|Home | About | Journals | Submit | