Article Text

Download PDFPDF

Clinical course of patients with low back pain following an emergency department presentation: a systematic review and meta-analysis
  1. Danielle M Coombs1,2,3,
  2. Gustavo C Machado1,2,
  3. Bethan Richards1,4,
  4. Crystian B Oliveira5,
  5. Robert D Herbert6,
  6. Chris G Maher1,2
  1. 1 Institute for Musculoskeletal Health, The University of Sydney and Sydney Local Health District, Sydney, New South Wales, Australia
  2. 2 Sydney School of Public Health, Faculty of Medicine and Health, The University of Sydney, Sydney, New South Wales, Australia
  3. 3 Physiotherapy Department, Royal Prince Alfred Hospital, Sydney Local Health District, Sydney, New South Wales, Australia
  4. 4 Rheumatology Department, Royal Prince Alfred Hospital, Sydney Local Health District, Sydney, New South Wales, Australia
  5. 5 Faculdade de Ciências e Tecnologia, Universidade Estadual Paulista, Presidente Prudente, São Paulo, Brazil
  6. 6 Neuroscience Research Australia (NEURA), Sydney, New South Wales, Australia
  1. Correspondence to Ms Danielle M Coombs, PO Box M179, Missenden Road, Camperdown, New South Wales, Australia; danielle.coombs{at}


Introduction Low back pain, and especially non-specific low back pain, is a common cause of presentation to the emergency department (ED). Although these patients typically report relatively high pain intensity, the clinical course of their pain and disability remains unclear. Our objective was to review the literature and describe the clinical course of non-specific low back pain after an ED visit.

Methods Electronic searches were conducted using MEDLINE, CINAHL and EMBASE from inception to March 2019. We screened for cohort studies or randomised trials investigating pain or disability in patients with non-specific low back pain presenting to EDs. We excluded studies that enrolled participants with minimal pain or disability scores at baseline. Two reviewers independently screened the full texts, extracted the data and assessed risk of bias and quality of evidence. Estimates of pain and disability were converted to a common 0–100 scale. We estimated pooled means and 95% CIs of pain and disability as a function of time since ED presentation.

Results Eight studies (nine publications) with a total of 1994 patients provided moderate overall quality evidence of the expected clinical course of low back pain after an ED visit. Seven of the eight studies were assessed to have a low risk of bias. At the time of the ED presentation, the pooled estimate of the mean pain score on a 0–100 scale was 71.0 (95% CI 64.2–77.9). This reduced to 46.1 (95% CI 37.2–55.0) after 1 day, 41.8 (95% CI 34.7 to 49.0) after 1 week and 13.5 (95% CI 5.8–21.3) after 26 weeks. The course of disability followed a similar pattern.

Conclusions Patients presenting to EDs with non-specific low back pain experience rapid reductions in pain intensity, but on average symptoms persisted 6 months later. This review can be used to educate patients so they can have realistic expectations of their recovery.

  • musculo-skeletal
  • spine non trauma

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known on this subject

  • Low back pain is the leading cause of disability internationally and a top reason for presentation to the ED.

  • All major guidelines for the management of low back pain recommend reassurance of good prognosis and education; however, it is unclear whether these guidelines can be applied in the ED creating a significant challenge for clinicians.

  • This is largely because prognostic information specific to the ED setting is limited in the literature, and the clinical course of these patients is not known.

What this study adds

  • Our findings show that patients presenting to the ED with non-specific low back pain experience rapid reductions in pain intensity and disability within 1 week, but on average mild symptoms persist.

  • These findings can be used to provide prognostic information to patients and assist clinicians with patient education.


