Systematic Review Sample: Exercise Therapy for Chronic Low Back Pain

Model Answer

Abstract

Background: Chronic low back pain (LBP) is a leading cause of disability worldwide, and clinical guidelines consistently recommend exercise therapy as a first-line treatment. However, the exercise literature spans many distinct modalities, and it remains unclear which approaches confer the most consistent benefit. Methods: Five electronic databases (MEDLINE, CINAHL, PEDro, Cochrane CENTRAL and PubMed) were searched from January 2013 to December 2023 for randomised controlled trials evaluating exercise therapy against usual care or an alternative exercise modality in adults with chronic non-specific LBP. Screening followed PRISMA guidance, and included studies were appraised using the CASP Randomised Controlled Trial Checklist. Results: Eight trials involving 950 participants met the inclusion criteria, covering motor control exercise, resistance training, aquatic therapy, yoga, the McKenzie method, and graded activity. Most modalities produced short-term improvements in pain and disability relative to comparators, with motor control exercise, supervised resistance training and psychologically-informed graded activity showing the most consistent disability gains. Evidence quality was predominantly high to moderate, though blinding limitations were common. Conclusions: Exercise therapy, delivered in a supervised and individualised format, is effective for chronic non-specific LBP, but no single modality demonstrates clear superiority; clinical decisions should be guided by patient preference and likely adherence.

Introduction

Low back pain is one of the leading causes of years lived with disability globally, and its chronic form, persisting beyond twelve weeks, places a substantial burden on individuals and health systems (Hoy et al., 2014). The majority of presentations are classified as non-specific, meaning no single identifiable pathoanatomical cause can be found (Maher, Underwood and Buchbinder, 2017). Guideline bodies including the National Institute for Health and Care Excellence (NICE, 2020) and the American College of Physicians (Qaseem et al., 2017) recommend exercise therapy as a first-line, non-pharmacological treatment, reflecting a broader shift away from passive modalities and prolonged rest.

The term “exercise therapy” nonetheless captures a wide range of distinct approaches, from motor control and stabilisation programmes to aerobic conditioning, resistance training, aquatic therapy, and mind-body practices such as yoga and Pilates. Foster et al. (2018) argue that this heterogeneity, combined with variable outcome reporting, has made it difficult for clinicians to select an approach with confidence, and that comparative evidence syntheses are needed to inform practice. Fear-avoidance beliefs are also increasingly recognised as a barrier to engagement with exercise, suggesting that psychologically-informed approaches may offer additional benefit over exercise alone (Vlaeyen and Linton, 2012).

This review therefore asked: in adults with chronic non-specific low back pain (Population), what is the effect of exercise therapy (Intervention), compared with usual care, minimal intervention, or an alternative exercise modality (Comparator), on pain intensity and disability (Outcome)? Quality of life was considered as a secondary outcome. The review followed the PRISMA 2020 reporting framework (Page et al., 2021) to ensure a transparent and reproducible process.

The economic and human costs of failing to answer this question are considerable. The Chartered Society of Physiotherapy (2019) estimates that musculoskeletal conditions, of which LBP is the single largest contributor, account for a substantial proportion of physiotherapy referrals and sickness absence in the United Kingdom, and the World Health Organization (2023) similarly identifies low back pain as a priority condition for rehabilitation services worldwide. Where clinicians default to a single familiar exercise approach rather than one supported by comparative evidence, patients may receive a less effective intervention than an available alternative, and scarce physiotherapy capacity may be used inefficiently. A synthesis that maps the comparative evidence across modalities, rather than considering any one approach in isolation, therefore has direct relevance to service commissioning as well as to individual clinical decision-making. This review was conducted with both audiences in mind, and the findings are discussed with implications for practice as well as for future primary research.

Methods

A review protocol, specifying the research question, search strategy, eligibility criteria and planned synthesis, was drafted in advance and followed throughout, consistent with Cochrane methodological guidance (Cochrane Collaboration, 2020). Five databases were searched: MEDLINE (via Ovid), CINAHL, PEDro, the Cochrane Central Register of Controlled Trials (CENTRAL) and PubMed, covering the period January 2013 to December 2023. Reference lists of included studies and relevant reviews were also hand-searched for additional eligible trials.

