FormalPara Key Points

High certainty evidence from 16 trials shows oral opioids provide small to no benefits above placebo for acute musculoskeletal pain.

Intravenous morphine likely provides large effects above placebo based on one study of sciatica.

Clinicians should consider that the benefits of prescribing an oral opioid are likely to be small to none, and weigh this against the well-known risks, even when using as a second- or third-line treatment.

1 Introduction

Musculoskeletal pain is responsible for the largest disability burden on individuals and the health care system, affecting approximately 1.7 billion people worldwide in 2019 [1]. Many people present to health care services seeking treatment for acute musculoskeletal pain, whether from an injury, a new condition, or an acute flare of a chronic condition. These people are often treated with pain medicines, advice and education, exercise, manual therapy, and occasionally surgery [2].

Opioids are recommended in guidelines for acute musculoskeletal pain when delivered in small doses over short periods of time [3]. Opioids are used frequently for musculoskeletal pain. For example, one study found that two-thirds of adults presenting to the emergency department with back pain receive an opioid medicine [4]. In paediatric populations, the proportion being prescribed opioids for musculoskeletal pain is less than adults but is still substantial (11% to 22%) [5]. Recommendations to prescribe opioids for acute pain, albeit cautiously, persist despite uncertain evidence of the benefits, and the known harms of opioid use (constipation, dizziness, nausea, somnolence, long-term use, dependence, overdose, etc.). Previous reviews of chronic musculoskeletal pain report that opioids typically provide small or no effect on pain and function compared to placebo [6, 7], and that there were high rates of withdrawal from the studies due to adverse events (AEs) or lack of effect [6]. However, it is not clear if this is the same for acute musculoskeletal pain. One previous review examined opioids for acute musculoskeletal pain [8]; however, they excluded the key diagnosis of low back pain, which is the most prevalent musculoskeletal pain condition [1]. Another examined the efficacy of opioids for musculoskeletal pain in the emergency department (assumed to be acute pain) and found a statistically but not clinically significant benefit of opioids over placebo [9], but the follow-up times in this review were very short (most in the 30- to 120-minute range) and therefore not entirely relevant to other settings like primary care, where the goals are for pain relief for longer periods. A recent placebo-controlled trial of orally administered opioids for acute spinal pain found no or negative effects [10] but has not been systematically reviewed alongside all available evidence.

In order for clinicians to make informed decisions about prescribing opioids for acute musculoskeletal pain, such as fractures, sprains, dislocations, aches and other injuries, we need an unbiased estimate of the benefits and harms of opioids for this population. The gold standard way to measure this is to systematically review randomised trials that compare an opioid to a placebo. Our research question is: are opioids (alone or in combination with other medicines) effective and safe for the management of acute musculoskeletal pain compared with placebo?

2 Methods

We prospectively registered this review on PROSPERO (CRD42021249346). We report this review as per the PRISMA 2020 statement recommendations [11]. A list of post-registration protocol deviations (with justifications) can be found in Appendix 1.

2.1 Identification and Selection of Trials

We searched electronic databases MEDLINE (via Ovid), EMBASE (via Ovid), Cochrane Central Register of Controlled Trials, and the World Health Organization International Clinical Trials Registry Platform (ICTRP) for eligible randomised controlled trials from their inception until February 22nd, 2023. We also did citation tracking from relevant systematic reviews and included trials.

The search strategy included words describing the study type (e.g., randomised controlled trial, based on the key words recommended by the Cochrane Back and Neck Group in their updated Search Strategies from January 2013) [12], and words describing the target population (e.g., acute musculoskeletal pain, back pain, fracture, sprain, etc.). Words describing the intervention (opioids, oxycodone, morphine, etc.) were determined by the research team by reviewing the strategies used in other similar reviews. See Appendix 2 for search strategies.