Low back pain is the leading cause of disability internationally.1 The extent of low back pain as a major public health problem has become increasingly evident over the past few decades.2 It provides a significant impact on healthcare resources, with over 87 billion dollars spent each year on low back and neck pain in the USA alone.3 The number of outpatient and ambulatory care visits for low back pain is growing.4 Low back pain is responsible for 4.4% of emergency department (ED) presentations worldwide,5 putting it in the top five reasons for ED attendance.6 Most presentations of low back pain cannot be definitively linked to a specific anatomical structure or disease. Such presentations are classified as non-specific low back pain.2

Although low back pain causes a large health burden, it can be self-limiting and typically has a favourable prognosis. Systematic reviews of the clinical course of low back pain in the primary care setting have shown pain typically improves markedly within the first 6 weeks.7–9 Low back pain does, however, seem to be more of a challenge in the ED setting. Management is often not consistent with guideline recommendations and frequently includes avoidable imaging tests, high rates of opioid use and prescription, and costly hospital admissions.10 11 Low back pain patients report relatively high pain scores in the ED12 creating a significant problem for treating clinicians. Published systematic reviews have not specifically investigated the prognosis of low back pain following an ED presentation.8 9 13–15 Thus, it is not clear if patients who report to the ED with low back pain have the same favourable clinical course as patients who report to primary care with low back pain.

It is unlikely for someone who presents to the ED to receive no intervention. Therefore, this systematic review aims to describe the clinical course (ie, the progress of a condition when treated), not the natural history (ie, progress of a condition in the absence of treatment), of pain and disability in adults with non-specific low back pain who present to an ED. The focus is on the clinical course of low back pain (ie, how the condition evolves over time), not on prognostic factors or risk factors.


The review protocol was registered with PROSPERO (registration number CRD42018110271). The report structure was written in accordance with the Preferred Reporting for Systematic Reviews and Meta-analyses16 and the Meta-analysis of Observational Studies in Epidemiology17 statements.

Search strategy

Electronic searches were conducted using MEDLINE, CINAHL and EMBASE databases from inception to March 2019. We combined relevant search terms and their variations to construct a search strategy for each database (online supplementary appendix 1). We also completed a manual search of references and relevant systematic reviews to identify any additional trials not retrieved by the above methods. There was no language restriction for articles. Initial screening of titles and abstracts was conducted by a single reviewer (CBO) and clearly irrelevant articles were removed. Two independent reviewers (DMC and CBO) then screened the full text of relevant articles to determine inclusion. In the case of disagreement, a third reviewer (GCM) was consulted.

Supplemental material

Study selection

Quantitative studies with longitudinal designs (eg, cohort studies and controlled trials) were eligible for inclusion if they reported the time course of pain and disability in adults (aged 18 years or older) with a diagnosis of non-specific low back pain following an ED presentation. We had planned to include only prospective studies in our registered protocol, but due to the small number of studies in this area, we decided to broaden the inclusion criteria and included retrospective cohort studies. The inclusion of both controlled trials and cohort studies has been previously used in other systematic reviews describing the clinical course.18 Studies investigating mixed pain populations (ie, including conditions other than non-specific low back pain, such as sciatica, thoracic or neck pain) were included if the majority (>75%) of the sample had non-specific low back pain or if data were presented separately for the subsample with non-specific low back pain. When the type or types of low back pain being studied were unclear, for example, in studies which stated that they were investigating ‘low back pain’ or ‘low back pain with or without leg pain’, we included them on the basis of the assumption that >75% were likely non-specific low back pain. We did this because previous cohort studies investigating low back pain in the ED have found that >85% of low back pain presentations in the ED were non-specific.10 19 There was no restriction associated with the duration of low back pain symptoms on presentation to the ED, but studies were classified as acute (less than 6 weeks), subacute (6–12 weeks) or chronic (greater than 12 weeks) based on the average duration of symptoms at the time of ED presentation in the included cohort. Studies had to include a measure of pain intensity or disability at baseline and at least one follow-up time point to be included in this review. Follow-up time points could be any time after the presentation to the ED. There were no language restrictions.

