Exercise therapy for treatment of non‐specific low back pain

Types of studies

We included published reports of completed randomized controlled trials.

Types of participants

We included studies involving adult participants with acute (less than six weeks), subacute (six to 12 weeks), or chronic (longer than 12 weeks) non‐specific low‐back pain. We excluded studies that involved individuals with low‐back pain caused by specific pathologies or conditions.

Types of interventions

Exercise therapy was defined as "a series of specific movements with the aim of training or developing the body by a routine practice or as physical training to promote good physical health" (Abenhaim 2000). We included studies that compared exercise therapy to a) no treatment or placebo treatment, b) other conservative therapy, or c) another exercise group.

Types of outcome measures

Outcomes of interest were self‐reported pain intensity, condition‐specific physical functioning and global improvement, and return to work/absenteeism. Outcome assessment data were abstracted for three time periods: short‐term (post‐treatment assessment closest to six weeks after randomization, not longer than 12 weeks), intermediate (six months), and long‐term follow‐up (12 months or more).

Study selection and data abstraction

A standard protocol was followed for study selection and data abstraction (van Tulder 2003). This included two reviewers' independent assessment of study eligibility, data extraction, assessment of trial quality and clinical relevance. Consensus and, if necessary, a third reviewer were used to resolve disagreements. We extracted population characteristics (patient population source or setting, study inclusion criteria, duration of low‐back pain episode, and age of patients), intervention characteristics (description and types of exercise therapy, duration and number of treatment sessions, intervention delivery type, and co‐interventions) outcome data, and overall conclusions about the effectiveness of the exercises onto pre‐tested standardized forms. Assessment of quality included: appropriate randomization, adequate concealment of treatment allocation, adequacy of follow‐up, and outcome assessment blinding (Jadad 1996). High quality studies were defined as those in which all of these key quality criteria were met. Clinical relevance of each trial was assessed with four items: participants described in detail to assess clinical comparability, interventions and treatment settings adequately described to allow repetition, clinically relevant outcomes measured and reported, and are likely treatment benefits worth potential harms. Reviewers were not blinded to authors, institution or journal of publication due to feasibility and because they were familiar with most of the literature. Authors of published trials were contacted to clarify or provide additional information if the study provided insufficient information.

Analysis

We discussed the analyses of study results with clinical content experts. We synthesized the earliest outcomes provided for acute, subacute and chronic low‐back pain, comparing exercise to no treatment and to other conservative treatment, and overall for short, intermediate and long‐term follow‐up periods. Due to important gaps in the reporting of return‐to‐work/absenteeism and global assessment, quantitative analyses were only possible for pain and functioning outcomes. In the low‐back pain literature, several outcome measures are used to assess the constructs of pain intensity (for example, 10 mm or 100 mm visual analogue scales [VAS], or 0 to 10 numerical rating scale [NRS]) (see recent review by Von Korff et al (von Korff 2000)) and condition‐specific functioning (for example, the 24‐point Roland Morris Disability Questionnaire, or the Oswestry Disability Index scored out of 100) (see recent review by Kopec (Kopec 2000)). There is moderate to high correlation between the different measures of the two constructs. In this review, individual trial outcomes for pain and functioning were re‐scaled to 0 to 100 points [for example a VAS pain score (standard deviation) of 5.1 (2.3) out of 10 was re‐scaled to 51 (23) out of 100], where positive mean effect sizes indicated improvement (i.e. decreased pain, and decreased functional limitations). Re‐scaling is common (Kopec 2000) and facilitates comparison and interpretability of the syntheses. On the basis of current literature on minimal clinically important differences, we considered that a 20‐point (/100) improvement in pain (Salaffi 2004) and a 10‐point (/100) improvement in functioning outcomes (Bombardier 2001) were clinically important. Differences were considered statistically significant at the five percent level. The adequacy of sample size to detect these differences in each trial was assessed assuming a power of 90%.

