LBP indicates lower back pain; PEDro, Physiotherapy Evidence Database; and RCT, randomized clinical trial.
Studies are ordered chronologically within prevention strategies. Short-term indicates follow-up of 12 months or less; long-term, follow-up evaluation of more than 12 months.
aOnly the baseline sample size was available.
Studies are ordered chronologically within prevention strategies. Short-term indicates follow-up of 12 months or less; long-term, follow-up evaluation of more than 12 months.
eTable 1. Search Strategy
eTable 2. PEDro Scores of Included Randomized Controlled Trials
eTable 3. Low Back Pain Prevention Strategies Estimates for Short- and Long-term Low Back Pain Episode and Sick Leave Outcome Measures
Customize your JAMA Network experience by selecting one or more topics from the list below.
Steffens D, Maher CG, Pereira LSM, et al. Prevention of Low Back Pain: A Systematic Review and Meta-analysis. JAMA Intern Med. 2016;176(2):199–208. doi:10.1001/jamainternmed.2015.7431
Existing guidelines and systematic reviews lack clear recommendations for prevention of low back pain (LBP).
To investigate the effectiveness of interventions for prevention of LBP.
MEDLINE, EMBASE, Physiotherapy Evidence Database Scale, and Cochrane Central Register of Controlled Trials from inception to November 22, 2014.
Randomized clinical trials of prevention strategies for nonspecific LBP.
Data Extraction and Synthesis
Two independent reviewers extracted data and assessed the risk of bias. The Physiotherapy Evidence Database Scale was used to evaluate the risk-of-bias. The Grading of Recommendations Assessment, Development, and Evaluation system was used to describe the quality of evidence.
Main Outcomes and Measures
The primary outcome measure was an episode of LBP, and the secondary outcome measure was an episode of sick leave associated with LBP. We calculated relative risks (RRs) and 95% CIs using random-effects models.
The literature search identified 6133 potentially eligible studies; of these, 23 published reports (on 21 different randomized clinical trials including 30 850 unique participants) met the inclusion criteria. With results presented as RRs (95% CIs), there was moderate-quality evidence that exercise combined with education reduces the risk of an episode of LBP (0.55 [0.41-0.74]) and low-quality evidence of no effect on sick leave (0.74 [0.44-1.26]). Low- to very low–quality evidence suggested that exercise alone may reduce the risk of both an LBP episode (0.65 [0.50-0.86]) and use of sick leave (0.22 [0.06-0.76]). For education alone, there was moderate- to very low–quality evidence of no effect on LBP (1.03 [0.83-1.27]) or sick leave (0.87 [0.47-1.60]). There was low- to very low–quality evidence that back belts do not reduce the risk of LBP episodes (1.01 [0.71-1.44]) or sick leave (0.87 [0.47-1.60]). There was low-quality evidence of no protective effect of shoe insoles on LBP (1.01 [0.74-1.40]).
Conclusion and Relevance
The current evidence suggests that exercise alone or in combination with education is effective for preventing LBP. Other interventions, including education alone, back belts, and shoe insoles, do not appear to prevent LBP. Whether education, training, or ergonomic adjustments prevent sick leave is uncertain because the quality of evidence is low.
Low back pain (LBP) is one of the most burdensome health problems worldwide,1 generating enormous costs in treatments and time lost from work.2 The global point prevalence of LBP is 12%; with the aging population, the number of people affected is likely to increase over the coming years.3Quiz Ref ID A key contributor to the burden is the high recurrence rate: approximately one-half of patients experience a recurrence of LBP within 1 year after recovering from a previous episode.4-6 It is therefore important to know whether it is possible to prevent LBP and, if so, which interventions are most effective.