Studies that only recruited patients with the cause of pain classified as serious spinal pathology (eg, fracture, infection, cancer, cauda equina syndrome or inflammatory arthropathy) were excluded. Studies that only recruited patients with lumbar radicular pain that is likely from a specific cause (eg, nerve root compression or lumbar spinal stenosis) were also excluded. We excluded studies that explored serious spinal pathology or lumbar radicular pain because we expect that the prognosis of these cohorts is likely to be different from the prognosis of those with non-specific low back pain. Studies were excluded if potential participants were excluded based on pain intensity or disability levels at baseline. We excluded these studies to reduce selection bias and ensure we explored cohorts where all participants with non-specific low back pain were included and not only those with higher pain and disability scores at baseline. Studies were also excluded if they were not reported as a full report in a peer-reviewed journal.

Data extraction

A standardised form was constructed and two reviewers (DMC and CBO) independently extracted data from the included studies. Disagreements were resolved by consensus. Data extracted from included studies included study design, sample size, sample characteristics (ie, age, gender and duration of symptoms at ED presentation), type of intervention received (ie, pharmacological, non-pharmacological or a combination of those) and measures of central tendency and dispersion of pain and disability at baseline and all follow-up time points. Where outcome data were missing or incomplete, the authors of the included studies were contacted and asked for summary data.

The primary outcomes were pain and disability, reported as continuous variables. Measures for pain included, for example, numerical rating scale scores (range, 0–10) or visual analogue scale scores (range, 0–100). These two pain measures are highly correlated and can be used interchangeably when transformed.20 Disability measures included, for example, Roland-Morris Disability Questionnaire scores (range, 0–24) or Oswestry Disability Index scores (range, 0–100). These measures were converted to a common 0–100 scale, where 0 is no pain or disability and 100 is the worst pain or disability. Converting pain and disability scores to a common 0–100 scale has been frequently used in systematic reviews of the clinical course of other conditions.8 13 18 21 Where data were reported using dichotomous or categorical scales (eg, mild, moderate and severe), the author was contacted and summaries of continuous data were provided where available, otherwise the data were not used in the analysis.

Risk of bias assessment

Each study in the review was assessed for risk of bias using the Quality In Prognosis Studies (QUIPS) tool.22 Although the QUIPS tool was originally designed for prognostic factor reviews, Hayden et al 22 suggested that it can also be used in clinical course reviews by selecting only the relevant domains. Other overall prognosis reviews have also done this.23 In this review, we used the study participation, study attrition, outcome measurement, and statistical analysis and reporting domains. Using the criteria described in the QUIPS tool,22 each of these domains was assessed as being at a low, moderate or high risk of bias. Two reviewers (DMC and CBO) independently conducted the risk of bias assessment. In the case of disagreement, a third reviewer (GCM) was consulted.

We used the Grading of Recommendations of Assessment, Development and Evaluation (GRADE) approach to assess the certainty in the body of evidence presented in this systematic review, as either high, moderate, low or very low quality of evidence. The higher the rating of the quality of evidence, the more confident we are that the true prognosis lies close to the estimate we report. Although the GRADE approach was originally developed to judge the quality of evidence in intervention studies, Iorio et al demonstrated that the same principles can be applied to assessing bodies of evidence regarding overall prognosis.24 Two reviewers (DMC and CBO) conducted GRADE assessment exploring the domains: risk of bias, inconsistency, imprecision, indirectness and publication bias. Any disagreement was resolved by a third reviewer (GCM). The guide that was used for the GRADE assessment was adapted from Iorio et al and Huguet et al 24 25 and is presented in the online supplementary appendix 2. We originally planned to only report the I2 statistic as a measure of heterogeneity, but considering the limitations of the I2 statistic,26 we also decided to calculate the tau statistic (the square root of the between-study variance) to evaluate the between-study variance.

