INTRODUCTION

Low back pain (LBP) is one of the most prevalent health problems worldwide and is estimated to affect 60%–70% of people in industrialized countries during their lifetime.1 More than 2.5 million patients present to U.S. emergency departments (EDs) annually with LBP.2 A total of 22%–46% of ED patients with acute, new-onset LBP report persistent moderate or severe pain 1 week after the ED visit.3-6 Nonsteroidal anti-inflammatory drugs (NSAIDs) are an effective treatment of acute LBP, although their impact is only modest. One high-quality meta-analysis reported that on average, the pain relief afforded by NSAIDs is not clinically relevant.7 Although NSAIDs are recommended as first-line pharmacological treatment for LBP, there is currently no evidence to recommend one NSAID over another and selection of an NSAID for treatment is often based on provider experience.

Given the uncertainty surrounding the choice of NSAID, we designed a clinical trial to compare ibuprofen, ketorolac, and diclofenac for the treatment of acute, nontraumatic, and nonradicular LBP. Based on our unpublished clinical experience, we anticipated that ketorolac would be more effective than the other two medications. We tested the following hypothesis: A daily regimen of ketorolac would provide greater relief of LBP functional impairment than ibuprofen or diclofenac 5 days after an ED visit, as measured by the Roland-Morris Disability Questionnaire (RMDQ), a 24-item questionnaire that surveys the impact of LBP on one's daily life.8

METHODS Overview

This was a randomized, three-armed, double-blind, comparative effectiveness study, in which we enrolled participants during an ED visit for musculoskeletal LBP and followed them by telephone 2 and 5 days later. Participants received a 5-day supply of ibuprofen, ketorolac, or diclofenac. All participants provided written informed consent. This study was reviewed and approved by the Albert Einstein College of Medicine Institutional Review Board. It was registered online at http://clinicaltrials.gov (NCT03861611).

Subject selection

Our goal was to include in this study a broad representation of patients with musculoskeletal back pain who were likely to respond to the investigational medications. We included patients who presented to the ED primarily for management of LBP, defined as pain originating between the lower border of the scapulae and the upper gluteal folds. We excluded patients with flank pain (pain originating from tissues lateral to the paraspinal muscles). We included patients aged 18–65 who were to be discharged home from the ED. Patients were excluded if the pain radiated below the gluteal folds in a radicular pattern, if the pain lasted >2 weeks (336 h), if they had a history of LBP occurring once per month or more frequently, or if they experienced substantial and direct trauma to the back within the previous month. Participants were required to have functionally impairing LBP, as determined by a score of >5 on the baseline RMDQ (Data Supplement S1, Appendices S1 and S2, available as supporting information in the online version of this paper, which is available at http://onlinelibrary.wiley.com/doi/10.1111/acem.14321/full), so we obtained baseline RMDQ scores on all study participants. We excluded patients who were not available for follow-up, were pregnant, had a chronic pain syndrome as evidenced use of any analgesic medication on a daily or near-daily basis, or were unable to take the medication.

Intervention

Participants in arm 1 received 600 mg of ibuprofen, orally, every 8 h as needed. Participants in arm 2 received 10 mg of ketorolac, orally, every 8 h as needed. Participants in arm 3 received 50 mg of diclofenac, orally, every 8 h as needed. All study participants were dispensed 15 tablets of the investigational medication. Research personnel provided each patient with a 15-min educational intervention. This was based on NIAMS's “Handout on Health: Back Pain” information webpage (http://www.niams.nih.gov/Health_Info/Back_Pain/default.asp). No other medications or prescriptions were given to the patients at the time of discharge.

The pharmacist performed randomization in blocks of six based on a sequence generated at http://randomization.com. The investigational medication was masked by placing tablets into identical capsules, which were packed with scant amounts of lactose and sealed. The masking took place in a secure location inaccessible to ED personnel. Thus, researchers, outcome assessors, all clinical staff, and patients were blinded. We presented patients with one bottle of investigational medication labeled with dosing instructions.

Outcome measures

All data were collected by salaried research associates, who are fluent in English and Spanish. We assessed outcomes using the following measures: (1) RMDQ (reproduced in the Appendices S1 and S2). This 24-item LBP functional scale with good psychometric properties is recommended for use in LBP research.9 Its yes/no format is amenable to telephone follow-up. (2) Study participants described their worst back pain in the previous 24 h using an ordinal pain scale (severe, moderate, mild, or none). (3) We assessed LBP frequency by asking: “Over the last 24 h, how often were you in pain? (not at all, rarely, sometimes, usually, always). (4) We assessed medication requirements by asking “What medications did you use to treat your LBP in the previous 24 h?”