Although there have been several systematic reviews of strategies to prevent LBP, most have major limitations. Many of the existing reviews are out-of-date,7,8 report data from randomized clinical trials (RCTs) of symptomatic participants,9 do not consider the strength of evidence (eg, using the Grading of Recommendations Assessment, Development, and Evaluation [GRADE] system),8,10 are restricted to a particular type of intervention11 or setting, or do not follow a prespecified, publicly accessible protocol.7,8
Therefore, a comprehensive, high-quality review that includes the most recent publications is needed to provide a current overview of the effectiveness of prevention strategies. The aim of this systematic review was to evaluate the evidence on the effectiveness of interventions for prevention of episodes of LBP and use of sick leave due to LBP.
The PRISMA Statement was used to guide the conduct and reporting of the study.12 This study searched the following electronic databases from the earliest record to November 22, 2014: MEDLINE, EMBASE, Physiotherapy Evidence Database (PEDro), and the Cochrane Central Register of Controlled Trials. A sensitive search strategy was used based on the recommendations of the Cochrane Back Review Group13 for randomized controlled trials and back pain as well as search terms for prevention.14 The full search strategy is outlined in eTable 1 in the Supplement. The reference lists of relevant reviews and trials were screened for additional studies, and we also used citation tracking of all included trials.
During the first screening, 2 reviewers (D.S. or M.J.H. with V.C.O. or M.C.) evaluated the titles and abstracts of each citation and excluded clearly irrelevant studies. For each potentially eligible study, 2 reviewers (D.S. or M.J.H. with V.C.O. or M.C.) examined the full-text article and assessed whether the study fulfilled the inclusion criteria. In cases of disagreement, a decision was made by consensus or, if necessary, a third reviewer (C.G.M.) was consulted.
We included RCTs assessing the effectiveness of prevention strategies for nonspecific LBP. To be eligible, trials needed to meet the following criteria: (1) included participants without LBP at study entry or at least 1 outcome was not present at baseline (eg, some participants had mild LBP, but all were working and the study outcome was an episode of work absence due to LBP); (2) aimed to prevent future episodes of LBP; (3) compared intervention group with groups that received no intervention, placebo, or minimal intervention; and (4) reported a measure of a new episode of LBP (eg, episode of LBP or episode of sick leave due to LBP). Studies that used a quasi-randomized design or reported the comparison of 2 prevention strategies (eg, exercise vs lumbar support) were excluded. No restrictions were placed on the setting or context of the included studies, languages, or date of the RCT report.
We assessed the quality of the trials’ methods using the PEDro scale15,16 by either downloading the available scores from the PEDro database (http://www.pedro.org.au) or rating the trial ourselves. Scores on the PEDro scale range from 0 (very low methodologic quality) to 10 (high methodological quality); methodologic quality was not an inclusion criterion of this review.
Two independent reviewers (D.S. or M.J.H. with V.C.O. or M.C.) extracted the characteristics and intervention outcomes of each trial using a standardized data extraction form. When possible, we extracted the raw outcomes (number of persons having an episode of LBP) for each group (intervention and control) and calculated the estimates of treatment effect using methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions, Version 220.127.116.11
To evaluate the overall quality of the evidence, we used the GRADE system.18 The GRADE classification was downgraded from high quality by 1 level for each factor that we encountered: (1) design limitation (>25% of participants from studies with low methodologic quality: PEDro score <7), (2) inconsistency of results (wide variance of point estimates across studies or large heterogeneity between trials: I2 >50%), and (3) imprecision (<400 participants for each outcome). We did not consider the indirectness criterion in this review because we included a specific population with relevant outcomes and direct comparisons. A GRADE profile was completed for each pooled estimate and for single trials comparing an LBP prevention strategy with controls. When only single RCTs were available, evidence from studies with fewer than 400 participants was downgraded for inconsistency and imprecision (ie, sparse data) and rated as low-quality evidence. These trials could be further downgraded to very low–quality evidence if limitations of study design were found (PEDro score <7). Two reviewers (D.S. or M.J.H. with V.C.O. or M.C.) judged whether these factors were present for each outcome. The quality of evidence was defined as (1) high (further research is unlikely to change our confidence in the estimate of effect and there are no known or suspected reporting biases: all domains are fulfilled); (2) moderate (further research is likely to have an important effect on our confidence in the estimate of effect and might change the estimate: 1 of the domains is not fulfilled); (3) low (further research is likely to have an important effect on our confidence in the estimate of effect and is likely to change the estimate: 2 of the domains are not fulfilled); and (4) very low (we are uncertain about the estimate: 3 of the domains are not fulfilled).19
Outcome data were extracted for short-term (follow-up evaluations ≤12 months) and long-term (follow-up evaluations >12 months) follow-up. When multiple time points fell into the same category, we used the longest follow-up period.
