Systemic corticosteroids for radicular and non‐radicular low back pain

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

To determine the benefits and harms of systemic corticosteroids compared with placebo or no systemic corticosteroid for patients with acute, subacute, or chronic radicular or non‐radicular low back pain.

Background

Low back pain is highly prevalent and can result in major adverse impacts on quality of life and function (Deyo 2006; Hoy 2012). Low back pain is associated with high direct costs and is a common reason for missed work or reduced productivity while at work, resulting in high indirect costs as well (Luo 2004; Martin 2008). A number of treatments are available for low back pain, including various pharmacological treatments, non‐invasive non‐pharmacological treatments, non‐surgical injection and interventional therapies, and surgical treatments.

Corticosteroids are a class of medications that are structurally similar to the human adrenal hormone cortisol, with potent anti‐inflammatory and immunosuppressant effects. For low back pain, corticosteroids can be administered directly into spinal structures through injections, as well as systemically, most commonly via the oral, intravenous, or intramuscular routes.

Systemic corticosteroids have been used for treatment of radicular low back pain for a long time. One of the earliest studies on systemic corticosteroids for low back pain, published in 1975, reported on 100 consecutive patients with radicular pain due to a herniated disc who were treated with a tapering course of dexamethasone (Green 1975). It found that all patients reported relief of pain within 24 to 48 hours, and 80% did not undergo surgery after 15 months of follow‐up. There was no non‐corticosteroid control group. Subsequently, a number of placebo‐controlled randomised trials have evaluated the effects of systemic corticosteroids on non‐radicular low back pain. Results have been inconsistent, with most trials reporting no clear benefits.

Clinical practice guidelines published in 2007 from the American Pain Society and American College of Physicians recommended against the use of corticosteroids for radicular or non‐radicular low back pain (Chou 2007). Nonetheless, 14.5% of patients with chronic low back pain who had seen a provider in the last year reported use of steroids in the past 30 days (Carey 2009), and 5% of patients evaluated in an emergency department for low back pain received corticosteroids (Friedman 2010).

Additional trials have recently been published on the benefits and harms of systemic corticosteroids for radicular and non‐radicular low back pain (Eskin 2014; Friedman 2008; Goldberg 2015; Holve 2008). Given the ongoing use of this class of medications for patients with low back pain, and discordance between clinical practice guidelines and clinical practice, our aim was to perform the first Cochrane Review on this topic, in order to provide up‐to‐date information on the benefits and harms of systematic corticosteroids for radicular and non‐radicular low back pain, compared with placebo.

Description of the condition

Low back pain is defined as pain or discomfort occurring between the lower posterior margin of the rib cage and the horizontal gluteal fold (Deyo 2014). The presence of radiating symptoms to the lower extremities characterises radicular low back pain; in this condition the leg symptoms are often more severe than the back pain (Chou 2007). Symptoms include pain or paraesthesias that occur in a dermatomal distribution corresponding to the affected nerve root, with or without motor deficits in the corresponding myotomal distribution. Radicular low back pain is most commonly due to a herniated disc, which occurs in about 4% to 5% of patients with low back pain and often presents acutely. Other causes of radicular low back pain include spinal stenosis and foraminal stenosis, often due to degenerative conditions within the spine, which tend to present more chronically. Symptoms of spinal stenosis often occur in the bilateral lower extremities and may be associated with neurogenic claudication (development of lower extremity symptoms when the patient assumes a standing position, which accentuates narrowing in the spinal canal).

Non‐radicular low back pain refers to low back pain without radicular symptoms. Some patients with non‐radicular low back pain have a clear pathoanatomic cause for their low back pain. However, non‐radicular back pain is most commonly non‐specific in the sense that it cannot be attributed to a specific spinal pathology or condition (Jarvik 2002). Although degenerative findings such as degenerated discs, facet joint arthropathy, and bulging discs are common in patients with non‐radicular low back pain, such findings are only weakly associated with the presence and severity of low back symptoms (Chou 2011). Serious conditions, such as cancer, fracture, inflammatory arthritis, and infection, are uncommon causes of low back pain.