The primary outcome was the change in RMDQ score between the baseline ED visit and the 5 day follow-up (Roland-Morrisbaseline – Roland-Morrisday 5). Secondary outcomes were as follows: (1) improvement in RMDQ between baseline and 2 days, (2) worst pain intensity scores at 2 and 5 days, (3) pain frequency at 2 and 5 days, and (4) medication requirements at 2 and 5 days. The following outcomes were listed in the study protocol but not registered online: (1) absolute RMDQ scores at 5 days and (2) medication side effects. We assessed medication side effects by asking all participants, “Did you have any side effects from the medications you've been taking for LBP?” Those who answered affirmatively were asked to provide detailed information. We also asked all participants, “Did the medication irritate your stomach?” and asked them to respond with “no,” “a little,” or “a lot.”

Data analysis

We report baseline characteristics as means with SD, medians with interquartile range (IQR), or absolute numbers with percentages, as appropriate. We performed an intention-to-treat analysis among all patients for whom primary outcome data were available. For the primary outcome, we looked for differences between groups using a one-way ANOVA after assuring that the data grossly distributed parametrically using a normal Q–Q plot. For between-group comparisons of improvement in RMDQ, we report mean difference with 95% CI. We also report median improvement in RMDQ with IQR to demonstrate how each quartile in each study arm faired. For ordinal secondary data, we first grouped outcomes into clinically meaningful categories that have been used previously in ED-based LBP research.3-6, 10, 11 For worst pain intensity this was none/mild versus moderate/severe. For pain frequency, this was never/rarely/sometimes versus frequency/always. For analysis of these ordinal variables, we report the absolute risk reduction with 95% CI or chi-square p-value. To help contextualize the RMDQ data, we also report how many participants achieved a minimum clinically important difference (MCID) and compare these rates among the study arms.

Sample size calculation

We based assumptions on previous data.3 The mean improvement in RMDQ among those who receive an NSAID alone was 10.2. The SD was 8.9. A widely accepted clinically important improvement of five points on the RMDQ would require those randomized to active medication to demonstrate a mean improvement of 15.2 on the RMDQ.12 Using a standard alpha of 0.05 and a beta of 0.20, we determined the need for 60 subjects in each arm. To account for protocol violations and patients lost to follow-up (our typical lost-to-follow-up rate is 5%–10%) and to ensure sufficient power for the per-protocol analysis (in our previous ED-based LBP studies, up to one-third of enrolled patients have not used the medication more than once), we intended to enroll 66 patients in each arm (total n = 198).

RESULTS

Enrollment commenced in July 2019, was halted between March and June of 2020 due to COVID-19, and concluded in January 2021. During this time, 868 patients were screened for participation and 198 were enrolled, 66 in each of the three arms of the study (Figure 1). Baseline characteristics of these patients are presented in Table 1. In general, these study participants reported a fair amount of baseline impairment, as evidenced by median RMDQ scores of 19 or 20 (Table 1).

Consort flow diagram. LBP, low back pain; RMDQ, Roland-Morris Disability Questionnaire

TABLE 1. Baseline characteristics Variable Ibuprofen (n = 66) Ketorolac (n = 66) Diclofenac (n = 66) Age (years), mean (±SD) 40.3 (±10.4) 39.0 (±11.2) 39.9 (±10.1) Female (%) 24 (36%) 25 (38%) 25 (38%) Back pain duration (h), median (IQR) 72 (24–96) 48 (24–96) 72 (38–101) Baseline RMDQ, median (IQR) 20 (15–22) 19 (15–22) 19 (15–23) Frequency of previous back pain Never before 10 (15%) 19 (29%) 16 (24%) A few times before 37 (56%) 29 (44%) 36 (55%) A few times per year 19 (29%) 18 (27%) 14 (21%) Note RMDQ, Roland-Morris Disability Questionnaire, a 24-item assessment of LBP-related functional impairment on which 0 = no impairment and 24 = severe impairment. A score of at least 6 was required for entry into this study. Abbreviations: IQR, interquartile range; LBP, low back pain.