Trials considered homogeneous were grouped according to the prevention strategy, comparison group, outcome (LBP episode and sick leave), and outcome assessment time points (short-term and long-term). We calculated relative risks (RRs) and 95% CIs and used the random-effects model to pool estimates for each analysis obtained with Comprehensive Meta-analysis, version 2.2.064 (Biostat). For trials that did not report the sample size at the end of the follow-up period, we calculated the RR using the baseline sample size.
The initial electronic database search identified 6133 potentially eligible studies. After screening citations by title and abstract, we considered 159 potentially eligible studies for inclusion and retrieved full-text articles. Twenty-three published reports (21 different RCTs including 30 850 unique participants) met the inclusion criteria and were included in this review.20-42 Two RCTs were reported in 4 articles22,30,39,40 (2 with 12-month data22,39 and 2 with 36-month data30,40). Figure 1 outlines the flow of RCTs through the review.
The included trials investigated 6 different LBP prevention strategies: exercise, education, exercise and education, back belts, shoe insoles, and other prevention strategies. Most of the trials focused largely or completely on working-age populations. The sample size of the trials ranged from 30 to 4325 participants. A comprehensive description of each trial is provided in Table 1.
Methodologic quality assessment was conducted using the PEDro scale. The mean (SD) score was 5.1 (1.5), with the key problem items being blinding, concealed allocation, and loss to follow-up (eTable 2 in the Supplement).
Estimates of the effects of LBP prevention strategies on LBP episode or sick leave due to LBP were calculated for 21 trials. The number of events, sample size, and RRs (95% CIs) for the trials are presented in eTable 3 in the Supplement. Trials were grouped according to the prevention strategy, outcome (episode of LBP or sick leave), and follow-up time point (short- or long-term). Table 2 provides a summary of the findings and GRADE quality ratings.
Four trials reporting data on 898 participants were included in the meta-analysis to estimate the short-term (ie, ≤12 months) efficacy of exercise on incident cases of LBP (presented as RR [95% CI]).21,25,26,42 The pooled results provide low-quality evidence of a protective effect of exercise (0.65 [0.50-0.86]). In the long-term (ie, >12 months), the pooled results of 2 trials (334 participants) provide very low–quality evidence of no effect of exercise (1.04 [0.73-1.49]) (Figure 2).21,33 Two trials presented data from 128 participants and provide very low–quality evidence that exercise reduces the risk of sick leave due to LBP in the long-term (0.22 [0.06-0.76]) (Figure 3).30,42
The effect of exercise and education was investigated in 4 trials (442 participants) at short-term follow-up,22,35,39,42 and in 2 trials (138 participants) at long-term follow-up (LBP episode).30,40 The pooled results (presented as RR [95% CI]) of 4 trials provide moderate-quality evidence that exercise and education reduce the risk of an episode of LBP at short-term follow-up (0.55 [0.41-0.74]). The long-term results are based on 2 trials30,40 and provide low-quality evidence of a protective effect (0.73 [0.55-0.96]) (Figure 2).
For prevention of sick leave due to LBP, 3 trials (228 participants)22,39,42 presented short-term data and 2 trials (138 participants)30,40 presented long-term data. The pooled results (presented as RR [95% CI]) provide low-quality evidence of no protective effect at short-term follow-up (0.74 [0.44-1.26]) or long-term follow-up (0.72 [0.48-1.08]) (Figure 3).