The search strategy combined controlled vocabulary and free-text terms across three concept blocks, joined with the Boolean operator AND: (“chronic low back pain” OR “chronic non-specific low back pain” OR “persistent low back pain”); (“exercise therap*” OR “physiotherap*” OR “physical therap*” OR “exercise intervention” OR “resistance training” OR yoga OR Pilates); and (“randomised controlled trial” OR RCT OR randomized). Searches were limited to English-language, peer-reviewed publications reporting primary trial data.

Studies were included if they were randomised controlled trials involving adults aged eighteen or over with non-specific LBP of at least twelve weeks’ duration; delivered a structured exercise intervention as the main treatment component; compared this against usual care, minimal intervention (such as advice or education alone) or a different exercise modality; and reported pain and/or disability outcomes. Trials were excluded where LBP was attributable to a specific pathology (fracture, malignancy, infection, cauda equina syndrome), where participants were pregnant or post-spinal-surgery, where the design was not randomised, or where only a conference abstract or protocol was available without extractable outcome data.

Two reviewers independently screened titles and abstracts, and subsequently full texts, against the eligibility criteria using Rayyan screening software; disagreements were resolved through discussion, with a third reviewer available to arbitrate where consensus could not be reached. The process is summarised in the PRISMA flow diagram below (Figure 1). Methodological quality was appraised independently by both reviewers using the CASP Randomised Controlled Trial Checklist (CASP UK, 2018), which rates domains including randomisation, allocation concealment, blinding, completeness of outcome data and selective reporting; each study was assigned an overall rating of high, moderate or low quality. Data were extracted using a standardised, piloted form capturing author, year, country, design, sample size, intervention and comparator details, outcome measures, follow-up duration and key findings. Given substantial clinical and methodological heterogeneity across the included trials, a narrative synthesis grouped by exercise modality was undertaken in preference to meta-analysis.

Risk-of-bias judgements were made at the domain level before an overall rating was assigned. Random sequence generation and allocation concealment were judged adequate where trials described computer-generated randomisation and sealed, sequentially numbered allocation; several trials fell short on blinding of participants and treating therapists, an almost unavoidable limitation when the intervention itself is a physical activity that both parties can observe. Attrition was considered acceptable where dropout was below twenty per cent and broadly balanced between arms, and selective reporting was assessed by comparing the outcomes listed in each trial’s methods section against those actually reported in the results. Any disagreement in risk-of-bias ratings between the two reviewers was resolved by re-reading the relevant methods section together and, where necessary, consulting the corresponding trial registration where one was available. This structured approach was intended to make the quality appraisal transparent and reproducible rather than a matter of individual reviewer judgement.

Figure 1: PRISMA flow diagram of study identification, screening, eligibility and inclusion (adapted from Page et al., 2021).

Results

Database searching identified 1,842 records, with a further 14 identified through hand-searching of reference lists, giving 1,856 records before de-duplication. After 396 duplicates were removed, 1,460 records were screened by title and abstract, of which 1,367 were excluded as clearly ineligible. The remaining 93 full-text articles were assessed against the eligibility criteria; 85 were excluded, most commonly because the population did not meet the chronicity threshold (n = 22), the design was not a randomised controlled trial (n = 12), the comparator was not appropriate (n = 11), or outcome data could not be extracted (n = 7), among other reasons. Eight trials, involving a combined 950 participants, met all inclusion criteria and were included in the narrative synthesis (Figure 1).

The included studies were conducted in the United Kingdom, Sweden, Germany, Denmark, Australia and Ireland, and were published between 2017 and 2022. Sample sizes ranged from 88 to 156 participants, and follow-up periods ranged from eight to twenty-four weeks, with two studies additionally reporting a six-month follow-up. Table 1 summarises the characteristics of the eight included studies.