Low back pain is also often classified according to duration as acute (less than four weeks), subacute (4 to 12 weeks), or chronic (longer than 12 weeks) (Chou 2007). The natural history in most patients with acute radicular or non‐radicular low back pain is for there to be marked improvement in symptoms over days to weeks, whereas there is often little improvement over time in patients with chronic symptoms (da C Menezes Costa 2012); subacute back pain represents a transition state between acute and chronic low back pain.

Description of the intervention

The term 'corticosteroids' generally refers to hormones that resemble cortisol and are associated with stronger glucocorticoid (anti‐inflammatory) effects relative to mineralocorticoid (sodium retention) effects (van der Laan 2008). A number of corticosteroids are available, including prednisone, prednisolone, methylprednisolone, dexamethasone, betamethasone, and hydrocortisone. Corticosteroids vary in their relative potency, duration of action, and glucocorticoid relative to mineralocorticoid effects. Corticosteroids can be administered orally as well as parenterally (e.g. intravenously or intramuscularly). Doses of corticosteroids for radicular and non‐radicular low back pain are not well standardised and vary widely across studies, ranging from a single dose to various tapering regimens in which the dose is reduced over days to weeks. Generic forms are available for all of the corticosteroids listed above. Hyperglycaemia is a common adverse event associated with short‐term corticosteroid use. Corticosteroids are also associated with a number of other less common or longer‐term adverse events, including infection, osteoporosis, psychiatric effects, and adrenal suppression (McDonough 2008).

How the intervention might work

The most common cause of radicular low back pain is a herniated intervertebral disc (Ropper 2015). The pathophysiology of radiculopathy from a herniated disc is related to compression of the nerve, as well as the local release of inflammatory cytokines. The effects of corticosteroids are presumed to be related to their anti‐inflammatory effects, which may help reduce swelling and related compression on the affected nerve. For non‐radicular low back pain and spinal stenosis, the rationale for use of systemic corticosteroids is less clear (Eskin 2014), though it is possible that some patients may have an inflammatory component to their symptoms.

Why it is important to do this review

The use of systemic corticosteroids for low back pain remains relatively common despite recommendations against their use in clinical practice guidelines (Chou 2007). Randomised trials and a prior systematic review, Roncoroni 2011, are available on this topic, but some recently published randomised trials were not included in the prior systematic review (Eskin 2014; Goldberg 2015), including the largest trial published to date (Goldberg 2015). Also, the prior systematic review focused on radicular low back pain, only evaluated dichotomous outcomes, and did not stratify outcomes according to duration of follow‐up (e.g. shorter‐ versus longer‐term outcomes). Therefore, a new systemic review that addresses benefits and harms of systemic corticosteroids for radicular as well as non‐radicular low back pain, includes newer trials, evaluates continuous and dichotomous outcomes, and stratifies outcomes according to duration of follow‐up is needed to better inform clinicians, patients, and policy‐makers making decisions regarding their use.

Objectives

To determine the benefits and harms of systemic corticosteroids compared with placebo or no systemic corticosteroid for patients with acute, subacute, or chronic radicular or non‐radicular low back pain.

Methods

Criteria for considering studies for this review

Types of studies

We will restrict inclusion to randomised controlled trials. We will exclude controlled trials that were not randomised or used quasi‐random allocation (e.g. alternating allocation, allocation by day of week), because such studies tend to report biased estimates of treatment effect (Higgins 2011).

Types of participants

Inclusion criteria

• Trials that enrolled adult patients 18 years of age or older with acute, subacute, or chronic non‐radicular or radicular low back pain.

• Trials that recruited patients from primary care settings, specialty settings, or emergency department settings.

Exclusion criteria

• Trials that focused on patients with serious spinal pathology (e.g. cancer, fracture, cauda equina syndrome, inflammatory diseases) or in which such patients accounted for > 10% of patients.

• Trials that focused on treatment of pregnant women or in which pregnant women accounted for > 10% of participants.

• Trials in which patients received systemic corticosteroids for the purpose of reducing postoperative pain following lumbar spine surgery. We will also exclude trials in which > 10% of patients underwent spinal surgery within three months of enrolment.

Types of interventions

We will include studies that administered any dose or regimen of systemic corticosteroids (prednisone, prednisolone, methylprednisolone, dexamethasone, betamethasone, hydrocortisone) through the oral, intravenous, or intramuscular routes. We will exclude studies that evaluate epidural administration of corticosteroids.

Types of outcome measures

We will focus on patient‐centred clinical outcomes. We will not evaluate physiological, biomechanical (e.g. range of motion, measures of muscular strength), and laboratory parameters.

Primary outcomes

• Pain, measured as the percentage of patients with pain relief, based on improvements recorded using a visual analogue scale (VAS) or numerical rating scale.

• Pain, measured as mean improvement from baseline.

• Back‐specific disability (function), measured as the percentage of patients with a successful functional outcome on the Oswestry Disability Index, the Roland Morris Disability Questionnaire, or another scale for back‐specific disability.

• Back‐specific disability (function) measured as mean improvement from baseline.

Secondary outcomes

• Quality of life, as measured by mean improvement in the Short‐Form 36 or another validated quality of life measure.

• Percentage of patients with a successful composite outcome (e.g. improvement in pain and function).

• Percentage of patients with a successful global measure of improvement.

• Percentage of patients who underwent surgery.

• Percentage of patients who experienced any adverse event.

• Percentage of patients who experienced a serious adverse event.

• Percentage of patients who withdrew because of an adverse event.

• Percentage of patients with hyperglycaemia.

Search methods for identification of studies

Electronic searches

We will search for trials in the following databases from inception to the current date:

• the Cochrane Central Register of Controlled Trials (CENTRAL; latest issue) in the Cochrane Library (includes the Cochrane Back and Neck Group Specialized Register);

• MEDLINE via OvidSP;

• Embase via OvidSP;

• MEDLINE In‐Process & Other Non‐Indexed Citations via OvidSP;

• International Pharmaceutical Abstracts (IPA) via OvidSP; and

• Web of Science.

There will be no language or date restrictions. Search strategies will be developed using the methods of Cochrane Back and Neck (Furlan 2015), as well as the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). A draft strategy for MEDLINE can be found in Appendix 1. It will be translated as closely as possible across the other databases.

Searching other resources

We will check the reference lists of relevant articles for additional citations. We will also search the World Health Organization (WHO) International Clinical Trials Registry Platform (apps.who.int/trialsearch) and the US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov) for unpublished or ongoing studies.

Data collection and analysis

Selection of studies

Two authors will independently screen citations identified by the literature search after all members of the review team have pilot tested a sample of 100 citations to determine the inter‐rater reliability. We will modify inclusion/exclusion criteria if needed to improve reliability. We will resolve disagreements using final inclusion/exclusion criteria by discussion or through the arbitration of a third author.

Data extraction and management

Two authors will independently extract data using a standardised data extraction form. We will resolve discrepancies by discussion or through the arbitration of a third author. We will extract the following information:

• study characteristics (author, year of publication, language, country, clinical setting, funding source, number randomised and analysed, loss to follow‐up);

• patient characteristics (age, sex, race, duration and severity of symptoms, use of imaging for patient selection);

• intervention characteristics for the treatment (corticosteroid used, dose, duration, and cointerventions) and placebo;

• timing of follow‐up assessments, analysed as immediate (< 2 weeks since initiation of treatment), short‐term (two weeks to three months since initiation of treatment), intermediate term (> 3 months to < 12 month), and long‐term (≥ 12 months); and

• study results for each primary and secondary outcome at the specified time frames.

