Acute low back pain: systematic review of its prognosisBMJ 2003; 327 doi: http://dx.doi.org/10.1136/bmj.327.7410.323 (Published 07 August 2003) Cite this as: BMJ 2003;327:323
- Liset H M Pengel, PhD student1,
- Robert D Herbert, senior lecturer ()1,
- Chris G Maher, associate professor1,
- Kathryn M Refshauge, associate professor1
- Correspondence to: R D Herbert
- Accepted 4 June 2003
Objectives To describe the course of acute low back pain and sciatica and to identify clinically important prognostic factors for these conditions.
Design Systematic review.
Data sources Searches of Medline, Embase, Cinahl, and Science Citation Index and iterative searches of bibliographies.
Main outcome measures Pain, disability, and return to work.
Results 15 studies of variable methodological quality were included. Rapid improvements in pain (mean reduction 58% of initial scores), disability (58%), and return to work (82% of those initially off work) occurred in one month. Further improvement was apparent until about three months. Thereafter levels for pain, disability, and return to work remained almost constant. 73% of patients had at least one recurrence within 12 months.
Conclusions People with acute low back pain and associated disability usually improve rapidly within weeks. None the less, pain and disability are typically ongoing, and recurrences are common.
Clinical practice guidelines promote the view that acute low back pain has a favourable prognosis—the 2000 UK guideline states that “90% [of cases] will recover within six weeks.”1 2 Yet these estimates are either unsubstantiated or based on individual studies. To date evidence of the prognosis of acute low back pain has not been systematically reviewed.
Many guidelines for acute low back pain advocate identification of adverse prognostic factors such as fear avoidance behaviours, leg pain, or low job satisfaction. Previous reviews of prognostic factors have been descriptive, do not use strict inception cohorts, or do not provide quantitative information of the predictive value of the factors.3–6
We aimed to systematically review published data on the course of acute low back pain and to identify clinically important prognostic factors. The term course refers to both the natural course and the clinical course of low back pain.
To be included studies had to be of a prospective design, describe the source of participants and method of sampling, have an inception cohort of participants with low back pain or sciatica for less than three weeks, have a follow up period of at least three months, and report on symptoms, health related quality of life, disability, or return to work. Studies were excluded that recruited patients with specific diseases such as arthritis, fracture, tumour, or cauda equina syndrome (but not sciatica).
Identification of studies and assessment of methodological quality
Studies were identified through searches of Medline, Embase, and Cinahl to March 2002. We also searched personal files and tracked references of included studies through the Science Citation Index. The search strategies were those recommended by the Cochrane Back Review Group together with a strategy for searching Medline for prognostic studies.7 8 Keywords used were inception, survival, logistic, Cox, life tables, and log rank. We had no language restrictions.
Despite there being no widely accepted method for assessing methodological quality of prognosis studies and no empirical evidence of bias related to various methodological features of such studies, validity criteria have been proposed.8 Methodological quality was assessed by six criteria (table 1). Two raters independently assessed the quality. A third reviewer resolved disagreements.
Data extraction and analysis
Study characteristics extracted from eligible papers were target population, sample size, duration of low back pain at time of enrolment, description of interventions, duration of follow up, prognostic factors, and outcome measures. Outcome data extracted were pain, disability, return to work, and recurrences. Data were extracted for time points where follow up was at least 80%. Data on return to work were obtained from the stratum of participants off work at baseline. To facilitate comparison, pain and disability scores were converted to a 100 point scale. Ten studies were controlled trials. For these studies, data were extracted for the control group, defined as the group receiving the least active intervention. In one trial, outcomes were reported only for the whole study sample because at follow up no differences were found between the groups receiving manual therapy, intensive training, or medical care.9 Prognostic data from this study are therefore based on the outcomes of the three groups.
The Wilson score method was used to calculate the confidence intervals for a single proportion.10 When it was possible to pool data across studies we obtained n weighted pooled means for continuous data and variance weighted pooled proportions for dichotomous data. The n weighted mean was used in preference to the variance weighted mean (the usual method of meta-analysis) because several studies did not provide variance data. Variance weighted pooled proportions were calculated using a random effects model.11
Studies evaluating prognostic factors used a range of modelling procedures and many different covariates, making pooling across studies problematic.8 Prognostic data were therefore not pooled. Data on prognostic factors were extracted only if the study reported on at least 80% of participants. If possible, odds ratios with 95% confidence intervals were extracted or calculated from the data. A second reviewer checked the data extraction.
The search retrieved 4458 articles, of which only 159 12–30 fulfilled all inclusion criteria and were included in our review (table 2). Five studies were described in more than one report.9 12 13 16 17 19 20 22 23 28 Of the 15 studies, only one monitored patients with sciatica.30 The studies included nine randomised controlled trials that evaluated exercise,9 15 16 18 22 23 25 28 manual therapy,9 28 an educational pamphlet,21 medical care,9 16 22 23 28 non-steroidal anti-inflammatory drugs,30 and bed rest25 26; one controlled trial that evaluated an early intervention in the work-place 12 13; and five cohort studies,14 19 20 24 27 29 one of which included an intervention by general practitioners.14 Patients were recruited from primary care,9 14 16 18 24 26–30 specialists,18 26 hospital emergency departments,9 15 18 22 23 28 and occupational healthcare providers.9 12 13 19–21 25 28 30
The two reviewers scored 84 quality criteria and agreed on 64 (76%). The intraclass correlation coefficient (2,1) for the total score was 0.52. Most studies defined the sample (87%). Five studies (33%) explicitly described methods for assembling a representative sample. Eleven studies (73%) had follow up of at least 80%. All but one study quantified prognosis.18 Six studies reported prognostic factors. Of the six studies, one (17%)19 20 used blinded assessment and four (67%)13 14 27 28 performed statistical adjustment for prognostic factors.