For the primary outcome, median (IQR) improvement in RMDQ score from initial ED visit to day 5 was as follows for each study arm: ibuprofen 11 (2–18), ketorolac 14 (5–18), and diclofenac 11 (4–19). The ANOVA p-value was 0.34. Between-group differences are presented in Table 2.

TABLE 2. Five-day outcomes Variable Ibuprofen (n = 66) Ketorolac (n = 66) Diclofenac (n = 66) Difference (95% CI) Ibuprofen vs. ketorolac Ibuprofen vs. diclofenac Ketorolac vs. diclofenac Improvement in RMDQ, mean (±SD), number of participants included in the analysis 9.4 (±9.5), 61 11.9 (±8.8), 59 10.9 (±9.4), 62 –2.5 (–5.8 to 0.8) –1.4a (–4.8 to 2.0) 1.1a (–2.2 to 4.4) Worst LBP in previous 24 hb None 20 (33%) 20 (34%) 20 (32%) 18% (2%–34%) 10% (–7% to 26%) 8% (–7% to 23%) Mild 17 (28%) 27 (46%) 25 (40%) Moderate 16 (27%) 11 (19%) 10 (16%) Severe 7 (12%) 1 (2%) 8 (13%) Missing 6 7 3 Frequency of LBP in previous 24 hc Never 16 (27%) 17 (29%) 21 (33%) 16% (1%–32%) 11% (–5% to 27%) 5% (–9% to 20%) Rarely 8 (13%) 14 (24%) 14 (22%) Sometimes 15 (25%) 17 (29%) 13 (21%) Frequently 14 (23%) 7 (12%) 6 (10%) Always 7 (12%) 4 (7%) 9 (14%) Missing 6 7 3 Took medication for LBP in previous 24 h No 25 (41%) 21 (36%) 27 (43%) 5% (–12% to 23%) 2% (–15% to 19%) 7% (–10% to 25%) Yes 36 (59%) 38 (64%) 36 (57%) Missing 5 7 3 Note RMDQ, Roland-Morris Disability Questionnaire, a 24-item assessment of LBP-related functional impairment on which 0 = no impairment and 24 = severe impairment. Abbreviation: LBP, low back pain. a Rounded. b We compared mild/none versus moderate/severe. c We compared never/rarely/sometimes versus frequently/always

Secondary outcomes are reported in Table 2 (5-day outcomes) and Table 3 (2-day outcomes). To help contextualize the clinical relevance of the difference reported in improvement in RMDQ between the ED visit and the 2-day follow-up between the ibuprofen arm and the ketorolac arm, we also calculated how many patients in each arm failed to experience at least a MCID of 5. This failure was reported by 27 of 62 (44%) of ibuprofen patients versus 15 of 60 (25%) ketorolac patients (the 95% CI for the difference of 19% was 2, 35%). Five days after the ED visit, a large minority of patients still reported substantial functional impairment on the RMDQ—in the ibuprofen arm, the median (IQR) RMDQ score was 3 (0–20), in the ketorolac arm it was 2 (0–10), and in the diclofenac arm it was 2 (0–22).

TABLE 3. Two-day outcomes Variable Ibuprofen (n = 62) Ketorolac (n = 62) Diclofenac (n = 62) Difference (95% CI) Ibuprofen vs. ketorolac Ibuprofen vs. diclofenac Ketorolac vs. diclofenac Improvement in RMDQ since ED visit, mean (±SD), number of participants included in the analysis 5.6 (±9.1), 62 9.9 (±8.5), 60 9.0 (±10.0), 64 –4.3 (–7.5 to –1.1) –3.4 (–6.8 to 0.0) 0.9 (–2.4 to 4.2) Worst LBP in previous 24 hb None 10 (16%) 8 (13%) 13 (21%) 15% (–2% to 33%) 15% (–2% to 32%) 0% (–17% to 18%) Mild 18 (29%) 29 (48%) 25 (40%) Moderate 23 (37%) 18 (30%) 18 (29%) Severe 11 (18%) 6 (10%) 7 (11%) Missing 4 5 3 Frequency of LBP in previous 24 hc Never 7 (11%) 6 (10%) 10 (16%) 21% (5% to 36%) 10% (–6% to 27%) 11% (–4% to 25%) Rarely 10 (16%) 18 (30%) 14 (22%) Sometimes 21 (34%) 26 (43%) 21 (33%) Frequently 9 (15%) 5 (8%) 7 (11%) Always 15 (24%) 6 (10%) 11 (17%) Missing 4 5 3 Took medication for LBP in previous 24 h No 7 (11%) 9 (15%) 6 (10%) 4% (–8% to 16%) 2%a (–9% to 12%) 5% (–6% to 17%) Yes 55 (89%) 51 (85%) 57 (91%) Missing 4 6 3 Note RMDQ, Roland-Morris Disability Questionnaire, a 24-item assessment of LBP-related functional impairment on which 0 = no impairment and 24 = severe impairment. Abbreviation: LBP, low back pain. a Rounded. b We compared mild/none versus moderate/severe. c We compared never/rarely/sometimes versus frequently/always.