When there are multiple time points reported within a category, we will extract results closest to one week, eight weeks, and six months for immediate, short‐term, and intermediate‐term follow‐up, respectively. For long‐term follow‐up, we will extract the longest duration of follow‐up, as well as results closest to one year. We will convert daily and total corticosteroid doses to prednisone equivalents, using the calculator available at www.medcalc.com/steroid.html, which is based on the following conversion ratios (Chrousos 2014; Goodman 2011):

Corticosteroid	Approximate equivalent dose (mg)
Hydrocortisone	20
Methylprednisolone	4
Prednisolone	5
Prednisone	5
Betamethasone	0.67
Dexamethasone	0.80

Assessment of risk of bias in included studies

We will assess the risk of bias for each included study using the 'Risk of bias' assessment tool developed by Cochrane Back and Neck (Table 3; Table 4) (Furlan 2015). Two review authors will independently perform the 'Risk of bias' assessment and resolve disagreements through discussion or arbitration by a third author when consensus cannot be reached. We will score each of the 13 items of the 'Risk of bias' assessment as 'high', 'low', or 'unclear' risk. We will define a study with an overall low risk of bias as having low risk of bias on seven or more of the 13 items.

Table 1.

Sources of Risk of Bias

Bias Domain	Source of Bias	Possible Answers
Selection	(1) Was the method of randomization adequate?	Yes/No/Unsure
Selection	(2) Was the treatment allocation concealed?	Yes/No/Unsure
Performance	(3) Was the patient blinded to the intervention?	Yes/No/Unsure
Performance	(4) Was the care provider blinded to the intervention?	Yes/No/Unsure
Detection	(5) Was the outcome assessor blinded to the intervention?	Yes/No/Unsure
Attrition	(6) Was the drop‐out rate described and acceptable?	Yes/No/Unsure
Attrition	(7) Were all randomized participants analyzed in the group to which they were allocated?	Yes/No/Unsure
Reporting	(8) Are reports of the study free of suggestion of selective outcome reporting?	Yes/No/Unsure
Selection	(9) Were the groups similar at baseline regarding the most important prognostic indicators?	Yes/No/Unsure
Performance	(10) Were cointerventions avoided or similar?	Yes/No/Unsure
Performance	(11) Was the compliance acceptable in all groups?	Yes/No/Unsure
Detection	(12) Was the timing of the outcome assessment similar in all groups?	Yes/No/Unsure
Other	(13) Are other sources of potential bias unlikely?	Yes/No/Unsure

Furlan 2015

Table 2.

