Background: Previous systematic reviews have concluded that the effectiveness of motor control exercise for persistent low back pain has not been clearly established.
Objective: The objective of this study was to systematically review randomized controlled trials evaluating the effectiveness of motor control exercises for persistent low back pain.
Methods: Electronic databases were searched to June 2008. Pain, disability, and quality-of-life outcomes were extracted and converted to a common 0 to 100 scale. Where possible, trials were pooled using Revman 4.2.
Results: Fourteen trials were included. Seven trials compared motor control exercise with minimal intervention or evaluated it as a supplement to another treatment. Four trials compared motor control exercise with manual therapy. Five trials compared motor control exercise with another form of exercise. One trial compared motor control exercise with lumbar fusion surgery. The pooling revealed that motor control exercise was better than minimal intervention in reducing pain at short-term follow-up (weighted mean difference=−14.3 points, 95% confidence interval [CI]=−20.4 to −8.1), at intermediate follow-up (weighted mean difference=−13.6 points, 95% CI=−22.4 to −4.1), and at long-term follow-up (weighted mean difference=−14.4 points, 95% CI=−23.1 to −5.7) and in reducing disability at long-term follow-up (weighted mean difference=−10.8 points, 95% CI=−18.7 to −2.8). Motor control exercise was better than manual therapy for pain (weighted mean difference=−5.7 points, 95% CI=−10.7 to −0.8), disability (weighted mean difference=−4.0 points, 95% CI=−7.6 to −0.4), and quality-of-life outcomes (weighted mean difference=−6.0 points, 95% CI=−11.2 to −0.8) at intermediate follow-up and better than other forms of exercise in reducing disability at short-term follow-up (weighted mean difference=−5.1 points, 95% CI=−8.7 to −1.4).
Conclusions: Motor control exercise is superior to minimal intervention and confers benefit when added to another therapy for pain at all time points and for disability at long-term follow-up. Motor control exercise is not more effective than manual therapy or other forms of exercise.
Low back pain (LBP) is one of the main causes of disability, and, despite its high prevalence, the source of pain is not established in the majority of cases and the term “nonspecific low back pain” is used.1–4 One factor that has been proposed as important in the genesis and persistence of nonspecific LBP is stability and control of the spine.4 Studies of individuals with LBP have identified impairments in the control of the deep trunk muscles (eg, transversus abdominis and multifidus) responsible for maintaining the stability of the spine.5–8 For example, activity of the transversus abdominis muscles9 and the multifidus muscles7 is delayed during arm movements (that challenge the stability of the spine) in individuals with LBP. Furthermore, there is evidence of decreased cross-sectional area10 and increased fatiguability11 and a suggestion of increased intramuscular fat in the paraspinal muscles of individuals with LBP.12 Therefore, theoretically, an intervention that aims to correct the changes occurring in the deep trunk muscles and that targets the restoration of control and coordination of these muscles should be effective in the management of persistent LBP.
Motor control exercise was developed based on the principle that individuals with LBP have a lack of control of the trunk muscles. The idea is to use a motor learning approach to retrain the optimal control and coordination of the spine. The intervention involves the training of preactivation of the deep trunk muscles, with progression toward more complex static, dynamic, and functional tasks integrating the activation of deep and global trunk muscles.13,14
Although a number of laboratory studies supporting the underlying mechanism of action of motor control exercises have been published in the last decades,5,9,15 the clinical effectiveness of motor control exercise for persistent LBP is still unclear.5,9,15 Three systematic reviews of motor control exercise have been published16–18; however, the authors of these reviews searched the literature only up until October 2005. Hauggaard and Persson,17 the authors of the latest published review, included 10 trials testing the efficacy of motor control for acute, subacute, and chronic LBP. The review used a simple descriptive approach to summarize the results of each individual trial. Rackwitz et al18 summarized the results of 7 randomized controlled trials of acute, subacute, and chronic LBP, and although they used a better approach to summarize the available evidence, no meta-analytical analysis with pooling of the data was used. Ferreira et al16 summarized the results of 13 randomized controlled trials of recurrent, acute, subacute, and chronic LBP and cervical pain. This review was the only one that included a meta-analytical approach; however, only a few trials were pooled, limiting the generalization of the results. A meta-analytical approach is superior to the other forms of analysis for systematic reviews because it provides a treatment effect size with 95% confidence interval (CI).