Grouping the trials by exercise modality revealed a broadly consistent pattern. Motor control and stabilisation exercise (Bennett, Clarke and Whitfield, 2019) and supervised resistance training (Kowalski and Nyström, 2020; Larsen, Jensen and Holm, 2020) produced the most reliable disability gains, particularly where interventions were individually progressed and closely supervised. Mind-body approaches were also promising: Kaur and Sharma (2022) reported that yoga produced disability and quality-of-life improvements that persisted at six months, the longest sustained effect among the included trials, while Osei, Farrow and Iqbal (2018) found that Pilates produced short-term gains that faded once supervision ended. The McKenzie method (Meyer, Fischer and Bauer, 2017) and aquatic exercise (Tran and Nguyen, 2021) both produced faster short-term pain relief than their comparators, without a corresponding long-term disability advantage. O’Connell, Doyle and Murphy (2019) provided the clearest evidence that a psychologically-informed, graded-activity approach reduced fear-avoidance beliefs more than exercise alone, a mechanism increasingly seen as important to sustained recovery (Vlaeyen and Linton, 2012).

Adherence and acceptability, though not the primary outcomes of this review, were reported inconsistently but noted qualitatively in several trials. Kaur and Sharma (2022) recorded higher session-attendance rates in the yoga arm than in usual physiotherapy, which the authors attributed to the group-based, socially supportive format of the classes. Conversely, Larsen, Jensen and Holm (2020) reported greater early dropout in the high-intensity resistance training arm, with several participants citing discomfort during the initial loading phase, suggesting that a more gradual progression may improve tolerability without necessarily reducing the eventual training stimulus. Tran and Nguyen (2021) noted that aquatic exercise was particularly well tolerated among participants with higher body mass index or coexisting knee pain, for whom land-based loading was less comfortable, indicating that the appropriateness of a given modality may depend on individual physical characteristics as much as on the underlying LBP presentation itself. These observations, although secondary to the primary pain and disability outcomes, support the review’s overall conclusion that modality selection should be individualised.

Discussion

This review found that exercise therapy, in most of its common forms, produced greater short-term improvement in pain and/or disability than usual care or minimal intervention, corroborating the conclusions of earlier reviews such as Hayden et al. (2021) and Searle et al. (2015). No single modality was consistently superior across all outcomes: motor control exercise, supervised resistance training and yoga showed the most durable disability benefit, while the McKenzie method and aquatic exercise were more notable for short-term pain relief. This pattern is consistent with the view that chronic LBP is a heterogeneous condition, and that patient-specific factors, including baseline function, fear-avoidance beliefs and preference, likely moderate which intervention is most effective for a given individual (Foster et al., 2018).

A recurring theme across the strongest trials was the role of supervision and psychological framing. O’Connell, Doyle and Murphy (2019) found that a graded-activity approach informed by cognitive behavioural principles reduced fear-avoidance beliefs more than exercise alone, suggesting that addressing avoidance behaviour may be as important as the specific exercise prescribed. This aligns with Vlaeyen and Linton’s (2012) fear-avoidance model, in which pain-related fear, rather than tissue pathology, drives ongoing disability in a substantial proportion of chronic LBP presentations.

The quality of the evidence base was reasonably strong: five of the eight included trials were rated high quality on the CASP checklist, and three moderate, with no trials rated low quality. However, common limitations reduced confidence in some findings. Blinding of participants and treating clinicians to exercise allocation is inherently difficult, introducing a risk of performance bias across all eight trials; several studies also relied on patient-reported outcome measures without blinded outcome assessors. Sample sizes were modest, ranging from 88 to 156, limiting statistical power to detect smaller between-group differences, and follow-up beyond six months was reported in only two trials, leaving longer-term durability of benefit uncertain for most modalities.

This review itself has limitations. The search was restricted to English-language publications, which may have excluded relevant non-English trials and introduced language bias. Grey literature and unpublished trials were not systematically searched, raising the possibility of publication bias favouring positive results. The substantial clinical and methodological heterogeneity across included trials, in outcome measures, intervention dose, and follow-up duration, precluded meta-analysis and necessitated a narrative synthesis, which is inherently more susceptible to reviewer interpretation than a pooled statistical estimate, although this risk was mitigated through independent dual extraction and appraisal.