The efficacy of education compared with control was investigated in 3 trials (2343 participants) at short-term follow-up and in 2 trials (13242 participants) at long-term follow-up (LBP episode). The pooled results (presented as RR [95% CI]) provide moderate-quality evidence of no protective effect of education at either short-term follow-up (1.03 [0.83-1.27])37,41,42 or long-term follow-up (0.86 [0.72-1.04])20,34 (Figure 2). In addition, a single trial (3597 participants) not included in the meta-analysis because it did not report raw data provides moderate-quality evidence of no protective effect of education at long-term follow-up (rate ratio, 1.11 [95% CI, 0.90-1.37]) (eTable 3 in the Supplement).28
Two trials (366 participants)41,42 presented short-term data on sick leave prevention. The pooled results provide very low–quality evidence of no protective effect of education on sick leave due to LBP at short-term follow-up (RR, 0.87 [95% CI, 0.47-1.60]) (Figure 3).
The efficacy of back belts over control to prevent LBP episodes (short- and long-term) or sick leave owing to LBP (short-term) was reported in 3 trials.27,34,41 For episodes of LBP, pooling of 2 trials (329 participants) (presented as RR [95% CI]) provides very low–quality evidence of no short-term effect of back belts over controls (1.01 [0.71-1.44]) (Figure 2).27,41 At long-term follow-up, a single trial (8472 participants) provides moderate-quality evidence that back belts do not reduce the risk of LBP episodes when compared with controls (0.85 [0.64-1.14]) (Figure 2).34 For sick leave owing to LBP, a single trial (282 participants) provides low-quality evidence of no effect of back belts compared with controls at short-term follow-up (RR, 1.44 [95% CI, 0.73-2.86]) (Figure 3).41
Four trials reported data from 1833 participants on the short-term efficacy of shoe insoles compared with controls.23,24,36,38 For prevention of episodes of LBP, there is low-quality evidence that shoe insoles are not superior to control at short-term follow-up (RR, 1.01 [95% CI, 0.74-1.40]) (Figure 2). One trial reported the efficacy of semirigid shoe insole vs control and soft shoe insole vs control.24 Only the group from the semirigid shoe insole was included in the meta-analysis.
Two trials reported the short-term effect of other prevention strategies vs control for LBP episode (3047 participants),29 and sick leave due to LBP (360 participants).32 An ergonomic program (moderate-quality evidence) was not more effective than control in reducing episodes of LBP at short-term follow-up (odds ratio, 1.23 [95% CI, 0.97-1.57]) (Table 2). It is unclear whether sick leave due to LBP can be prevented by education, training, and ergonomic adjustments since there was very low–quality evidence (RR, 0.95 [95% CI, 0.51-1.76]) (Figure 3).
Quiz Ref IDThe results of this systematic review and meta-analysis indicate that exercise in combination with education is likely to reduce the risk of LBP. Exercise alone may reduce the risk of an episode of LBP and sick leave; however, it is uncertain whether the effects persist beyond 1 year.Quiz Ref ID Education alone, back belts, shoe insoles, and ergonomic adjustments probably do not prevent an episode of LBP or sick leave due to LBP. It is uncertain whether education, training, or ergonomic adjustments prevent LBP owing to the very low quality of evidence.
The strengths of this review include the use of a prespecified protocol registered on PROSPERO, inclusion of all prevention strategies from any setting, the use of the GRADE system to evaluate the overall quality of the evidence, and the use of a highly sensitive search strategy to identify LBP prevention trials. We assessed trials’ methodologic quality with the PEDro scale, which has been shown to have acceptable reliability and validity.15,16,44 All scores were available online at the PEDro website. These scores were rated by experienced PEDro researchers, which provided less chance of errors.