To be consistent with the previous review and to allow more complete use of available data, we used both a qualitative rating system and quantitative meta‐analyses. The latter were conducted by pooling weighted mean differences with random effects models and data from at least three studies (DerSimonian 1986). Exercise treatment groups from included trials were included in the syntheses if they had an independent no treatment or other conservative treatment comparison group. This requirement appropriately meant studies with no comparison group (i.e. trials that contrast multiple exercise therapy groups only) were not included and comparison groups were not "double counted" in the meta‐analyses. This latter criteria is necessary to avoid correlation in effect sizes resulting from the use of repeated comparison data. We extracted data on means or median follow‐up outcomes for study groups. To maximize the available data, missing variance scores were imputed using the mean variance from studies with similar duration. Sensitivity analyses were conducted to assess the impact of excluding studies reporting median values and excluding studies that did not adequately present variance scores. Statistical heterogeneity was assessed using I² statistics and confidence intervals (Higgins 2002). Publication bias was evaluated with Egger's test and funnel plots (Egger 1997).

Qualitative assessment of results was based on primary outcome measures and considered the methodological quality and the reviewers' overall conclusions for each exercise therapy group. Exercise therapy groups were included in the qualitative synthesis if the trial included a no treatment or other conservative treatment comparison group. Two reviewers independently rated the findings for each exercise therapy group. Studies were considered to be providing evidence of effectiveness if statistically significant improvement was observed in at least one of the key outcomes in favour of the exercise group and clinically important improvement was observed within or between groups. Studies were considered to be providing evidence that the exercise therapy was ineffective if there was statistically significant improvement of the comparison group and no clinically important improvement within the exercise group. We rated studies neutral if no statistically and clinically significant results were observed and unclear if insufficient data were presented. A consensus process was used to examine patterns in trial results. Levels of evidence, as recommended by the Back Group (van Tulder 2003), were used:

1. Strong evidence ‐ consistent findings* in multiple high quality trials

2. Moderate evidence ‐ consistent findings in multiple low quality trials and/or one high quality trial

3. Limited evidence ‐ one low quality trial

4. Conflicting evidence ‐ inconsistent findings in multiple trials.

5. No evidence ‐ no randomized trials available.

*Consistent findings were defined as 75% or more trials (66% in sensitivity analysis) showing results in the same direction.

Further analyses explored heterogeneity due to study‐level variables, such as population source and study quality. We characterized the population sources as healthcare (primary, secondary or tertiary care centres), occupational (patients presenting to occupational healthcare facilities or personnel in compensatory situations), or from a general or mixed population (e.g. including individuals recruited by newspaper advertisements) to differentiate the studies with patients in typical treatment settings (healthcare and occupational) from those including individuals with low‐back pain who may not normally present for treatment. Outcomes for subgroups of studies conducted in these populations were compared (Song 2003). The impact of study quality on effect sizes was assessed using subgroup analysis.

SAS for Windows Version 8 (for descriptive), STATA 8 (for publication bias), and Review Manager 4.2 packages were used for analyses.

Clinical relevance of the included studies

Assessment of clinical relevance found that many of the trial publications supplied inadequate information. The study population was adequately described by 90% of the publications, but only 54% adequately described the exercise intervention. There was adequate reporting of relevant outcomes in 70% of the trials. A small number of studies reported on the presence or absence of adverse events (16 studies, 26%). Twelve studies reported mild negative reactions to the exercise program, such as increased low‐back pain and muscle soreness, in a minority of patients. Due to limitations of reporting, it was not possible to assess the treatment benefit to harm ratio.

Effectiveness

Further Analyses

Analyses were conducted on studies from acute, subacute and chronic populations to assess the impact of study level variables. Test of statistical heterogeneity of pain outcomes found 57% (95% CI: 12 to 79), 37% (95% CI: 0 to 76) and 81% (95% CI: 72 to 87) for acute, subacute and chronic, respectively, of the heterogeneity not due to chance; function outcomes showed 80% (95% CI: 63 to 89), 47% (95% CI: 0 to 82) and 52% (95% CI: 19 to 71), respectively. To account for heterogeneity, random effects models were used and clinically relevant subgroups of studies investigated. A complete exploration of intervention heterogeneity is included in an earlier publication (Hayden 2005b).

Indirect subgroup comparisons using qualitative synthesis and meta‐analysis found trials examining healthcare study populations observed higher mean improvements in functioning and pain over their comparison groups than trials examining occupational or general populations. In chronic populations, there were mean improvements in healthcare settings of 13.3 points (95% CI: 5.5 to 21.1) on pain and 6.9 points (95% CI: 2.2 to 11.7) on function outcomes. The adjusted differences between studies with different source populations found significantly greater improvement in outcomes in healthcare populations compared to studies from general population or mixed populations, with a mean of 9.96 points (95% CI: 1.6 to 18.4) more improvement in pain, and 5.52 points (95% CI: 0.6 to 10.4) greater improvement in functioning.