Criteria for a Judgment of ‘‘Yes’’ for the Sources of Risk of Bias

1	A random (unpredictable) assignment sequence. Examples of adequate methods are coin toss (for studies with 2 groups), rolling a dice (for studies with 2 or more groups), drawing of balls of different colours, drawing of ballots with the study group labels from a dark bag, computer‐generated random sequence, preordered sealed envelopes, sequentially‐ordered vials, telephone call to a central office, and preordered list of treatment assignments.Examples of inadequate methods are: alternation, birth date, social insurance/security number, date in which they are invited to participate in the study, and hospital registration number.
2	Assignment generated by an independent person not responsible for determining the eligibility of the patients. This person has no information about the persons included in the trial and has no influence on the assignment sequence or on the decision about eligibility of the patient.
3	Index and control groups are indistinguishable for the patients or if the success of blinding was tested among the patients and it was successful.
4	Index and control groups are indistinguishable for the care providers or if the success of blinding was tested among the care providers and it was successful.
5	Adequacy of blinding should be assessed for each primary outcome separately. This item should be scored ‘‘yes’’ if the success of blinding was tested among the outcome assessors and it was successful or: ‐for patient‐reported outcomes in which the patient is the outcome assessor (e.g., pain, disability): the blinding procedure is adequate for outcome assessors if participant blinding is scored ‘‘yes’’ ‐for outcome criteria assessed during scheduled visit and that supposes a contact between participants and outcome assessors (e.g., clinical examination): the blinding procedure is adequate if patients are blinded, and the treatment or adverse effects of the treatment cannot be noticed during clinical examination ‐for outcome criteria that do not suppose a contact with participants (e.g., radiography, magnetic resonance imaging): the blinding procedure is adequate if the treatment or adverse effects of the treatment cannot be noticed when assessing the main outcome ‐for outcome criteria that are clinical or therapeutic events that will be determined by the interaction between patients and care providers (e.g., cointerventions, hospitalization length, treatment failure), in which the care provider is the outcome assessor: the blinding procedure is adequate for outcome assessors if item ‘‘4’’ (caregivers) is scored ‘‘yes’’ ‐for outcome criteria that are assessed from data of the medical forms: the blinding procedure is adequate if the treatment or adverse effects of the treatment cannot be noticed on the extracted data
6	The number of participants who were included in the study but did not complete the observation period or were not included in the analysis must be described and reasons given. If the percentage of withdrawals and drop‐outs does not exceed 20% for short‐term follow‐up and 30% for long‐term follow‐up and does not lead to substantial bias a ‘‘yes’’ is scored. (N.B. these percentages are arbitrary, not supported by literature).
7	All randomized patients are reported/analyzed in the group they were allocated to by randomization for the most important moments of effect measurement (minus missing values) irrespective of noncompliance and cointerventions.
8	All the results from all prespecified outcomes have been adequately reported in the published report of the trial. This information is either obtained by comparing the protocol and the report, or in the absence of the protocol, assessing that the published report includes enough information to make this judgment.
9	Groups have to be similar at baseline regarding demographic factors, duration and severity of complaints, percentage of patients with neurological symptoms, and value of main outcome measure(s).
10	If there were no cointerventions or they were similar between the index and control groups.
11	The reviewer determines if the compliance with the interventions is acceptable, based on the reported intensity, duration, number and frequency of sessions for both the index intervention and control intervention(s). For example, physiotherapy treatment is usually administered for several sessions; therefore it is necessary to assess how many sessions each patient attended. For single‐session interventions (e.g., surgery), this item is irrelevant.
12	Timing of outcome assessment should be identical for all intervention groups and for all primary outcome measures.
13	Other types of biases. For example: ‐When the outcome measures were not valid. There should be evidence from a previous or present scientific study that the primary outcome can be considered valid in the context of the present. ‐Industry‐sponsored trials. The conflict of interest (COI) statement should explicitly state that the researchers have had full possession of the trial process from planning to reporting without funders with potential COI having any possibility to interfere in the process. If, for example, the statistical analyses have been done by a funder with a potential COI, usually ‘‘unsure’’ is scored.

Furlan 2015

Measures of treatment effect

For dichotomous outcomes, we will quantify treatment effects by the relative risk (RR) and 95% confidence interval. Dichotomous outcomes for pain relief and successful functional (back pain‐specific disability), composite, or global outcomes are likely to be variably defined and reported across trials. We will define "success" on dichotomous outcomes as the following: 1) improvement of > 50% from baseline; 2) improvement of > 30% from baseline; 3) "good" or "excellent" results on a categorical pain rating scale, or similar; or 4) another definition for a successful outcome. For studies that report more than one dichotomous result for an outcome, we will select one result to analyse, which we will base on the prioritised order presented in this list. For patients with radiculopathy, we will analyse improvement in leg pain; if leg pain is not reported, we will analyse back pain or overall pain. We will analyse dichotomous outcomes using Review Manager 5 (RevMan 2014). For dichotomous outcomes, we will classify the magnitude of effects based on the relative risk estimate as small (RR < 1.25 or > 0.8), moderate (RR 1.25 to 2.00 or 0.5 to 0.8), or large (RR > 2.0 or < 0.05) (Furlan 2015; Ostelo 2008).