Data synthesis and analysis

The primary outcomes (pain and disability) were graphed as a function of time since the ED presentation. For controlled studies, data (means and SD) were extracted for all groups and combined as recommended by the Cochrane Handbook for Systematic Reviews of Interventions.27 In the case where one of the intervention arms was not considered reasonable practice in the emergency setting (eg, placebo or sham groups), that intervention arm was not used in the analysis. All available follow-up time points were used in the analysis. Fractional polynomial random-effects meta-regression was used to model the non-linear relationship between time (treated as a continuous variable) and summary outcomes (pain and disability). Each estimate from each study was assigned a weight equal to the inverse of the sum of the variance of the estimate and the between-study variance. A restricted maximum likelihood estimate of the between-study variance was obtained.28 Cluster-robust standard errors were used to account for the dependence of repeated observations within studies – this is a small departure from the protocol, which specified the use of generalised estimating equations. The model was used to estimate pooled means and 95% CIs of expected outcomes at 4 hours, 1 day, and 1, 6, 12 and 26 weeks after ED discharge.29 The model was also used to calculate 95% prediction intervals for the true mean outcome in a new study from an exchangeable population.30 We planned to conduct subgroup analyses to investigate whether different study level characteristics (eg, types of treatment and duration of pain) were associated with differences in the clinical course, however, we did not have enough studies to do this. Stata V.16 (StataCorp LP) was used for all analyses.

Patient and public involvement

This systematic review of published literature was completed without patient or public involvement in the design or conduct. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.


The titles and abstracts of a total of 8036 studies were screened, and the full text of 69 articles were reviewed (figure 1). Of these, nine publications (reporting on eight studies) were included in the review. The main reasons for exclusion were that the study was not conducted in an ED setting, participants did not have non-specific low back pain, relevant data were not reported (ie, no pain or disability data, or where participants were excluded based on pain or disability levels at baseline) and the report was not a full-text report (online supplementary appendix 3). We contacted authors for one included study where only categorical data were reported. The authors provided us with continuous data, so we were able to include this study in the analysis. There were four controlled trials31–34 and four cohort studies.35–39 One cohort study reported the same data in two publications.35 36 Overall, the eight studies investigated the clinical course of 1994 patients post-ED presentation for non-specific low back pain. The characteristics of the included studies and participants are described in table 1. The available follow-up time points used in each study and the number of participants at each time point are summarised in the online supplementary appendix 4.

Figure 1

Flow diagram of studies included in the systematic review.

Table 1

Characteristics of included studies

The clinical course of low back pain after ED presentation

Pain and disability behaved in similar ways (figure 2). On average, both pain and disability were moderately severe on presentation to the ED but subsequently decreased, initially rapidly then more slowly. Low levels of pain and disability were still present, and pain and disability were still decreasing, at 26 weeks. There was moderate between-study heterogeneity in the magnitude of pain and disability, but all studies showed qualitatively similar clinical courses of pain and disability.

Figure 2

Clinical course of pain and disability in patients with non-specific low back pain presenting to an ED. (A) Mean pain and (B) mean disability at all time points reported in all included studies. The areas of the bubbles are proportional to the weights assigned to the means in the meta-regression. (C) Mean pain and (D) mean disability, with a line joining means from a single study. Each study is represented with a different symbol. Error bars are 95% CIs. The lower row superimposes the regression line. The dark, inner shaded band is the 95% CI, representing uncertainty in the estimate of the mean course of (E) pain or (F) disability. The light, outer shaded band is the 95% prediction interval, representing uncertainty in predictions of the true mean course of pain or disability of a future study conducted on an exchangeable population.

At ED presentation, the pooled mean pain score on a 0–100 scale was 71.0 (95% CI 64.2–77.9). This decreased to 47.7 (95% CI 38.2–57.3) after 4 hours and to 46.1 (95% CI 37.2–55.0) after 1 day. At 1 week, the pooled mean pain score was 41.8 (95% CI 34.7–49.0) that further reduced to 31.8 (95% CI 27.8–35.8) at 6 weeks, 24.8 (95% CI 20.8–28.8) at 12 weeks, and to 13.5 (95% CI 5.8–21.3) at 26 weeks (figure 2).