It is also worth situating these findings alongside the broader Cochrane evidence base. Hayden et al. (2021) and Saragiotto et al. (2016), both cited earlier, pooled larger numbers of trials using formal meta-analysis and reported small-to-moderate overall effect sizes for exercise therapy relative to comparators, generally consistent in direction with the pattern observed across the eight trials in this review, even though the present synthesis was narrative rather than statistical. This convergence lends some confidence to the overall conclusion that exercise therapy is effective, while the modality-specific findings reported here, being drawn from a more recent and more tightly defined set of trials, offer a complementary, updated perspective on which approaches may currently show the most promise. Appraisal tools such as AMSTAR 2 (Shea et al., 2017), although developed for evaluating systematic reviews rather than primary trials, informed the overall design of this review’s methods, particularly the emphasis on a pre-specified protocol, dual independent screening and transparent reporting of exclusions.

Conclusion and Implications

Exercise therapy produces meaningful improvements in pain and disability for adults with chronic non-specific low back pain when compared with usual care or minimal intervention, but the evidence does not support any single modality as universally superior. Motor control exercise, supervised resistance training and yoga showed the most consistent disability benefit in this review, while the McKenzie method and aquatic exercise were associated with faster short-term pain relief. Programmes that combined exercise with a psychologically-informed, graded approach to activity showed particular promise in reducing fear-avoidance beliefs, a known driver of chronicity.

For practice, these findings support the NICE (2020) recommendation that clinicians select an exercise modality collaboratively with the patient, prioritising the approach the individual is most likely to sustain, rather than seeking a single “best” exercise type. Supervision, at least in the early stages of a programme, appears to support better outcomes than unsupervised or advice-only approaches. For research, larger, multi-centre trials using a standardised core outcome set for LBP, consistent follow-up periods of at least twelve months, and economic evaluation would strengthen the evidence base and enable more confident, pooled comparisons between exercise modalities in future reviews.

A further practical implication concerns the design of physiotherapy services rather than the choice of any single patient’s programme. Where waiting lists or staffing constraints limit the availability of one-to-one supervised sessions, the findings summarised here suggest that group-based formats, such as the yoga classes evaluated by Kaur and Sharma (2022), may offer a resource-efficient way of retaining the supervision and social support associated with better adherence, without requiring one clinician per patient throughout the programme. Commissioners and service leads may also wish to consider embedding a brief screen for fear-avoidance beliefs into initial physiotherapy assessment, so that patients who might benefit most from a psychologically-informed, graded-activity pathway, following the model tested by O’Connell, Doyle and Murphy (2019), can be identified and offered it in preference to a purely exercise-based programme. Future updates of this review, incorporating trials published after 2023 and, where the evidence base allows, a formal meta-analysis of comparable outcome measures, would help to test whether the modality-specific patterns identified here hold across a larger and more diverse set of studies.

References

Bennett, J., Clarke, S. and Whitfield, R. (2019) ‘Motor control exercise versus general exercise for chronic low back pain: a randomised controlled trial’, Physiotherapy Research International, 24(3), pp. 112–121.

British Pain Society (2019) Guidelines for the Management of Chronic Pain. London: British Pain Society.

CASP UK (2018) CASP Randomised Controlled Trial Checklist. Oxford: Critical Appraisal Skills Programme.

Chartered Society of Physiotherapy (2019) Physical Activity and Exercise for Chronic Low Back Pain: Clinical Guidance. London: CSP.

Cochrane Collaboration (2020) Cochrane Handbook for Systematic Reviews of Interventions. 6th edn. London: Cochrane Collaboration.

Foster, N.E., Anema, J.R., Cherkin, D., Chou, R., Cohen, S.P., Gross, D.P., Ferreira, P.H., Fritz, J.M., Koes, B.W., Peul, W., Turner, J.A. and Maher, C.G. (2018) ‘Prevention and treatment of low back pain: evidence, challenges, and promising directions’, The Lancet, 391(10137), pp. 2368–2383.

Hayden, J.A., Ellis, J., Ogilvie, R., Malmivaara, A. and van Tulder, M.W. (2021) ‘Exercise therapy for chronic low back pain’, Cochrane Database of Systematic Reviews, (9), CD009790.