Course of low back pain
Most studies reported that pain decreased rapidly (by between 12% and 84% of initial levels, pooled mean 58%) within one month. Pain continued to decrease, albeit more slowly, until about three months (1). Two studies that provided data beyond the three month follow up showed that pain levels remained nearly constant until the 12 month follow up.12 16 The pooled mean level of pain on a 100 point scale was 22 at one month and 15 between three and 12 months. A similar trend was seen for disability, which decreased by between 33% and 83% of initial levels (pooled mean 58%) within one month (1). One study reported data on six month follow up.12 The pooled mean level of disability on a 100 point scale was 24 at one month and 14 between three and six months.
Between 68% and 86% of participants initially off work returned to work within one month (pooled estimate 82%, 95% confidence interval 73% to 91%; 2). One study reported data on six month follow up.21 The pooled estimate of the proportion of participants who returned to work, extracted from studies that reported return to work at three to six months, was 93% (91% to 96%).
The cumulative risk (one study, 135 participants) of at least one recurrence within three months was 26% (19% to 34%).26 Two studies reported recurrences within 12 months.16 23 The cumulative risk of at least one recurrence within 12 months varied from 66% to 84% (pooled estimate 73%, 59% to 88%). One study reported a cumulative risk of recurrence after three years of 84%.23
One study included patients with sciatica.30 In this sample, both back pain and leg pain decreased, on average, by 69% of initial scores within one month. Disability decreased by 57% of initial scores within one month. Data on long term pain and disability were not available.
Three studies reported on prognostic factors for at least 80% of the population.14 19 27 With one exception, odds ratios of significant prognostic factors ranged from 0.04 to 10.4. One study reported that scores of 0.48 or more on the Vermont disability prediction questionnaire were predictive of return to work at three months (odds ratio 76.3, 9.6 to 604.9; positive likelihood ratio 5.7, 3.9 to 8.5; negative likelihood ratio 0.07, 0.01 to 0.50).19
Our review confirms the widely held view that most people with acute low back pain have rapid improvements in pain and disability within one month. Most of those off work with back pain also returned to work within one month. Further improvement occurred until about three months. Thereafter levels of pain, disability, and return to work remained almost constant, although only two studies provided follow up data beyond three months.12 16
Although most people return to work within 12 months, low levels of pain and disability persist. The studies did not report enough data to establish if levels of long term pain and disability reflect a small subgroup with high levels of pain and disability or a large subgroup with low levels of pain and disability. Nor is it clear whether chronic low levels of pain and disability are due to persistence of the original episode or to recurrent episodes.
Findings from previous reviews on prognostic factors of low back pain have been inconsistent.3–6 Putative prognostic factors include psychological factors such as distress,3 5 personal factors such as previous back pain,6 and work related factors such as job satisfaction.4 However, the evidence of the prognostic value of these factors comes mainly from studies that either did not recruit a relevant cohort or were methodologically weak. We located only one relevant, methodologically strong paper that provided evidence of a clinically useful predictor of outcome (in this case return to work) for primary care patients with acute low back pain. Hazard et al reported that scores of 0.48 or more on the Vermont disability prediction questionnaire were associated with a likelihood ratio of 5.7 and scores of less than 0.48 were associated with a likelihood ratio of 0.07.19 Given the low prevalence of failure to return to work at three months (pooled estimate of 6%), this predictor may be of limited clinical utility. Moreover, the cut-off score of 0.48 was chosen by inspection of the data, which is known to inflate predictive accuracy.31
What is already known on this topic
Clinical practice guidelines state that recovery from acute low back pain is rapid and complete
What this study adds
People with acute back pain experience improvements in pain, disability, and return to work within one month
Further but smaller improvements occur up to three months, after which pain and disability levels remain almost constant
Low levels of pain and disability persist from three to at least 12 months
Most people will have at least one recurrence within 12 months
Participants off work with low back pain have higher pain and disability scores than people who are working.32 Thus it may be sensible to consider separately the prognosis of those off work. It remains unclear if the prognosis of participants initially off work is worse than those who are not.
We included only studies that recruited inception cohorts of participants with low back pain or sciatica for less than three weeks. This policy may be sufficiently restrictive or too restrictive. A formal sensitivity analysis of participants with low back pain for less than one week and for less than three weeks showed that the reduction in pain and disability is similar in these two groups, justifying inclusion of studies with participants having pain for up to three weeks. However, inclusion of participants with low back pain for up to six weeks seems unjustified. Our data show that study participants had rapid improvements in pain and disability within one month. By six weeks, participants had already improved significantly; typically pain and disability were only a third of initial values. Moreover, many people no longer had back pain at six weeks, so those recruited with back pain for six weeks cannot be representative of all people who have back pain. We therefore believe it is justifiable to restrict our review to participants with low back pain for three weeks or less.
Contributors LHMP designed the study protocol, located and selected studies, extracted and interpreted the data, wrote the paper, and approved the final manuscript. RDH designed the study protocol, extracted and interpreted the data, advised on the statistical analysis, and revised and approved the final manuscript. CGM and KMR designed the study protocol, assessed the quality of the trials, interpreted the data, and revised and approved the final manuscript. LHMP will act as guarantor for the paper
Funding LHMP's scholarship was funded by the National Health and Medical Research Council and the Australasian Physiotherapy Low Back Pain Trial Consortium. The guarantor accepts full responsibility for the conduct of the study, had access to the data, and controlled the decision to publish.
Competing interests None declared