For disability, the pooled mean score on a 0–100 scale at ED presentation was 69.0 (95% CI 57.9–80.0). This decreased to 65.8 (95% CI 53.7–77.9) after 4 hours and to 59.6 (95% CI 45.4–73.9) after 1 day. At 1 week, the pooled disability score was 48.9 (95% CI 32.2–65.6) that further reduced to 33.5 (95% CI 17.6–49.3) at 6 weeks, 24.9 (95% 11.2–38.6) at 12 weeks, and 12.4 (95% CI 3.1–21.8) at 26 weeks (figure 2).

Risk of bias assessment

When assessing the risk of bias of each study using the QUIPs tool, seven studies were considered as low risk of bias and one study was considered as a moderate risk of bias (table 2). Seven of the eight studies adequately reported study participation and outcome measurement and were rated as low risk of bias in these domains. The domain most rated as a moderate or high risk of bias (44%) was statistical analysis and reporting. Sixty-seven per cent of studies were rated as a low risk of bias for study attrition.

Table 2

Risk of bias of included studies using QUIPS tool

Using the GRADE approach, the overall quality of evidence was assessed as moderate for both pain and disability outcomes. We downgraded the quality of evidence due to heterogeneity between studies (inconsistency). Heterogeneity was assessed with the tau statistic (the square root of the between-study variability), which was 11.5 for pain and 9.2 for disability. We also calculated the I2 statistic and found it to be 98.3% for pain and 96.4% for disability. The other GRADE domains (study design, risk of bias, imprecision, indirectness and publication bias) all fulfilled predetermined criteria for judging the quality of evidence (criteria are reported in the online supplementary appendix 2).


This systematic review with meta-analysis was able to quantify the clinical course of pain and disability over time following an ED presentation for non-specific low back pain. We included eight studies (nine publications) with 1994 participants in total. We found moderate-quality evidence that patients with non-specific low back pain experience an initial rapid reduction in symptoms after ED presentation. On the first day, there is a 35% reduction in pain intensity, 41% after 1 week and 53% at 6 weeks. However, on average these patients still have mild pain at 26 weeks. We found a similar pattern for disability.

To our knowledge, this is the first systematic review that has estimated the clinical course of pain and disability in patients with non-specific low back pain presenting to the ED. Other systematic reviews investigating clinical courses have focused on primary care settings7–9 and have shown rapid reduction after initial presentation. This is consistent with our review. A strength of our study is the modelling approach that permits estimation of pooled means over time. Furthermore, as we extracted data from all time points reported in the included studies, we were able to describe the initial period of the clinical course in detail, highlighting another strength of our review. For example, we report the pooled pain and disability scores at 4 hours, 1 day and 1 week, illustrating the most improvement shortly after the presentation to the ED (online supplementary appendix 5). This is important prognostic information for ED clinicians when discussing the clinical course of this condition with their patients.

A previous systematic review has investigated the clinical course of non-specific low back pain in primary care using a similar method. Menezes Costa et al 8 found pooled mean pain score (on a 0–100 scale) for patients with acute low back pain in primary care was 52 (95% CI 48–57) at the initial onset of pain, 23 (95% CI 21–25) at 6 weeks and 12 (95% CI 9–15) at 26 weeks.8 In the ED, we found higher mean pain scores at baseline and a more rapid pain reduction up to 1 week, then slower improvements until 26 weeks, when compared with primary care. The findings for disability were similar. By 26 weeks, the primary care and ED cohorts seem to have similar predicted pooled mean scores of pain and disability. It is possible that ED patients may have specific prognostic factors contributing to more severe symptoms and delayed recovery. Pain catastrophising, older age, poor general health, increased psychological or psychosocial stress, poor relations with colleagues, physically heavy work, worse baseline functional disability and the presence of compensation have been associated with poorer outcomes in low back pain patients.14 15 The literature, however, does not explore differences in prognostic factors between the ED and primary care populations. Interestingly, when using a prognostic screening tool in the ED, the majority of low back pain patients were classified as ‘moderate’ and ‘high’ risk,38 whereas in primary care, the majority of patients were classified as ‘low’ and ‘moderate’ risk.40 Caution is needed when comparing clinical course across different settings, as the differences seem small and are potentially due to between-study heterogeneity.