Meta‐analyses were conducted on the subgroup of high quality trials. The observed effectiveness of exercise therapy decreased and only remained significant for pain outcomes in the chronic population.

Summary

My concerns with this review stem not from its methodology, but its objectives. To this point, I refer specifically to the treatment of 'exercise therapy' as a single form of treatment, rather than a wide‐ranging and multifaceted modality that requires specific prescription. 'Exercise therapy' can mean many things to many people, and not just the unqualified. This point is illustrated by the wide range of interventions your reviewed studies include. Unfortunately, the generalisation of 'exercise therapy' and its efficacy in the management of low back pain also seems to be reflected in current guidelines to practitioners such as those published in the UK by NICE and the RCGP.

In my opinion, reviewing 'exercise therapy' as a single form of treatment detracts greatly from the interpretation of the results of this review. The attempt to stratify the data into 'flexion and/or extension exercises' and 'strength exercises', although noble, gives the reader little more information about the nature of the exercises undertaken. Although I appreciate the problems with finding sufficient similar trials on 'specific exercises' (which can also be sub‐divided into many forms; McKenzie; transversus abdominus and multifidus retraining, etc.), I would argue that by design, these studies may be too heterogeneous to combine. If the analogy could be made with drug treatments, 'general treatment with medication' would just not cut it as an objective for a systematic review of the pharmacological management of any condition.

The area of exercise prescription in the treatment of sub‐acute and chronic low back pain has undergone some major developments in the last 10 years. Yet unfortunately, growing bodies of high quality research into specific exercise therapy for LBP, such as that undertaken initially by a group of physiotherapists from the University of Queensland in Australia (P Hodges, C Richardson, G Jull and J Hides, and repeated successfully by other authors ‐ transversus abdominus and multifidus retraining in the treatment of low back pain), has been ignored by a large proportion of the medical community undertaking clinical trials (the UK Beam trial which recently concluded is a good example of this) and funding bodies alike. Similar highly specific exercise such as advocated by these authors requires rigorous assessment by RCT's rather than the 'blanket response' to exercise therapy most research and current 'good practice guidelines' seem to be focusing on. Time after time, insensitive interventions such as 'exercise' are tested in low back pain sufferers with understandably conflicting results, serving only to confuse practitioners and patients and fruitlessly drain research funds.

I and many of my colleagues would argue strongly that 'exercise prescription' is as broad a term as 'drug prescription'; and as such it's assessment in clinical trials requires explicit and repeatable measures such as required in drug trials (explicit description, type, dose and side‐effects etc.). I would argue that until we can assess each explicit exercise form, we have a 'general idea' what exercise can do but no more. Considering their ramifications, recommendations such as those given in your article must be given with great caution. At the very least they must realise their own limitations to interpretation and external validity. I would certainly like to see references to 'exercise' in ALL RCT's on this topic narrowed in their definitions and I feel large RCT's and subsequent systematic reviews need to be undertaken to investigate the growing body of evidence supporting the effectiveness of highly specific exercises in the management of LBP.

Reply

You have identified an argument that rages within the systematic review field ... when to 'lump' and when to 'split'. All exercises are again included in the upcoming update of the exercise review. To this point, the research question has been 'is exercise of any benefit to individuals with low back pain' ... there has not been a breakdown of each type of exercise for each duration of symptoms, due in part to the lack of data for each comparison, once one starts breaking it down to this degree, although I think the authors may have attempted some sub‐group analysis. As the literature increases, I suspect it will become more feasible to split into different research questions, addressing the efficacy of specific exercises for specific sub‐groups of individuals with low back pain. There have been some attempts to do this, but the data is still sparse and results must be treated with caution. The authors recognize that this continues to pose a challenge to clinicians who deliver exercise therapy. We cannot comment on how the summary of the scientific literature is used in the development of guidelines, since guidelines must take more into consideration than just the available evidence.

I will pass on your comments to the authors of the updated review for their consideration. Please do not hesitate to contact me should you have any further concerns once the updated review is published.

Contributors

Michael Noonan, Occupation Physiotherapist/Medical Student 
 Victoria Pennick, Back Group Coordinator