Consistent with the Cochrane Collaboration,19 we felt that an updated review incorporating new randomized controlled trials would make a useful contribution to the literature. In addition, a meta-analytical approach, which has not been widely used in the previous published systematic reviews, can potentially add useful information about the magnitude of the effect of motor control exercises. Because our main interest was to study persistent LBP and guidelines suggest that persistent and acute LBP should be considered separately,19–21 we included only trials studying patients with LBP that persisted beyond the acute phase. The term “persistent low back pain” is used to describe subacute, chronic, and recurrent pain. Thus, the objective of this study was to systematically review randomized controlled trials testing the effect of motor control exercise in patients with persistent, nonspecific LBP.
Data Sources and Searches
A computerized electronic search was performed to identify relevant articles. The search was conducted on MEDLINE (1950 to June 2008), CINAHL (1982 to June 2008), AMED (1985 to June 2008), PEDro (to June 2008), and EMBASE (1988 to June 2008). Key words relating to the domains of randomized controlled trials and back pain were used, as recommended by the Cochrane Back Review Group.19 Terms for motor control and specific stabilization exercises were extracted from the review by Ferreira et al.16 Subject subheadings and word truncations, according to each database, were used. There was no language restriction.
One reviewer (LGM) screened search results for potentially eligible studies, and 2 reviewers (LGM, CGM) independently reviewed articles for eligibility. A third independent reviewer (JL) resolved any disagreement about inclusion of trials. Authors were contacted if more information about the trial was needed to allow inclusion of the study. Researchers who published in the area were contacted to help identify gray literature and articles in press. Citation tracking was performed using ISI Web of Science, and a manual search of the reference lists of previous reviews and the eligible trials was performed.
The reviewers followed a research protocol, developed prior to the beginning of the review, that included a checklist for inclusion criteria. Articles were eligible for inclusion if they were randomized or quasi-randomized controlled trials comparing motor control exercise with a placebo treatment, no treatment, or another active treatment or when motor control exercise was added as a supplement to other interventions. When motor control exercise was used in addition to other treatments, motor control exercises had to represent at least 40% of the total treatment program. This criterion was judged by reading the description of the treatment with the reviewer making a global yes/no judgment.
Trials were considered to have evaluated motor control exercise if the exercise treatment was described as motor control or specific spinal stabilization or core stability exercise and where the protocol described exercise targeting specific trunk muscles in order to improve control and coordination of the spine and pelvis.
Randomized or quasi-randomized controlled trials were included if they explicitly reported that a criterion for entry was nonspecific LBP (with or without leg pain) of at least 6 weeks’ duration (nonacute LBP) or recurrent LBP. Studies evaluating individuals of all age groups of either sex were included. Trials were included if one of the following outcome measures had been reported: pain, disability, quality of life, return to work, or recurrence.
Data Extraction and Quality Assessment
The methodological quality of the trials was assessed using the PEDro scale,22 with scores extracted from the PEDro database. Assessment of quality of trials in the PEDro database was performed by 2 trained independent raters, and disagreements were resolved by a third rater.23 One study24 was extracted from a conference proceeding, and, therefore, the PEDro score was not available in the database. However, 2 PEDro raters evaluated the information available in the abstract and in an initial version of a manuscript, and a PEDro score was given. Methodological quality was not an inclusion criterion.
Three independent reviewers (LGM, CGM, JL) extracted data from each included study using a standardized extraction form. Mean scores, standard deviations, and sample sizes were extracted from the studies. When this information was not provided in the trial, the values were calculated or estimated using methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions.25 When there was insufficient information about outcomes to allow data analysis, the authors of the study were contacted, and all authors replied to our inquiries.24,26–28
Outcomes were extracted for pain and disability for short-term follow-up (less than 3 months after randomization), intermediate follow-up (at least 3 months but less than 12 months after randomization), and long-term follow-up (12 months or more after randomization). When there were multiple time points that fell within the same category, the one that was closer to the end of the treatment for the short-term follow-up, closer to 6 months for the intermediate follow-up, and closer to 12 months for the long-term follow-up was used. These references for time points were based on guidelines from the Cochrane Back Review Group. Scores for pain and disability were converted to a 0 to 100 scale.29
Data Synthesis and Analysis
The studies were grouped into 4 treatment contrasts: (1) motor control versus minimal intervention (no intervention, general practitioner care, education) or motor control as a supplement, (2) motor control versus spinal manipulative therapy, (3) motor control versus exercise, and (4) motor control versus surgery (lumbar fusion). Results were pooled when trials were considered sufficiently homogenous with respect to participant characteristics, interventions, and outcomes. I2 was calculated using RevMan 4.2* to analyze statistical heterogeneity. I2 describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error (chance). A value greater than 50% may be considered substantial heterogeneity.25 When trials were statistically homogeneous (I2<50%), pooled effects (weighted mean difference) were calculated using a fixed-effect model. When trials were statistically heterogeneous (I2>50%) pooled estimates of effect (weighted mean difference) were obtained using a random-effects model.25 When there was a single trial for the comparison, results were expressed as mean differences and 95% CI.