Clinical relevance

This review adds valuable prognostic information for clinicians when educating patients with low back pain in the ED setting. Education and advice are recommended in all major low back pain guidelines,41 but we have limited prognostic evidence on what advice to give in the ED. Given the clinical course illustrated in our review, ED clinicians can more confidently advise that on average, patients will see a rapid reduction in symptoms immediately after an ED presentation. However, it is important to note that these patients, similar to their primary care counterparts,8 on average still have mild pain at 26 weeks with a pooled mean pain score of 13.5 (95% CI 5.8–21.3) and a pooled mean disability score of 12.4 (95% CI 3.1–21.8) on a 0–100 scale. These results, together with the huge socioeconomic impact of chronic low back pain,1 highlight the need to address persistent pain and disability in this population. It may be beneficial to have a clearer referral pathway in place for follow-up of these patients from the ED to ensure that they fully recover. Patients with chronic low back pain benefit from multidisciplinary biopsychosocial rehabilitation,42 but access to these services is limited and wait times are long. Imaging, hospital admissions and surgery continue to be prevalent, exhausting healthcare resources on these low-value management options, rather than redirecting resources to evidence-based, high-value care.2 41 43 Health system change is needed to address this problem.44 ED clinicians can contribute to this change by adopting the evidence-based practice and by providing patients with appropriate education on their time course of recovery, as well as referral to high-value care on discharge, potentially leading to a reduction in high cost, unnecessary care during and following their ED presentation.


There are some limitations to our review. First, the results need to be interpreted within the context of a moderate degree of between-study heterogeneity, as indicated by the tau statistic. The I2 statistic reported should be interpreted with caution. It is important to note that the I2 statistic is very high, less because of high between-study heterogeneity and more because within-study variances are low, highlighting the limitations of the I2 statistic in this context.26 The small number of included studies precluded the prespecified subgroup analyses and prevented the exploration of sources of between-study heterogeneity. The definitions of low back pain and non-specific low back pain were clearer in some studies than others. Some studies in the review may have included mixed types of low back pain, which could have contributed to clinical heterogeneity between studies. Another limitation is that we could not find any studies that followed patients beyond 26 weeks, so the long-term clinical course of low back pain after an ED presentation remains unknown. Only two studies reported pain intensity data and only one reported disability scores at 26 weeks. This may be because low back pain in the ED has not been widely investigated and follow-up from the ED is seldom done. If follow-up occurs, it is often only in the short to medium term.

In conclusion, patients with non-specific low back pain, on average, experience a 29 point or 41% absolute reduction in symptoms in the first week after an ED presentation. This continues to reduce over time, but on average patients still have mild symptoms at 6 months. This review provides a quantitative summary of the clinical course of low back pain after an ED visit that can be used to inform patients of what they may expect of their recovery.

Ethics statements

Patient consent for publication



  • Handling editor Edward Carlton

  • Twitter @gustavocmachado, @crystianbso

  • Contributors All authors contributed substantially to conception and design, acquisition and interpretation of the data. CBO conducted the first search and initial screening of abstracts. CBO and DMC performed screening of full text determined inclusion and the main data collection. RDH, GCM and DMC performed the statistical analysis. BR and DMC provided clinical expertise in low back pain management. Study supervision was conducted by CGM and GCM. DMC drafted the manuscript. All authors participated substantially in the critical revision of the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests CGM is supported by a Principal Research Fellowship from Australia’s National Health and Medical Research Council (APP1103022) as well as a Program grant (APP1113532) and two Centre for Research Excellence grants (APP1134856, APP1171459). He has received research grants from various government and not for profit agencies. His expenses have been covered by professional associations hosting conferences he has spoken at. GCM is supported by an Early Career Fellowship from Australia’s National Health and Medical Research Council (APP1141272). FlexEze provided heat wraps at no cost for the SHaPED trial that DMC, GCM, BR and CGM are investigators on.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.