2.2 Eligibility Criteria

We included randomised controlled trials of opioids (alone or in combination) versus placebo, or like regimen minus opioids (e.g., non-steroidal anti-inflammatory drugs [NSAIDs] + opioid vs NSAID alone) where blinding was maintained, for patients with acute musculoskeletal pain. Acute pain is defined as pain that has persisted for equal to or less than 12 weeks. Acute musculoskeletal pain was defined as any pain, sprain or ache of the muscles, ligaments, tendons, bones and joints (including sciatica). This included people with a fracture and were waiting on conservative treatment or surgery. We excluded studies enrolling people in the immediate post-operative phase as this was likely conflated by pain from the incision and trauma of surgery. We included acute flares of chronic pain conditions such as osteoarthritis, but pooled them separately. We excluded studies of people with pregnancy-related musculoskeletal pain. Studies with participants of any age were included. Studies with mixed samples of acute and chronic musculoskeletal pain were included if subgroup results for the acute participants were available. Studies with participants recruited from any setting were included (emergency departments, primary care, etc.).

2.3 Data Extraction

All records identified in the initial search were screened by two independent reviewers for relevance by their title first, and abstract then full text if it was still unclear. Two independent reviewers extracted the following data using a standardised data extraction form: bibliometric data, study characteristics, characteristics of included participants, outcome measures and outcomes. Any disagreements between the two reviewers were resolved by consensus first, then by arbitration by a third, independent reviewer.

2.4 Primary Outcomes

Our dual primary outcomes were pain intensity measured by self-report (e.g., visual analogue scale [VAS]), and disability measured by self-report (e.g., Roland Morris Disability Questionnaire). Our secondary outcome was number (%) of patients reporting an AE and nature of the AE.

We grouped outcomes by the timing of follow-up: immediate term (≤ 24 h or closest to 12 h); short term (> 24 h to 7 days or closest to 4 days); intermediate term (> 7 days to 12 weeks or closest to 6 weeks; and long term (> 12 weeks or closest to 6 months). Our pre-specified primary timepoint was the immediate term.

2.5 Risk of Bias (Certainty) Assessment of the Outcome Measures

We assessed risk of bias using the Cochrane risk of bias 1 tool [13]. A study with an overall low risk of bias was defined as being ‘low risk’ in at least three of the six domains (domains listed in Appendix 7). We modelled this from a high-quality systematic review of a similar topic [14]. Studies with an ‘unclear’ or ‘high risk’ rating in any three or more of six domains were considered overall at high risk of bias.

We rated the certainty of the evidence for each outcome measure using the Grading Recommendations Assessment, Development an Evaluation (GRADE) framework. The domains evaluated by the GRADE framework included study design, risk of bias, inconsistency, indirectness, imprecision and publication bias. See Appendix 4 for further information about how we graded each domain. When reporting the findings, we use the labels “high certainty”, “moderate certainty”, “low certainty” and “very low certainty”. When discussing the findings, we use the terms recommended by the GRADE Working Group for communicating the findings of systematic reviews (e.g., “probably” when discussing moderate certainty evidence, “may” when discussing low certainty evidence, “unclear” when discussing very low certainty evidence) [15].

2.6 Strategy for Data Synthesis

Continuous pain severity and disability outcome measures were converted to a 0–100 scale to improve the comparability of findings between trials. We calculated the mean differences along with 95% confidence intervals (CI) between opioids versus control using a random effects model. Daily morphine milligram equivalent (MME) was calculated by first determining the total daily dose (stated dose for single-dose studies, or by multiplying the stated dose by the number of doses per day for multiple-dose studies) and then multiplying the daily total by the MME multiplier as per the Australian and New Zealand College of Anaesthetists & Faculty of Pain Medicine (ANZCA) opioid dose equivalence calculation table [16]. For studies that described dosing per body weight, we calculated the daily dose based on the mean body weight reported in the trial, or if the trial did not report participant body weight, we used the average body weight reported in a trial with a similar population, or where no similar trial existed, we used a weight of 70 kg for an average adult. When dosage was based on weight, we calculated the daily MME using the group mean weight of study participants. When prescribed dosing was described as a range (e.g., 1–2 tablets every 4–6 h) and no average number of daily tablets were reported, we calculated MME based on the highest allowable dose.