No serious side effects attributable to the investigational medications were observed. One participant in the ibuprofen arm was admitted to the hospital for angioedema. This participant was taking an angiotensin-converting enzyme inhibitor, which was thought to be the more likely cause of this clinical presentation. Reporting of any side effects were similar in all arms: ibuprofen eight of 63 (13%), ketorolac six of 61 (10%), and diclofenac eight of 64 (13%). The most common side effect was drowsiness, which was reported by three participants in each arm (Table 4). When specifically asked “Did the medication irritate your stomach?” a negative response was reported by 46 of 62 (74%) in the ibuprofen arm, 58 of 61 (95%) in the ketorolac arm, and 58 of 64 (91%) in the diclofenac arm (p < 0.01).

TABLE 4. Adverse events Adverse event Ibuprofen (n = 66) Ketorolac (n = 66) Diclofenac (n = 66) Any adverse event report at any timea Yes 8 (13%) 6 (10%) 8 (13%) No 55 (87%) 55 (90%) 56 (88%) Missing 3 5 2 Stomach irritationb,c, b,c No 46 (74%) 58 (95%) 58 (91%) A little 13 (21%) 3 (5%) 4 (6%) A lot 3 (5%) 0 (0%) 2 (3%) Missing 4 5 2 Note One patient randomized to ibuprofen was admitted to the hospital for angioedema. There were no other serious adverse events. The following side effects were reported by more than one participant: drowsiness, by three participants in each study arm, and headache, by one ketorolac participant and one diclofenac participant. These additional side effects were reported by participants who received ibuprofen: thirst, bubbles in urine, flushing, and palpitations. These additional side effects were reported by participants who received ketorolac: a bruise and dizziness. These additional side effects were reported by patients who received diclofenac: elevated blood pressure, rash, and nausea. a Study participants were asked: Did you have any side effects from the medications you've been taking for low back pain? b Study participants were asked: Did the medication irritate your stomach? c The chi-square p-value for stomach irritation was <0.01. DISCUSSION

In this randomized, comparative effectiveness study conducted among patients with acute, nonradicular LBP, there was no statistically significant difference between the groups with regard to the primary outcome. However, we believe that merely describing this study as a negative study does not do the data justice. Among several outcomes relevant to patients, participants who received ketorolac fared better than those who received ibuprofen, with diclofenac falling in the middle. Sometimes these differences were statistically significant and sometimes they were not. Therefore, these data raise the possibility that ketorolac was more efficacious. With regard to the RMDQ functional scale, participants who received ketorolac reported greater improvement on the RMDQ scale than participants who received ibuprofen 2 days after the ED visit (4.3 points, 95% CI = 1.1–7.5 points) and 5 days after the ED visit (2.5 points, 95% CI = –0.8 to 5.8). Fewer participants who received ketorolac reported moderate and severe pain than participants who received ibuprofen 2 days after the ED visit (between-group difference = 15%, 95% CI = –2% to 33%) and 5 days after the ED visit (18%, 95% CI = 2%–34%). Fewer participants who received ketorolac experienced pain frequently or always compared with participants who received ibuprofen 2 days after the ED visit (between-group difference = 21%, 95% CI = 5%–36%) and 5 days after the ED visit (16%, 95% CI = 1%–32%). One possible explanation of these data is that ketorolac is superior and that our study was underpowered for some of these outcomes. A second, less likely possible explanation is that there truly is no difference between the medications with regard to clinically important outcomes. Participants who received diclofenac generally reported outcomes that were not statistically distinguishable from ketorolac or ibuprofen.

