Epidural steroid injections compared with gabapentin for lumbosacral radicular pain: multicenter randomized double blind comparative efficacy study

Participants and settings

The study sites were four joint service military treatment facilities, three of which serve as teaching hospitals (Walter Reed, San Diego, and San Antonio) and one of which is located in Europe; a Veteran’s Administration hospital; and three civilian teaching hospitals (Johns Hopkins, Case Western, and Penn State).

All participants were treated between 15 December 2011 and 10 June 2014. Inclusion criteria were age ≥17; an average score for radicular leg pain of ≥4 on a 0-10 numerical rating scale over the preceding week or 3/10 if the leg pain was as bad as or worse than back pain; current symptoms had lasted for more than six weeks and up to four years; and signs (such as straight leg raising test) and/or symptoms (such as lower leg pain) of lumbosacral radicular pain. All participants were also required to have findings of a herniated disc or spinal stenosis on magnetic resonance imaging, concordant with their presentation. Patients were permitted to have symptoms in more than one dermatome.

Exclusion criteria were neuropathic pain for more than four years in duration; previous failed trial with, or adverse reaction to, gabapentin or pregabalin; epidural steroid injections within the past three years; cauda equina syndrome; referrals for surgical evaluation; previous lumbar spine surgery; pregnancy; allergic reaction to contrast dye; known secondary gain (such as active litigation); active infection; and serious medical (such as poorly controlled diabetes, cirrhosis, recent myocardial infarction, congestive heart failure, or unstable angina) or psychiatric condition (such as dementia, poorly controlled post-traumatic stress syndrome as indicated by a score ≥44 on the post-traumatic stress disorder checklist,27 substance abuse, somatization disorder, or depression as indicated by a Beck Depression score ≥2128) that might preclude an optimal response to treatment. Because our intention was to evaluate the relative effects of two commonly used treatments for radicular pain, we also excluded patients with neurogenic claudication from spinal stenosis who did not report lower leg pain (that is, those who had only weakness or paresthesias). Although heterogeneity is often desirable in comparative effectiveness trials, we performed briefings, debriefings, full time access to a senior investigator, and at one site practice enrollments to enhance selection consistency.

Randomization and interventions

We randomized 145 participants in a 1:1 ratio by computer generated randomization tables. An investigator physician enrolled participants, stratified by study site. Research nurses performed allocation in groups of 36 at Walter Reed and Johns Hopkins as these sites were expected to enroll more patients, and in groups of 18 at other sites, with treatment allocated with a sequentially numbered opaque sealed envelope before injection. Larger allocation blocks were used to promote allocation concealment with investigators. Participants at each site were suballocated separately on a 1:1 ratio based on the type of epidural steroid injection they received: those with unilateral pain received unilateral transforaminal epidural steroid injections, while those with bilateral pain underwent interlaminar epidural steroid injections. The patient, research nurse, and evaluating physician were blinded to assignment.

Epidural injections

A board certified pain medicine physician conducted or supervised all procedures using fluoroscopic guidance. The segmental level at which the injection was administered was selected based on signs, symptoms, and radiological findings. For interlaminar injections, a Tuohy needle was inserted in or near the midline and advanced into the epidural space with image guidance in the anteroposterior and lateral views by using the loss of resistance technique. For transforaminal epidural steroid injections, a 22 gauge spinal needle was inserted coaxially into the upper part of the targeted foramina with the imagine intensifier positioned in an oblique plane. Correct placement was confirmed with the injection of contrast, which showed bilateral spread for all interlaminar injections and proximal epidural uptake for all transforaminal procedures. Once the physician was satisfied with the pattern of contrast spread, a solution consisting of 60 mg of depomethylprednisolone + 1 mL of 0.25% bupivacaine was administered. For those who received interlaminar injections, the injectate was diluted in normal saline to a volume of 4 mL; in the transforaminal subgroup, the total volume administered was 3 mL. Whereas some physicians routinely perform multilevel transforaminal injections, we limited our procedures to a single level to minimize risks because there are no controlled studies validating the use of more than one injection and because a well placed transforaminal epidural steroid injection generally spreads to multiple levels.29 30

Sham injections and maintenance of blinding

Participants were instructed ahead of time that they might or might not experience paresthesias during the procedure and were visually shielded from the image screen. The same technique (such as trajectory and use of multiplanar fluoroscopy) was used for injections except in sham injections the needle was positioned 1-2 cm proximal to the epidural space into the posterior ligaments. A small volume of saline was then injected in lieu of contrast, followed by an additional 3 mL to simulate the injectate. A generic note was entered into the medical record without radiographs.

Pharmacotherapy

A central research pharmacy over-capsulated 300 mg gabapentin and placebo capsules to appear identical. Before each shipment, the capsules were tested to ensure potency, or lack thereof for group 2. Titration schedules were prepared on a case-to-case basis in accordance with standard practice, but dosing targets generally ranged from 1800 mg/day to 3600 mg/day three times a day. Drugs were generally up-titrated over a period of 15 to 24 days, with the only caveat being that a therapeutic dose range had to be obtained at least five days before follow-up.