We separated studies according to the age of the population; paediatric (< 18 years) and adults (≥18 years). In the main analysis, we separated studies by route of administration (oral or intravenous [IV]). Effect size was considered small if 5–10 points on a 0–100 scale, moderate if > 10 to 20 points, and large if ≥ 20 points (based on the American College of Physicians guidelines for low back pain) [17]. Dichotomous outcomes (AEs) were calculated as risk differences (absolute) and risk ratios. To avoid duplication error in meta-analysis, we split the sample sizes of groups that were acting as a control to more than one comparator. We used Comprehensive Meta-Analysis Version 4 software for all analyses [18].

2.7 Exploratory Analyses

We conducted post hoc exploratory analyses on the primary outcome (oral opioids for pain in the immediate term) to explore reasons for statistical heterogeneity. We performed subgroup analyses per condition group, per opioid type, and opioid composition (opioids alone vs opioids in combination with simple analgesics). We also performed meta-regression on daily MME against pain outcomes, and a sensitivity analyses removing studies at high risk of bias. We also explored reasons for heterogeneity in the harms outcome by performing subgroup analysis per opioid type and opioid composition, meta regression of total daily MME on risk of harm and sensitivity analysis removing studies at high risk of bias. We performed a post hoc sensitivity analysis on our primary analysis only, excluding one study that examined acute-on-chronic populations.

3 Results

The search retrieved 15,144 articles, 17 of which (reporting 17 studies) were found to be eligible for inclusion [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36]. All studies were of parallel design. Four studies were not included in the meta-analysis as they provided outcomes in formats that were not able to be pooled. See Figure 1 for PRISMA flow diagram. See Appendix 5 for a list of excluded studies and reasons. See Table 1 for study characteristics. See Appendix 6 for individual results. The studies were from 6 countries (USA = 11, Australia = 2, Canada = 2, China = 1, New Zealand = 1, Turkey = 1).

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 flow diagram

Full size image
Table 1 Study characteristics
Full size table

3.1 Participants

There were 14 adult studies with a total of 3124 participants. The mean age of participants across studies was 46.1 years (SD = 12.3, n = 9 studies). The lowest study mean age was 29.9 [19] and the highest study mean was 69.6 [35]. The mean proportion of female participants across studies was 51.4% (n = 13 studies). The mean duration of pain upon admission to the study ranged from 120 min to up to 12 weeks. The mean pain intensity of participants at baseline was 69 out of 100 and ranged from 56 to 83. Four studies were categorised as ‘mixed presentations’ [22,23,24, 37], five studies were categorised as ‘spinal pain including sciatica’ [10, 20, 27, 31, 33, 35], two studies were categorised as ‘an acute flare of osteoarthritis’ [32, 34], and two studies were categorised as ‘soft tissue injury’ [19, 25]. Three of the 14 studies investigated populations with acute flares of chronic pain conditions [31, 32, 34]. All studies investigated oral opioids except for one, which investigated an intravenous opioid [33].

There were 3 paediatric studies enrolling a total of 427 participants [26, 28, 29] [Table 1]. The mean age of participants across studies was 11.5 years (SD = 2.9). The mean proportion of female participants across studies was 42.2%. The mean duration of pain upon admission to the study was not reported numerically but was defined as acute [28], < 24 h [26], and < 72 h [29]. The mean pain intensity of participants at baseline was 61.6/100 and ranged from 59 to 65. Two studies were categorised as ‘mixed presentations’ (combining fractures, sprains, contusions, etc.) [28, 29], and one study was categorised as ‘bony injuries’ [26]. All three studies examined acute pain conditions (not acute flares of chronic conditions).

3.2 Interventions

Adult studies tested seven types of opioid analgesics, with 5 studies with multiple arms testing different opioids (all orally administered unless stated: oxycodone = 6 trials, tramadol = 5 trials, codeine = 3 studies, butorphanol = 1 study, hydrocodone = 1 study, IV morphine = 1 study, and tapentadol = 1 study). Six studies delivered the opioid intervention as a single ingredient and eight used a combination of an opioid with a non-opioid. There were two types of controls (placebo alone = 9 and like regimen minus opioids where blinding was maintained = 5). Five trials had more than two arms [22, 24, 25, 33, 35]. Ten of 14 studies followed-up participants for longer than 1 day, and follow-up ranged from 30 min to 12 months in duration.