For continuous outcomes, we will quantify treatment effects by the mean difference in change from baseline and 95% confidence interval. We will analyse results adjusted for baseline differences (e.g. analysis of covariance or multiple regression) if available. If mean difference in change from baseline is not available, we will analyse mean difference in follow‐up scores. We will attempt to perform analyses on continuous outcomes using Review Manager 5, but for analyses that we cannot perform in Review Manager 5 (e.g. analysis of covariance or using estimates from multiple regression), we will use Stata/SE 14.1. For pain, we will convert scales to a common 0 to 100 scale. For function, given the variability in available scales, for the primary analysis we will report the standardised mean difference (SMD). For continuous outcomes, we will define the magnitude of effects as small (< 10% difference on the scale or SMD of 0.2 to < 0.5), medium (10% to 20% or 0.5 to 0.8), and large (> 20%, > 0.8), as recommended in the 2009 updated Cochrane Back and Neck methods guidelines (and used in recent Cochrane Back and Neck reviews) (Cohen 1988; Rubinstein 2011). We will analyze outcomes at predefined time points (immediate, short‐term, intermediate‐term, and long‐term).

Unit of analysis issues

If studies include multiple corticosteroid treatment arms (e.g. treatment arms using different corticosteroids or different doses of the same corticosteroid), we will combine the arms if results are similar; if results differ, we will analyse the treatment arms separately, dividing the placebo comparison group by the number of treatment arms to avoid double‐counting of data.

Dealing with missing data

We will contact study authors for any missing or incompletely reported data. In cases where the data are reported as the median and interquartile range and we are unable to obtain the mean and 95% confidence intervals from the study authors, we will assume that the median is the mean and that the width of the interquartile range is equivalent to 1.35 times the 95% confidence interval (Higgins 2011). We will estimate data from graphs and figures when the information is not presented in tables or text. If information regarding standard deviations is missing and cannot be obtained from study authors, we will impute these from the confidence intervals or standard errors presented for other outcomes reported in the same study. If no measure of variability is reported for any outcome, we will estimate the standard deviation from the most similar trial in the review, taking the sample size and risk of bias of individual studies into consideration.

Assessment of heterogeneity

We will assess statistical heterogeneity based on visual inspection of the forest plot (e.g. for overlapping confidence intervals and the direction of effect) and the I² statistic. We will consider an I² of > 75% to represent substantial heterogeneity, 40% to 75% to represent moderate heterogeneity, and < 40% to represent low heterogeneity (Higgins 2011).

Assessment of reporting biases

We will perform comprehensive literature searches, including searches of multiple electronic databases and review of reference lists, in order to reduce the possibility of reporting bias. We will also generate funnel plots and perform the Egger test to evaluate small sample effects for meta‐analyses with at least 10 trials (Sterne 2011).

Data synthesis

We will combine the results from individual studies through meta‐analysis. We will analyse separately results for participants with radicular low back pain and non‐radicular low back pain. We will pool results for each primary and secondary outcome at different follow‐up times as described above. The premise for pooling is that based on application of the inclusion criteria, the studies will be similar in terms of the intervention evaluated (a corticosteroid) and use of a placebo comparator. We will separately pool studies of patients with radicular pain due to herniated disc, radicular pain due to spinal stenosis, and non‐radicular pain due to assumed variability in the underlying pathophysiology, which could impact responsiveness to corticosteroids. Because herniated disc is the most common cause of radiculopathy, we will analyse trials of patients with radiculopathy that do not perform imaging confirmation together with studies of confirmed herniated disc, unless there are clinical factors that suggest spinal stenosis (e.g. primarily bilateral symptoms, chronic presentation in older adults, presence of neurogenic claudication). Within each of these groups, we anticipate at least some heterogeneity because of likely differences between studies in terms of patient characteristics, settings, and interventions (e.g. specific corticosteroid used and dose); therefore, we will use a random‐effects model. If statistical heterogeneity is low, results of the random‐effects model will be similar to a fixed‐effect model. If studies are too dissimilar to pool, we will summarise the results qualitatively.