Quiz Ref IDThis review was designed to be comprehensive with a robust search strategy; however, it is possible that not all studies were identified. Some identified trials did not have the term prevention in either the title or the abstract.27,33,34 For several prevention strategies, we could identify only a small number of trials; this combined with the quality of the trials means the level of evidence for several prevention strategies is very low or low.
To our knowledge, this review is the first to have included a variety of LBP prevention strategies and conducted a meta-analysis of RCTs. Several reviews have investigated the effectiveness of an exercise and/or education program on LBP prevention. All are out-of-date, included at least 1 RCT with symptomatic participants at baseline (ie, the trial evaluated treatment, not prevention), and presented data descriptively.7-9,45-49 The most recent review we know of investigating the effectiveness of exercise for preventing a LBP episode,11 presented data from 3 trials. One was included in the meta-analysis of the current review (ie, exercise vs control),33 one was excluded because the trial included symptomatic participants at baseline,50 and one was included in a different LBP prevention strategy (ie, exercise and education vs control).39 That review by Choi et al11 reported a 50% (2 RCTs with 130 patients) reduction in future LBP episodes when compared with no intervention, which is a larger effect than our estimate of a 35% reduction (4 RCTs with 898 patients).
Previous reviews investigating the efficacy of exercise on the prevention of LBP episodes have not distinguished between studies that included education with the exercise from those just including exercise.11,45,46 In our review, the combination of exercise and education was effective at long-term follow-up (RR, 0.73 [95%CI, 0.55 to 0.96]), while exercise alone was not (RR, 1.04 [95% CI, 0.73 to 1.49]), suggesting that the distinction between exercise alone and exercise combined with education may be important.
Quiz Ref IDThe present review’s finding that back belts do not prevent LBP is consistent with results of a previous systematic review.51 There are a few previous systematic reviews10,45,52 investigating the use of shoe insoles in the prevention of an LBP episode. Findings from these reviews are in line with the results of our study: shoe insoles are not effective for the prevention of back pain. The most recent review by Chuter et al10 included 6 trials; our review included 4. We excluded 2 trials because the participants were symptomatic at the time of study entry.53,54
Although our review found evidence for both exercise alone (35% risk reduction for an LBP episode and 78% risk reduction for sick leave) and for exercise and education (45% risk reduction for an LBP episode) for the prevention of LBP up to 1 year, we also found the effect size reduced (exercise and education) or disappeared (exercise alone) in the longer term (>1 year). This finding raises the important issue that, for exercise to remain protective against future LBP, it is likely that ongoing exercise is required. Prevention programs focusing on long-term behavior change in exercise habits seem to be important.
The results of this systematic review and meta-analysis of RCTs indicate that exercise in combination with education is likely to reduce the risk of LBP and that exercise alone may reduce the risk of an episode of LBP and sick leave due to LBP, at least for the short-term. The available evidence suggests that education alone, back belts, shoe insoles, and ergonomics do not prevent LBP. It is uncertain whether education, training, or ergonomic adjustments prevent sick leave due to LBP because the quality of evidence is very low.
Corresponding Author: Daniel Steffens, PhD, Musculoskeletal Division, The George Institute for Global Health, Sydney Medical School, The University of Sydney, GPO Box 5389, Sydney, New South Wales, Australia 2000 (email@example.com).
Accepted for Publication: November 9, 2015.
Published Online: January 11, 2016. doi:10.1001/jamainternmed.2015.7431.
Author Contributions: Drs Steffens and Maher had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Steffens, Maher, Stevens, Teixeira-Salmela, Hancock.
Acquisition, analysis, or interpretation of data: Steffens, Maher, Pereira, Oliveira, Chapple, Teixeira-Salmela, Hancock.
Drafting of the manuscript: Steffens, Maher, Teixeira-Salmela, Hancock.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Steffens, Maher, Chapple.
Administrative, technical, or material support: Steffens, Oliveira, Teixeira-Salmela, Hancock.
Study supervision: Steffens, Maher, Pereira, Teixeira-Salmela, Hancock.
Conflict of Interest Disclosures: None reported.