Co-interventions, outcome measures, follow-up, and missing data

No contact with the investigative team was permitted during the study. Tramadol and non-steroidal anti-inflammatory drugs could be prescribed on an “as needed” basis as rescue medications (or opioids could be increased by up to 20% for those taking opioids), but no other co-interventions were permitted. Participants were provided with instructions on how to taper their analgesic drugs based on response. An investigator blinded to treatment carried out the first follow-up visit one month after the start of treatment. The primary outcome measure was the average leg pain score on a 0-10 numerical rating scale recorded at one and three months, reflecting the average pain experienced during the week before follow-up. Predefined secondary outcome measures included worst leg pain over the past week, average and worst back pain, score on the Oswestry disability index (version 2.0, MODEMS, Des Plaine, IL),31 adverse effects and complications, reduction in analgesic drugs (>20% reduction in opioid use or complete cessation of non-opioid analgesics), and global perceived effect, which was defined as not requiring further non-rescue interventions along with an affirmative response to the following two statements32: “My pain has improved/ worsened/stayed the same since my last visit” and “I am satisfied/not satisfied with the treatment I received and would/would not recommend it to others.”

Because patients cannot accurately distinguish between neuropathic pain (which is more likely to respond to epidural steroid injection and gabapentin)12 33 and mechanical pain, we separately recorded leg (neuropathic) and back (mechanical) pain scores as surrogates. Whereas a small percentage of cases of axial back pain might be neuropathic in nature,34 and many forms of mechanical pain such as facet arthropathy and degenerative disc disease can radiate into the leg,35 validated instruments have shown that leg pain emanating from the back is generally indicative of radicular pain.7

The Oswestry disability index is a 10 question survey used to assess function in people with low back and/or leg pain in which higher scores indicate greater levels of disability. The classification of spinal stenosis was made if the participant had moderate or severe canal stenosis (<12 mm) not attributable to a herniated disc. We assessed complications by asking fixed and open ended questions one day after injections and at all follow-up visits. In addition to individual variables, we predesignated a positive composite outcome (that is, successful procedure) to be a decrease of ≥2 points in average leg pain score coupled with a positive global perceived effect.36

In those individuals who experienced a positive outcome at one month, the final follow-up occurred at three months. In these individuals, in addition to rescue medications, the study drug could also be titrated upwards. For ethical reasons, those with a negative outcome at one month left the study “per protocol” to receive non-study interventions, which is consistent with other randomized interventional studies.26 32 37 38 We imputed missing data points for pain scores, Oswestry disability index, and the composite outcome using the “last observation carried forward” method, which can underestimate effect sizes when “dropouts” occur because of lack of efficacy.39

Statistical analysis

We used an intention to treat strategy for all analyses. We calculated differences in treatment effects and 95% confidence intervals for pain and disability scores with analysis of covariance (ANCOVA) adjusting for baseline values of outcome measures. An indicator of the treatment group was coded such that positive values favored the epidural steroid injection group. No correction was prespecified for multiple comparisons. Because of a difference in sex distribution at baseline, we adjusted post hoc analysis of outcomes for sex. We used logistic regression models to compare the proportion of patients with adverse events in the first month and factors associated with binary outcomes in post hoc analysis. We used Poisson regression models with robust standard errors to analyze adverse event rates. Effectiveness of blinding in each treatment group was evaluated by using two indices. In the James blinding index40 (range 0-1), 0 indicates total absence of blinding, 1 indicates complete blinding, and 0.5 indicates completely random blinding. In the Bang blinding index41 (range −1-1), −1 indicates that all patients guessed the incorrect treatment, 0 indicates all patients randomly guessed, and 1 indicates all patients guessed the correct treatment. We conducted subgroup analyses to ascertain which demographic and clinical variables were associated with outcome. Variables examined included etiology (such as stenosis v herniated disc), level of injection, pain duration ≥3 months, injection type (for instance, transforaminal epidural steroid injection for unilateral pain v interlaminar epidural steroid injection for bilateral pain), smoking status, military status, presence of psychiatric disease, obesity, age, sex, and gabapentin dose. We had intended to examine the effect of inciting events related to war, such as traumatic brain injury, but there were too few patients to render any analysis meaningful. We used Bonferroni-corrected significance thresholds for post hoc subgroup analyses, with corrected P values calculated as 0.05 divided by the number of comparisons being made. All other reported P values were based on two sided tests, with <0.05 considered significant.

The study was powered to evaluate the effectiveness of epidural steroid injections compared with gabapentin. Assumptions include a difference of 1 point in pain scores between groups at one month, standard deviation of each group of 2 based on data from previous studies, a retention rate of 87%, and a two sided α level of 0.05. Although a 30% or 2 point decrease in pain has been shown to represent a clinically meaningful benefit to an individual patient, the same IMMPACT (Initiative on Methods, Measurement and Pain Assessment in Clinical Trials) guidelines note that smaller differences between groups can be considered to be clinically important in clinical trials.36 We chose a 1 point difference in pain scores because it is consistent with US Food and Drug Administration requirements for approval of adjuvant analgesic drugs for low back pain and gabapentin for neuropathic pain.42 43

Patient involvement

Although non-scientists from the surrounding communities served on many of the approving review boards at participating institutions, patients did not participate in the design of this study. All participants will receive the results of the study and a copy of the final manuscript from their study team by email or post.