The initial electronic database search resulted in a total of 1,052 articles. Of these, 42 were selected as potentially eligible based on their title and abstract. Through a Web of Science search of these articles, 3 other potentially eligible articles were identified. A total of 45 potentially eligible articles were considered for inclusion, with only 14 eligible for inclusion in this review (Fig. 1). Reasons for exclusion are shown in Figure 1 for those articles2,3,15,30–57 that were excluded from this review. Only 1 of the 26 experts contacted sent information to us on a new trial for inclusion.
A number of randomized controlled trials that were included in previous systematic reviews of motor control exercises were not included in this review. Reasons for exclusion included: patients had acute but not persistent back pain,15,51,53 patients had neck pain and headache but not back pain,58 the trial did not use a motor control intervention according to our review definition,56 and the trial did not have the outcomes of interest.59,60 Four new trials13,24,26,61 that were not included in any of the previously published reviews were included in this review, accounting for the addition of 560 patients.
The methodological quality assessment using the PEDro scale revealed a mean score of 6 (range=2–8). Blinding of the therapist and blinding of the subject were not used in any of the trials, as would be expected for an exercise therapy study. An intention-to-treat analysis was used in 36% of the trials, and allocation concealment was present in 58% of the trials. One of the articles24 included in the review was from a conference proceeding, and, therefore, not much information on the conduct of the trial was available. With the limited information available, this trial received a score of 2 on the PEDro scale and was the only trial that was a quasi-randomized controlled trial.24
The 14 randomized controlled trials included in this review compared motor control exercise against another treatment or against no treatment (Tabs. 1 and 2). No placebo-controlled trials were identified. Trials were grouped into 4 treatment contrasts: (1) motor control exercise versus minimal intervention or motor control exercise as a supplement, (2) motor control exercise versus manual therapy, (3) motor control exercise versus other forms of exercise, and (4) motor control exercise versus surgery.
Seven trials (603 patients) were included in the first treatment contrast: 4 trials (343 patients) that compared motor control exercise with minimal intervention (no intervention, general practitioner care, or education)14,27,62,63 and 3 trials (260 patients) that used motor control exercise as a supplement to other treatment (general exercise or usual physical therapy.28,64,65 Four trials (523 patients) compared motor control exercise with manual therapy (high- or low-velocity trust).13,26,64,66 Five trials (508 patients) compared motor control exercise with another form of exercise therapy (pain management, general exercises, or the McKenzie approach).13,24,26,61,67 One trial (61 patients) compared motor control exercise with lumbar fusion surgery.68 The characteristics of the motor control exercise programs that were evaluated in each trial are provided in Table 2.
Motor Control Exercise Versus Minimal Intervention or Motor Control Exercise as a Supplement
Of the 7 studies included in this treatment contrast, 4 compared motor control exercise with a minimal intervention program (usual general practitioner care or no intervention)14,27,62,63 and 3 compared motor control exercise as a supplement to another intervention versus this other intervention alone.28,64,65 Methodological quality of the articles ranged from 4 to 8. Data for pain, disability, and quality of life were available for pooling at short-term, intermediate, and long-term follow-up. Data were pooled using a random-effects model for all comparisons except for quality of life at intermediate and long-term follow-ups, where a fixed-effects model was used because I2 was smaller than 50%.
The pooled results favored motor control exercise for pain and disability outcomes at each follow-up, with 4 of the 6 estimates of treatment effect being statistically significant. The random-effects model showed a statistically significant decrease in pain favoring motor control exercise at short-term follow-up (weighted mean difference [on a 0–100 scale]=−14.3 points, 95% CI=−20.4 to −8.1), intermediate follow-up (weighted mean difference=13.6 points, 95% CI=−22.4 to −4.1), and long-term follow-up (weighted mean difference=−14.4 points, 95% CI=−23.1 to −5.7) and in reducing disability at long-term follow-up (weighted mean difference=−10.8 points, 95% CI=−18.7 to −2.8) (Fig. 2). There was no evidence that motor control exercise was effective for improving quality of life.