It is worthwhile discussing the clinical relevance of two statistically significant findings. At 2 days, the RMDQ score favored ketorolac over ibuprofen by 4.3 (95% CI = 1.1–7.5), a difference that was less than the MCID of 5 that we identified in the sample size calculation. However, it should be understood that biological phenomena cannot be explained by simple cut points. When one considers the distribution around this mean difference, it seems that the improvement in low back impairment experienced with ketorolac 2 days after the ED visit was more than that from ibuprofen. For some patients, this was clinically relevant and for others it was not. This is seen more clearly in our analysis of the frequency with which patients in each arm failed to improve by at least 5.0: in the ketorolac arm, 25% of participants failed to achieve this threshold versus 44% of participants in the ibuprofen alone arm corresponding to a number needed to treat of nearly five. With regard to the frequency of moderate or severe pain 5 days after ED discharge, the between-group difference favored ketorolac over ibuprofen by 18%, also corresponding to a number needed to treat of nearly five. The overall rates of adverse events were comparable but far more patients who received ibuprofen reported stomach irritation (number needed to harm of five).

Comparisons of NSAIDs in the literature for LBP have been limited by heterogeneous and lower-quality data. An updated 2020 Cochrane review on acute LBP found moderate evidence of no clinically relevant difference in the NSAIDs compared head to head with regard to pain intensity and disability, although these investigators only identified one RCT involving ketorolac.1 Direct comparison of our results to prior work is hindered by heterogeneity in study methodologies including different investigational medications, routes of administration, dosing, assessment methods, and biases including selection bias. Plapler et al.13 reported that ketorolac provides faster pain relief versus naproxen for LBP patients in an outpatient setting at 60 min but did not differ in RMDQ score or pain scale at day 5. In a dental model, Olmedo et al.14 found increased pain relief and decreased need for rescue medication for ketorolac over ketoprofen. Our work contributes to the existing literature supporting ketorolac as an appropriate treatment for acute LBP.

In this study and in previous ED-based work, a large proportion of patients continue to suffer from frequent, severe, and functionally disabling pain for many days after the ED visit—nearly one-third of our sample reported moderate or severe pain at 5-day follow-up and more than one-quarter reported substantial functional impairment. The addition of other classes of medication (opiates, muscle relaxers, acetaminophen, and benzodiazepines) to NSAIDs does not improve pain relief or functional outcomes among patients with acute LBP.3, 4, 6, 10, 11 It is not clear if recommending nonpharmacological therapy to ED patients will improve these outcomes.15 Therefore, the modest benefit that ketorolac might afford LBP patients is potentially important and, when applied on a population level, should have a noticeable impact on this highly prevalent disease.

LIMITATIONS

Limitations of this study include the population in which it was conducted––our study was conducted in two urban EDs in one of the most socioeconomically depressed counties in the United States. Whether or not these data are generalizable to other populations is uncertain. Also, the 600-mg dose of ibuprofen used in this study may have increased the risk of side effects without necessarily offering an opportunity for benefit, although the optimal dose of ibuprofen for LBP has not been established.16 We made an error in the trial registration, describing the pain intensity outcome as worsening LBP after an ED visit rather than worst LBP after an ED visit. We did not assess baseline pain scores nor did we record which analgesics study participants received in the ED. Thus, it is unknown whether or not these variables impacted outcomes. And finally, this trial was underpowered for some of the differences we reported, thus raising the possibility that a larger trial would have reported more definitive results.

CONCLUSIONS

In conclusion, there were no important differences between groups with regard to the primary outcome. These data do not rule out that possibility that ketorolac results in better pain relief and less stomach irritation than ibuprofen.

AUTHOR CONTRIBUTIONS

Eddie Irizarry, Maha Salama, and Benjamin W. Friedman conceived the study concept and designed the trial. Andrew Restivo, Michelle Davitt, Carmen Feliciano, and Alexis Cortijo-Brown supervised the acquisition of the data. Benjamin W. Friedman provided statistical expertise on study design and analyzed the data. Eddie Irizarry drafted the manuscript and all authors contributed substantially to the critical revision of the manuscript for important intellectual content. Eddie Irizarry takes responsibility for the paper as a whole.

Filename Description acem14321-sup-0001-DataSupplementS1.pdfPDF document, 746.2 KB Supplementary Materials

Please note: The publisher is not responsible for the