Regardless of whether evidence is synthesised quantitatively or qualitatively, we will assess the overall quality of the evidence for each comparison and outcome using the GRADE approach, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions, Higgins 2011, and adapted in the updated Cochrane Back and Neck methods guidelines (Furlan 2015) (Appendix 2). For comparisons and outcomes with data, we will grade the overall evidence as high, moderate, low, or very low quality. Because the review is restricted to randomised controlled trials, we will initially assume the overall quality of evidence to be high. Factors that will decrease the overall quality include the study type and assessments of risk of bias, imprecision, inconsistency, indirectness, and publication bias (Appendix 2).

The interpretation of the overall evidence grade categories (high, moderate, low, very low, or no evidence) is shown in Appendix 2.

'Summary of findings' table

A 'Summary of findings' table will summarise the findings for systemic corticosteroid versus placebo or no systemic corticosteroid for each of the primary and secondary outcomes specified above. We will stratify results for participants with radicular and non‐radicular low back pain, as well as outcomes reported for predefined time periods (immediate (< 2 weeks since initiation of treatment)), short‐term (two weeks to three months), intermediate‐term (> 3 months to < 12 month), and long‐term (≥ 12 months). We will analyse dichotomous and continuous outcomes as described in the Measures of treatment effect section.

Subgroup analysis and investigation of heterogeneity

We plan to perform subgroup analysis based on the following factors.

• Risk of bias: high risk of bias versus low risk of bias; we will also stratify trials according to use of a blinded versus non‐blinded design.

• Dose of corticosteroid: high or low total dose based on whether it is above or below the median total dose administered in prednisone equivalents.

• Number of corticosteroid doses: single dose, multiple corticosteroid doses over < 1 week, multiple corticosteroid doses over one week or longer.

• Route of administration: intravenous or intramuscular versus oral.

• Clinical setting: emergency room versus non‐emergency room setting.

• Duration of symptoms: acute (< 4 weeks) versus non‐acute (four weeks or longer).

• Imaging confirmation of herniated disc or spinal stenosis (for studies of radiculopathy): required imaging confirmation versus no imaging confirmation required.

As described above, we will stratify all analyses according to whether patients had radicular or non‐radicular low back pain, and we will analyse outcomes according to the prespecified categories for duration of follow‐up. For effects on function using a continuous measure, the primary analysis will report the standardised mean difference, and we will perform a subgroup analysis on the actual improvement of the RDQ, the ODI, and other measures of function reported in at least three trials.

We anticipate that there will be too few studies to perform meta‐regression, but will attempt meta‐regression on the factors described above if at least 10 studies meet inclusion criteria.

Sensitivity analysis

For continuous outcomes for pain and function, we will base the primary analysis on mean differences in improvement from baseline, adjusted for baseline differences when available. We will perform a sensitivity analysis based on mean differences at follow‐up, and use mean differences at follow‐up if mean differences from baseline are not available.

For long‐term outcomes, we will base the primary analysis on the longest duration of follow‐up available. We will perform a sensitivity analysis based on outcomes recorded closest to one year of follow‐up.

We will also examine results for outlier studies and perform sensitivity analyses excluding outliers, to help assess the robustness of findings.

Appendices

Appendix 1. Draft MEDLINE strategy

1. randomized controlled trial.pt.

2. controlled clinical trial.pt.

3. pragmatic clinical trial.pt.

4. clinical trial.pt.

5. comparative study.pt.

6. random$.ti,ab,kw.

7. placebo.ti,ab,kw.

8. controlled.ti,ab,kw.

9. prospective.ti,ab,kw.

10. compar$.ti,ab,kw.

11. trial.ti,ab,kw.

12. groups.ti,ab,kw.

13. or/1‐12

14. (animals not (humans and animals)).sh.

15. 13 not 14

16. back disorder$.ti,ab,kw.

17. dorsalgia.ti,ab,kw.

18. exp Back Pain/

19. (backache or back ache).ti,ab,kw.

20. ((back or lumb$) adj3 (pain or radicul$ or polyradicul$)).ti,ab,kw.

21. coccyx.ti,ab,kw.

22. coccydynia.ti,ab,kw.

23. sciatic$.ti,ab,kw.