Motor Control Exercise Versus Manual Therapy
Four trials13,26,64,66 compared motor control exercise with manual therapy, with pain and disability outcomes measured at short-term, intermediate, and long-term follow-ups and quality of life measured at intermediate and long-term follow-ups. The methodological quality of the articles ranged from 4 to 8. Because I2 was smaller than 50% for all time points, a fixed-effects model was used to pool the results. The pooled effects for pain and disability outcomes favored motor control exercise, but the effects were always small and reached statistical significance for only 2 of the 6 estimates. There was a significant difference between treatment groups favoring motor control exercise for pain and disability at intermediate follow-up (weighted mean difference=−5.7 points, 95% CI=−10.7 to −0.8 for pain and weighted mean difference=−4.0 points, 95% CI=−7.6 to −0.4 for disability) (Fig. 3). The pooled estimates of treatment effects on quality of life were small, favoring motor control exercise at short-term follow-up and favoring manual therapy at long-term follow-up.
Motor Control Exercise Versus Other Forms of Exercise
Five trials13,24,26,61,67 compared motor control exercise with another form of exercise therapy. The methodological quality of the trials ranged from 2 to 8. The trial with a methodological quality score of 2 had its PEDro score assessed from a conference proceeding and some information given by the authors.24 Results were pooled for pain and disability at short-term, intermediate, and long-term follow-ups. Because I2 was greater than 50% for pain at short-term follow-up and for disability at long-term follow-up, pooled effects for these time points were calculated using a random-effects model. All other pooled effects were calculated using a fixed-effects model. All estimates of treatment effect were small. Five of the 6 estimates favored motor control exercise; however, only one effect was statistically significant. The results showed that motor control exercise was better than other forms of exercises only for reducing disability at short-term follow-up (weighted mean difference=−5.1 points, 95% CI=−8.7 to 1.4) (Fig. 4). The results of a single trial26 showed no difference between treatment groups for quality of life at short-term follow-up.
Motor Control Exercise Versus Surgery
Only one study68 compared motor control exercise with surgery, with a methodological quality score of 8. Surgery consisted of lumbar fusion with transpedicular screws of the L4–L5 segments or the L5–S1 segments. Brox et al68 found no statistically significant differences for pain (mean difference [on a 0–100 scale]=−9 points, 95% CI=−22.1 to 3.5), disability (mean difference=−3.3 points, 95% CI=−12.8 to 6.2), and quality of life (mean difference= 0.4 points, 95% CI=−1.6 to 0.8) at the long-term follow-up (Fig. 5).
This systematic review provides evidence that motor control exercise, alone or as a supplement to another therapy, is effective in reducing pain and disability in patients with persistent, nonspecific LBP. We did not find convincing evidence that motor control exercise was superior to manual therapy, other forms of exercise, or surgery.
Figure 2 shows that there was some variation among studies in the effect sizes for motor control exercise. Features that could influence the treatment effect sizes are characteristics of the patients (eg, symptom duration), characteristics of treatment implementation (eg, program duration, experience of the therapist), and the methodological quality of the trial. Unfortunately, there are too few trials to systematically evaluate the effects of these features using techniques such as meta-regression.
An intriguing finding of this review was that motor control exercise was as effective in reducing pain and increasing quality of life as a less-complex form of exercise therapy that did not incorporate the retraining of specific muscles that often is time consuming to therapists and patients. When taking in consideration the results for disability, motor control exercise was more effective than other forms of exercise only at short-term follow-up, but the point estimate was small (5.1 out of 100), showing differences between interventions that may not be clinically important.
The results of a single trial68 showed that motor control exercise was not more effective than surgery. This finding is interesting because both interventions target the restoration of spinal stability, and although spinal stability was not directly measured, the findings suggest that the motor control approach is as effective in maintaining stability as an invasive intervention that creates stability by fusing the spine. However, this was the finding of a single trial, and more research is needed to confirm the results.
Although a motor control intervention has been shown to reduce pain, it is still unknown whether these changes are accompanied by improvements in measures of motor control. Tsao and Hodges69 have shown improvements in motor control (anticipatory contraction of the transversus abdominis muscle during arm movement) after a single treatment session where the isolation of the transversus abdominis muscle was trained. In a different trial, Hall and colleagues70 did not find that motor control (anticipatory contraction of the transversus abdominis muscle during arm movement and a walking task) changed after training the trunk muscles in a nonisolated manner. Therefore, the results of these 2 studies support the principles of a motor control intervention where the isolated training of the deep trunk muscles is emphasized. However, there has not been a published randomized controlled trial that used clinical and physiological measures to detect improvements in motor control that can be associated with improvements in pain and disability and the maintenance of these changes.
