewhat better, a noticeable improvement”.

^bNegative GPE, medication reduction and successful outcome at 4 weeks carried over to 12 weeks.

^cMedication reduction defined as cessation of non-opioid analgesic or >20 % decrease in opioid use.

^dSuccessful procedure defined as ≥ 2-point decrease in leg pain coupled with positive GPE and no medication increase or need for adjunct therapy.

3.3. Categorical outcomes

Positive GPE rates did not significantly differ between groups, with 57.14 % and 40 % of those with normal MRIs being satisfied with their outcomes at 4- and 12-weeks, respectively, vs. 71.43 % and 54.54 % of control patients. At 12-weeks post-ESI, 44.44 % of control patients were able to reduce analgesic intake vs. only 11.11 % of those with near-normal MRIs (P = 0.30). For successful outcomes at 4 weeks, 28.57 % (N = 4) of patients with near-normal MRIs met criteria for a positive outcome vs. 50 % (N = 7) in the control group (P = 0.22). Compared to the success rate of the entire cohort from which the propensity-matched controls were drawn, the success rate in those with near-normal MRIs was lower (56.85 % vs. 28.57 %; P = 0.037). At 12 weeks, 4 patients in each group (28.57 %) continued to experience a positive outcome.

4. Discussion

The main findings in this study are that while some people with an absence of or minimal pathology fare well with ESI, they may be less likely to benefit than historical controls and the entire control cohort, though the outcomes between our propensity-matched control cohort and those with near-normal MRIs did not significantly differ [2,14,19,[22], [23], [24]]. This is because the computer-generated control group also fared worse than both historical controls and the entire cohort from which it was drawn, who had a 4-week success rate of 56.85 %. The effect in those with normal MRIs in this study is comparable to that reported after sham procedures [23,27,28].

4.1. Comparison to extant literature

As noted above, although some studies have reported poorer outcomes for ESI with non-compressive than with compressive etiologies [19,20], a systematic review failed to find any consistent relationship between outcomes and pathology [21]. However, this review found 2 studies in which “low-grade” nerve compression responded better to “high-grade”, and two others showed better outcomes when electromyographic studies were positive. In contrast, based on two studies spinal decompression surgery outcomes are more likely to be positive when the pathology is more pronounced [16,17]. Reasons for this disparity are likely related to the intervention. Since an ESI is not removing the offending source of decompression, low-grade pathology which causes mechanical or chemical inflammation [2,29] rather than a persistent compressive injury may be more likely to respond than when there is high-grade, continuous compression of the offending nerve root that persists even after the intervention.

The lack of an association between procedural outcomes and objective findings is not limited to ESI. For carpal tunnel release, a systematic review found that 8 of 10 studies failed to find a strong correlation between electrodiagnostic tests and surgical outcomes, with many finding no correlation [30]. For cubital tunnel release, one systematic review found no clear trend emerging on the predictive value of preoperative electrodiagnostic testing (4 of 9 studies were positive), while a more recent one found low-certainty evidence that very mild or absent findings on nerve conduction studies is predictive of poor surgical outcome [31,32]. Subsequent studies have similarly yielded mixed results [33,34]. In a highly publicized study examining the correlation between lumbar decompression surgery outcomes and postoperative MRIs, the authors found no association between having a visible disc herniation and clinical outcome [35].

4.2. Explanation of findings

Absent frank compression, there are other causes of nerve root irritation including chemical irritation from inflammatory cytokines and ischemia, which is a hallmark of neurogenic claudication, though our patients had no evidence of spinal stenosis. Chemical irritation from tumor necrosis factor-alpha (TNF-α) and other cytokines is a well-known etiology for radiculopathy, though studies examining the injection of cytokine inhibitors have yielded mixed results [3,23]. Although present in non-degenerated discs, the concentration of these inflammatory mediators (TNF-α, interleukins 1 and 6, etc.) increases exponentially with degeneration [36], and without frank herniation or full-thickness annular disruption, the path to egress and chemically stimulation of a nerve root(s) closes off. Although the changes are small, studies have also shown an increased disc protrusion volume in the standing vs. sitting position and in morning vs. evening MRIs [37,38]. Although we do not know time of day the MRIs were conducted, no patient in our study had a standing MRI.

These patients may suffer from central sensitization (and/or somatization), in which normal pain signals are amplified. By definition, these patients all had non-organic signs (e.g., regional sensory disturbances inconsistent with an anatomical pattern). Non-organic signs have been shown to be associated with poor ESI and other procedural outcomes, and at least some have been postulated to be secondary to central sensitization [39,40]. For somatization, research suggests that these patients may also suffer from decreased descending inhibition and attentional failure that can amplify (lumbar) sensory signals which studies show all individuals experience when sufficiently queried [41,42].

There is ongoing controversy regarding the efficacy of ESI, which stems in part from overuse and conflicting results from randomized controlled trials [1,2]. The placebo effect is indubitably greater than the intrinsic effect for pain therapies, being even stronger for procedures such as ESI and surgery [27,28]. Although patients with somatization and central sensitization are more likely to fail interventional therapies [2,39], their placebo response rates are paradoxically similar to patients with organic pathology [43,44]. Since the physiological effects of epidural corticosteroids do not typically extend beyond 6–8 weeks [45], the decline in the responder rate in the concordant MRI group but not the normal MRI group is probably attributable to either a difference in the placebo effect or a difference in the natural history of the two groups. Examining the re-injection rates and the response to subsequent procedures over longer time periods might have provided additional insight on both the mediators of response and long-term outcomes.

4.3. Limitations

There are several limitations to this analysis. First, although most patients were recruited as part of clinical studies with others accumulated based on referrals, the patients were not necessarily consecutive (i.e., some may have been missed), and the control patients were garnered from a database containing slightly under 500 patients. Thus, data that might have affected outcomes such as instruments measuring somatization or central sensitization was not recorded. Second, these patients presented with presumptive radicular pain without supporting MRI evidence of pathology. The use of electrodiagnostic testing, psychometric testing or even quantitative sensory testing may have provided more information regarding mechanisms, including peripheral, central, and psychological contributors. Last, the small numbers included, a byproduct of both the rarity of the event and obstacles in obtaining authorization for ESI in the context of normal MRIs, precluded obtaining a larger dataset that might have provided greater statistical power.

5. Conclusions

For the main outcomes, there were no significant differences between outcomes for those with near-normal MRIs and the control group with abnormal imaging suggesting that some of these patients benefit. The lack of a significant difference likely stems from the small number of patients with near-normal MRIs who underwent ESI, the lower success rates in the propensity-matched controls than in the entire cohort from which it was drawn, and the high placebo effects associated with interventional procedures which may drown out small signals of efficacy.

Funding statement

Funded in part by a grant from MIRROR, Uniformed Services University of the Health Sciences, U.S. Dept. of Defense, grant # HU00011920011. The sponsor did not play a role in study design or performance, or analysis or interpretation of data.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Steven Paul Cohen reports financial support was provided by US Department of Defense. Milan Stojanovic reports a relationship with International Pain & Spine Intervention Society that includes: board membership. Steven Paul Cohen reports a relationship with American Society of Regional Anesthesia and Pain Medicine that includes: board membership. SPC: In the past 2 years, Dr. Cohen served as a consultant and received research funds (paid to institution for a product not related to epidural injections) from Scilex, which has an injectable steroid being studied for epidurals in development. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.