Paediatric studies tested three types of oral opioid analgesics (oxycodone, morphine and codeine). All three studies delivered the opioid orally and in combination with an NSAID. All three controls were NSAIDs with placebo. All three studies followed participants up for 120 min.

3.3 Risk of Bias

All three paediatric studies were rated as low risk of bias [26, 28, 29]. Eight of the 14 adult studies were scored as being at low risk of bias [10, 19, 20, 23, 25, 26, 33, 34]. All other studies had at least three domains that were high or uncertain risk of bias (see Appendix 7 for individual domain ratings for each study). The most frequent reasons for down-grading were high loss to follow-up and lack of prospective registration to confirm there was no selective reporting.

3.4 Outcomes

3.4.1 Pain

For adults, oral opioids provided a small effect compared to placebo at our primary timepoint (immediate term) (MD − 8.8, 95% CI − 12.0 to − 5.6, n = 1385, 7 studies, high certainty, includes studies testing codeine, immediate-release oxycodone, modified-release oxycodone, tramadol, and hydrocodone) (Fig. 2, Table 2). At our secondary timepoints, opioids provided a small effect in the short term (MD − 9.2, 95% C − 13.9 to − 4.4, n = 1676, 5 studies, moderate certainty, includes studies testing tramadol, butorphanol, hydrocodone and modified-release oxycodone), no effect in the intermediate term (MD − 9.0, 95% CI − 30.4 to 12.4, n = 305, 3 studies, moderate certainty, studies testing modified-release oxycodone and tramadol), and a small negative effect in the long term (6 months) (MD 8.0, 95% CI 2.6–13.4, n = 247, 1 study, moderate certainty, study testing modified-release oxycodone) [Appendix 8].

Fig. 2
figure 2

Pain outcomes for oral opioids at primary (immediate) timepoint, sorted by daily morphine milligram equivalent (MME) dose. Graudins a—oxycodone, Graudins b—codeine, Hewitt a—tramadol, Hewitt b—hydrocodone, Gilbert a—butorphanol tartrate 4 mg, Gilbert b—butorphanol 8 mg, Gilbert c—codeine

Full size image
Table 2 Outcomes and GRADE ratings
Full size table

A single study investigated single dose of IV morphine with a daily MME of 21.0 and found moderate certainty evidence of a large effect (MD − 42.5, 95% CI − 49.9 to − 35.1, n = 197) for sciatica pain at the immediate timepoint.

For paediatric patients, there is moderate certainty evidence that opioids did not provide greater pain relief compared with placebo at our primary timepoint (immediate term) (MD 6.1, 95% CI − 1.7 to 13.8, n = 236, 3 studies, moderate certainty, studies testing morphine, oxycodone and codeine) (Fig. 2, Table 2). There were no paediatric studies providing evidence for the pain outcome at our secondary timepoints of short-, intermediate-, and long-term.

3.4.2 Pain Exploratory Analyses

Exploratory subgroup analyses were conducted on the studies investigating oral opioids at the primary timepoint (immediate) to explore effect modifiers [Appendix 9]. When separated by condition group, opioids had a small effect on mixed musculoskeletal pain (7 comparisons, n = 397, I2 = 0%), and soft tissue injuries (2 comparisons, n = 595, I2 = 0%) in the immediate term. There may be no effect on spinal pain (2 comparisons, n = 393, I2 = 87%); however, this estimate is of low certainty and the heterogeneity remains very high. When separated by opioid type, all oral opioids had small-to-moderate effects except hydrocodone which showed no effect in a single study of ankle sprains. Studies investigating opioids alone had a moderate effect compared to studies investigating opioids in combination with simple analgesics which had a small effect, but the actual difference between the two mean estimates was only 4 points (7.7 vs 11.9). Meta-regression of the effect of daily MME on pain scores showed that MME was not a significant treatment effect modifier (p = 0.224) [Appendix 9]. A sensitivity analysis removing all studies at high risk of bias found a smaller effect compared to the main analysis (MD −4.6, 95% CI −8.7 to −0.4, n = 488, 3 studies, high certainty).