One question that is still to be answered is whether individuals with reduced motor control respond best to this intervention or whether there are other clinical features that can be used to define a subgroup of patients who will respond best to this type of intervention.
A standard protocol and definitions for motor control exercise are yet to be established, and this is reflected in the wide variation among trials in how the exercise was named and implemented (Tab. 2). Although in most cases O'Sullivan et al14 and Richardson et al71 were cited as references, it is apparent from inspection of the articles that the interventions in the trials were quite heterogeneous. There was variation in the duration of the exercise program, progression rule, use of home exercise programs, and type of feedback used with the motor control intervention. As an illustration, the program lasted 10 weeks in the trial by O'Sullivan et al, whereas the program lasted 18 to 20 weeks in the trial by Stuge et al.28 In the trial by Ferreira et al,13 ultrasound was used for feedback, and Stuge et al28 used Terapi Master exercise equipment†: 2 elements missing from the trial by O'Sullivan and colleagues.
Detailed comparison among trials is difficult because in many trials the authors did not thoroughly describe the motor control intervention that was evaluated. Accordingly, although we can conclude from this review that motor control exercise is an effective treatment for persistent LBP, the optimal way to implement this intervention is not yet clear.
When looking at the quality of the trials included in this review, a mean score of 6 can be considered a high score because these trials were exercise trials where it is impossible to blind the treatment provider and subjects, and, therefore, the maximum PEDro score that can be achieved is 8. However, because some trials were of lower methodological quality, they potentially present biased (and overly optimistic) estimates of treatment effects. To assess the impact of the lower-quality studies on the review conclusions, a sensitivity analysis with exclusion of trials with scores lower than 524,64 was performed. When the lower-quality studies were deleted, the effect size unexpectedly increased slightly for pain and disability outcomes (we did not conduct a sensitivity analysis for quality of life because the exclusion of these trials would leave only one trial in the treatment contrast). Therefore, we do not believe that our conclusion that motor control exercise is effective (compared with minimal intervention or when used as a supplement) is an artifact of the inclusion of low-quality trials.
This review not only includes 4 new trials that were not included in previous reviews, accounting for the addition of 560 patients, but also allowed the use of a meta-analytical approach with the inclusion of a greater number of articles into each treatment contrast. The pooled results of this systematic review showed smaller and more-precise estimates of treatment effects when compared with the pooled results of Ferreira et al.13 This difference among studies can be seen when looking, for example, at the motor control exercise versus minimal intervention contrast. For this contrast, Ferreira et al13 included 2 trials and found an effect of −21 on a 0 to 100 scale (95% CI=−32 to −9) for pain, whereas we found, based on 5 trials, an effect of −14.3 (95% CI=−20.4 to −8.1).
Although it has been only recently that reviews of motor control exercises have been published, this type of intervention is widely accepted and used in the clinical field around the world. Therefore, it is still crucial that further studies in the area be developed, such as a placebo-controlled trial and trials aiming to identify subgroups of patients who will benefit more from a motor control intervention. More fundamental studies in LBP to establish reliable and valid clinical assessment tools to identify deficits in motor control also are needed.
The results of this systematic review suggest that motor control exercise is more effective than minimal intervention and adds benefit to another form of intervention in reducing pain and disability for people with persistent LBP. The optimal implementation of motor control exercise at present is unclear. Future trials evaluating issues such as dosage parameters, feedback approaches, and effects in defined subgroups are a high priority.
Ms Macedo, Dr Maher, and Dr Latimer provided concept/idea/research design and data collection. Ms Macedo and Dr Maher provided writing and data analysis. Ms Macedo, Dr Maher, and Dr McAuley provided project management. Dr Latimer provided clerical support and consultation (including review of manuscript before submission).
Ms Macedo holds a PhD scholarship jointly funded by The University of Sydney and the Australian Government. Dr Maher's research fellowship is funded by Australia's National Health and Medical Research Council.
↵* Copenhagen, Denmark: The Nordic Cochrane Centre, The Cochrane Collaboration, 2003.
↵† Nordisk Terapi A/S, Kilsund 4290, Staubo, Norway.
- Received April 3, 2008.
- Accepted October 10, 2008.
- American Physical Therapy Association