24. exp sciatic neuropathy/

25. spondylosis.ti,ab,kw.

26. lumbago.ti,ab,kw.

27. Spinal Stenosis/

28. ((spine or spinal or lumb$) adj3 stenosis).ti,ab,kw.

29. ((disc$ or disk$) adj degenerat$).ti,ab,kw.

30. ((disc$ or disk$) adj prolapse$).ti,ab,kw.

31. ((disc$ or disk$) adj herniat$).ti,ab,kw.

32. Intervertebral Disc/

33. exp Intervertebral Disk Displacement/

34. or/16‐33

35. exp Glucocorticoids/

36. Adrenal Cortex Hormones/

37. Steroids/

38. exp Betamethasone/ [includes Clobetasol]

39. exp Dexamethasone/ [includes Desoximetasone]

40. exp Methylprednisolone/

41. exp Prednisolone/ [includes Fluprednisolone, Methylprednisolone and prednimusine]

42. exp Prednisone/

43. exp Hydrocortisone/

44. exp Cortisone/

45. exp Triamcinolone/

46. (corticosteroid$ or glucocortic$ or adrenal cortex hormone$ or steroid$ or corticoid$ or prednis$ or methylprednis$ or dexamet$ or betamet$ or hydrocort$ or cortisone$ or triamcin$).mp.

47. or/35‐46

48. 15 and 34 and 47

Appendix 2. The GRADE approach to evidence synthesis

We will categorise the quality of evidence as follows.

• High (⊙⊙⊙⊙): further research is very unlikely to change our confidence in the estimate of effect.

• Moderate (⊙⊙⊙○): further research is likely to have an important impact in our confidence in the estimate of effect.

• Low (⊙⊙○○): further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

• Very low (⊙○○○): any estimate of effect is very uncertain.

We will grade the evidence available to answer each subquestion on the domains in the following manner.

1. Study design: we will restrict the review to randomised trials, so we will initially assume that the quality of evidence is high.

2. Risk of bias: we will downgrade by one level if > 25% but < 50% of the participants in an analysis were enrolled in trials assessed as being at high risk of bias, and we will downgrade by two levels if > 50% of the patients in an analysis were enrolled in trials assessed as being at high risk of bias.

3. Inconsistency: we will downgrade by one level if the I² is > 75% or if there is only one trial in an analysis (since it is not possible to assess inconsistency).

4. Indirectness: we will downgrade by one level if > 50% of the participants are assessed as being outside the target group. As all outcomes are clinical outcomes and we will not be performing indirect comparisons, this type of indirectness is not relevant for this review.

5. Imprecision: we will downgrade by one level if the confidence interval for a relative risk estimate crosses one and the upper limit of the confidence interval is > 1.20 or < 0.80, or if the mean difference estimate crosses 0 and the upper or lower limit of the confidence interval is > 10 points above or below 0. We have based the confidence interval thresholds on the predefined thresholds for clinically meaningful differences.

6. Publication bias: we will downgrade by one level if we detect reporting bias or publication bias, based on identification of unpublished trials, presence of funnel plot asymmetry, or discrepancies between study protocols and the completed trials.

Contributions of authors

Roger Chou drafted the protocol. Tracy Dana, Nicholas Henschke, Rafael Zambelli Pinto, and Robert Lowe reviewed the draft and made revisions.

Declarations of interest

Roger Chou has conducted systematic reviews funded by the Agency for Healthcare Research and Quality and the American Pain Society that included systemic corticosteroids, and he led a guideline from the American College of Physicians and the American Pain Society that addressed systemic corticosteroids.. Tracy Dana has no conflicts of interest. Nicholas Henschke has no conflicts of interest. Robert Lowe owns stocks in companies that make systemic corticosteroids. His declaration was referred to the Cochrane Funding Arbiter, and it was determined that there was no breach of Cochrane policy. Rafael Zambelli Pinto has no conflicts of interest. Rochelle Fu has no conflicts of interest.

New