We performed a post hoc sensitivity analysis on our primary analysis only, excluding one study that examined acute-on-chronic populations. Of the 3 included studies that examined acute-on-chronic conditions [31, 32, 34], only 1 [31] reported a result relevant to our primary timepoint (immediate term). The pooled estimate excluding Ma 2008 [31] was not substantially different from the primary analysis (MD − 7.4, 95% CI − 10.3 to − 4.4, n = 1269, 10 comparisons from 6 studies, I2 = 11%, high certainty). The individual result from Ma 2008 [31] was more favourable but of low certainty (MD − 14.1, 95% CI − 20.0 to − 8.2, n = 116, testing modified-release oxycodone for acute-on-chronic neck pain).

3.4.3 Disability

For adults, there was no difference between opioid and placebo groups at the primary timepoint (immediate term) (MD − 6.2, 95% CI − 17.8 to 5.4, n = 208, 2 studies, very low certainty) (Fig. 3). At our secondary timepoints, opioids provided no effect in the short term (MD − 5.8, 95% CI − 13.7 to 2.1, n = 416, 2 studies, moderate certainty), intermediate term (MD − 7.5, 95% CI − 30.7 to 15.7, n = 478, 3 studies, low certainty) or the long term (MD −15.0, 95% CI − 51.1 to 21.2, n = 208, 2 studies, very low certainty). No paediatric studies reported disability.

Fig. 3
figure 3

Adult disability outcomes at primary (immediate) timepoint, as well as secondary timepoints (short and intermediate), sorted by daily morphine milligram equivalent (MME) dose in adults. Gilbert a—butorphanol tartrate 4 mg, Gilbert b—butorphanol 8 mg, Gilbert c—codeine

Full size image

There were insufficient studies to conduct subgroup analysis or meta-regression on the disability outcome.

3.4.4 Harms

For adults, oral opioid groups were at more risk of harm compared to placebo groups (RD 14.3%, 95% CI 8.3–20.4%, n = 2649, 11 studies, very low certainty) [Appendix 10]. The risk ratio was 1.69 (95% CI 1.35 to 2.09). A single study investigating IV opioids found no difference (RD 4%, 95% CI − 0.3 to 8.3%) (RR 8.9, 95% CI 0.5–163.3).

For paediatrics, there was no evidence for a difference in risk of experiencing an AE, although there was a trend towards opioid groups being more at risk of harms (risk difference [RD] 10.4%, 95% CI [− 0.6 to 21.4%], n = 393, 3 studies, low certainty) (Fig. 3). The risk ratio was 2.8 (95% CI 1.5–53). One of the three studies is unclear about whether the reported harms are number of events or number of people reporting at least one AE. Excluding this study, the pooled risk difference for the remaining two studies (which state number of people reporting at least one AE) was not substantially different (RD 8.5%, 95% CI − 3.6 to 20.6%, n = 349, low certainty).

The reporting of harms was too inconsistent to analyse by the nature of the harm. Studies that identified the nature of the harm reported common opioid-related AEs, including somnolence, dizziness, vomiting and confusion.

3.4.5 Harms Exploratory Analyses

Subgroup analyses by opioid type did not reduce the statistical heterogeneity [Appendix 11]. When analysing by opioids alone (not in combination with simple analgesics), the pooled outcome estimate did not change substantially but I2 value reduced to 32%. Sensitivity analysis removing all studies at high risk of bias did not substantially change the estimate of risk difference, but it reduced I2 to 34%. Total daily MME was not a significant effect modifier for risk of harm. The risk of harm increased by 1.7% for every 10 MME increase in daily dose (p = 0.07) [Appendix 12].

4 Discussion

4.1 Summary of Findings

In adult studies, oral opioids provided a small effect compared to placebo for pain (high certainty) in the immediate term. There was low to very low certainty evidence of small benefits for pain in the short and intermediate term for musculoskeletal pain. Only one study reported outcomes in the long term and found a small negative effect of modified-release oxycodone for spinal pain. In contrast, one trial evaluated IV morphine and found moderate certainty evidence of a large effect on sciatica pain at immediate-term follow-up. When separated by condition in further exploratory analysis of the primary analysis (immediate term), the result was reasonably consistent across all condition categories. There was no effect for disability in the short term (moderate certainty), or immediate and long term (both very low certainty). We found very low certainty evidence of a 14% absolute increase in risk for opioids groups in the adult studies, where the maximum follow-up was 12 months. Meta regression showed that increased daily doses of opioids (MMEs) were not associated with greater effects on pain scores.