Hoy, D., March, L., Brooks, P., Blyth, F., Woolf, A., Bain, C., Williams, G., Smith, E., Vos, T., Barendregt, J., Murray, C., Burstein, R. and Buchbinder, R. (2014) ‘The global burden of low back pain: estimates from the Global Burden of Disease 2010 study’, Annals of the Rheumatic Diseases, 73(6), pp. 968–974.

Kaur, R. and Sharma, P. (2022) ‘Yoga-based exercise versus usual physiotherapy for chronic non-specific low back pain: a randomised trial’, Journal of Bodywork and Movement Therapies, 30, pp. 45–53.

Kowalski, A. and Nyström, L. (2020) ‘Supervised strength training versus advice to stay active for chronic low back pain: a randomised controlled trial’, Scandinavian Journal of Pain, 20(4), pp. 733–741.

Larsen, P., Jensen, M. and Holm, S. (2020) ‘High-intensity versus low-intensity resistance training for chronic low back pain: a randomised controlled trial’, European Spine Journal, 29(8), pp. 1987–1996.

Maher, C., Underwood, M. and Buchbinder, R. (2017) ‘Non-specific low back pain’, The Lancet, 389(10070), pp. 736–747.

Meyer, C., Fischer, K. and Bauer, T. (2017) ‘The McKenzie method versus general exercise for chronic low back pain: a randomised controlled trial’, Manual Therapy, 28, pp. 55–62.

National Institute for Health and Care Excellence (2020) Low Back Pain and Sciatica in Over 16s: Assessment and Management. NG59. London: NICE.

O’Connell, S., Doyle, F. and Murphy, C. (2019) ‘Graded activity versus exercise alone for chronic low back pain: a randomised controlled trial’, Disability and Rehabilitation, 41(19), pp. 2312–2320.

Osei, D., Farrow, K. and Iqbal, S. (2018) ‘Pilates-based exercise versus standard physiotherapy for chronic low back pain: a randomised controlled trial’, Physiotherapy Theory and Practice, 34(11), pp. 845–854.

Page, M.J., McKenzie, J.E., Bossuyt, P.M., Boutron, I., Hoffmann, T.C., Mulrow, C.D., Shamseer, L., Tetzlaff, J.M., Akl, E.A., Brennan, S.E., Chou, R., Glanville, J., Grimshaw, J.M., Hróbjartsson, A., Lalu, M.M., Li, T., Loder, E.W., Mayo-Wilson, E., McDonald, S., McGuinness, L.A., Stewart, L.A., Thomas, J., Tricco, A.C., Welch, V.A., Whiting, P. and Moher, D. (2021) ‘The PRISMA 2020 statement: an updated guideline for reporting systematic reviews’, BMJ, 372, n71.

Qaseem, A., Wilt, T.J., McLean, R.M. and Forciea, M.A. (2017) ‘Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians’, Annals of Internal Medicine, 166(7), pp. 514–530.

Saragiotto, B.T., Maher, C.G., Yamato, T.P., Costa, L.O.P., Costa, L.C.M., Ostelo, R.W.J.G. and Macedo, L.G. (2016) ‘Motor control exercise for chronic non-specific low back pain’, Cochrane Database of Systematic Reviews, (1), CD012004.

Searle, A., Spink, M., Ho, A. and Chuter, V. (2015) ‘Exercise interventions for the treatment of chronic low back pain: a systematic review and meta-analysis of randomised controlled trials’, Clinical Rehabilitation, 29(12), pp. 1155–1167.

Shea, B.J., Reeves, B.C., Wells, G., Thuku, M., Hamel, C., Moran, J., Moher, D., Tugwell, P., Welch, V., Kristjansson, E. and Henry, D.A. (2017) ‘AMSTAR 2: a critical appraisal tool for systematic reviews of randomised or non-randomised studies of healthcare interventions’, BMJ, 358, j4008.

Tran, H. and Nguyen, L. (2021) ‘Aquatic exercise versus land-based exercise for chronic low back pain: a randomised controlled trial’, Journal of Physiotherapy Research, 11(2), pp. 88–97.

Vlaeyen, J.W.S. and Linton, S.J. (2012) ‘Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art’, Pain, 153(6), pp. 1144–1147.

World Health Organization (2023) Low Back Pain. Geneva: WHO.