In paediatric studies, there was no difference between oral opioids and placebo in reducing pain in the immediate term for musculoskeletal pain (fractures or soft tissue injuries) (moderate certainty). There were no data for pain or disability for timepoints beyond 24 h. We found no difference in risk of experiencing an AE in paediatric studies (possibly due to maximum follow-up time being 120 min). For both adults and paediatric populations, clinicians should carefully consider that the benefits may be small-to-none, and that the harms are uncertain, when making decisions about using opioids to treat acute musculoskeletal pain.

4.2 Strengths and Limitations

This is the first systematic review of opioids versus placebo for acute musculoskeletal pain which included all musculoskeletal conditions (including low back pain). These findings provide up-to-date, clinically relevant information to clinicians and policy makers. Four out of five of the pain outcome timepoints (including oral and IV) had moderate or high certainty evidence. A limitation is the lack of data to inform paediatric outcomes. All but one disability outcome was of low or very low certainty, and AE outcomes were both low certainty. Also, more than half of the included studies were at high risk of bias, mainly due to unclear/poor reporting of randomisation and allocation concealment practices, and high rates of losses to follow-up. There was also a low rate of prospective registration, meaning we could not determine whether there was selective reporting. Although, sensitivity analyses confirmed these studies at high risk of bias were not inflating the results.

4.3 Interpretation

In addition to this evidence for opioids versus placebo, recent trials and reviews comparing opioids with non-opioid analgesics show that oral opioids do not provide superior pain-relieving effects in acute pain and are associated with increased AEs. A review of analgesics for musculoskeletal pain in the emergency department found that NSAIDs had equal benefits to opioids (including oral, IV and intramuscular routes), but less risk of harm [9]. The same review found that paracetamol was only slightly less effective than opioids, again with less risk of harm, and may also be considered in place of an opioid [9]. Another review of oral opioids for post-surgical pain (prescribed at discharge) found there were no benefits over some non-opioid analgesics such as NSAIDs [38]. These results are likely relevant to those with acute musculoskeletal pain in any setting, therefore clinicians may consider NSAIDs before opioids if there are no contraindications, even when rapid pain relief is required.

More studies are needed to properly assess whether opioids have a benefit above placebo and/or non-opioid analgesics in paediatric populations at all timepoints. Although the rates of prescribing in paediatric populations are lower than that of adult populations, there is some concern (based on limited data) that a prescription of an opioid during adolescent years may increase the risk of opioid misuse by 33% in later years [5]. It is important that opioids are only prescribed to paediatric populations when there is genuine clinical benefit to avoid contributing to the opioid overuse problems. Further investigation into whether people (paediatrics and adults) receiving an opioid are at higher of risk of ongoing use, misuse, and dependence in the long term compared to groups who were not prescribed an opioid for their acute pain would be beneficial. While we did not set out to collect these outcomes, we note that only one study reported them, finding that the opioid group was at higher risk of misuse at 12-months after a short course of opioids (up to 6 weeks) [10]. Another aspect worthy of investigation is whether the rates of prescription of other analgesics such as pregabalin or ketamine, which may be equally as harmful, are increasing in place of opioids.

5 Conclusion

There was moderate certainty evidence of no effect of oral opioids in paediatric populations for acute musculoskeletal pain in the immediate term, and no data at other timepoints. In adult populations, there was low to high certainty evidence that oral opioids provide small benefits in the shorter term (≤ 7 days) but not in the longer term. Intravenous opioids provided large effects on pain in the immediate term. There was very low certainty evidence of a 14% absolute higher risk of experiencing an AE in adults. Clinicians should carefully weigh up the short-term small-to-no benefits of oral opioids (low-to-moderate certainty) against the uncertain evidence on safety.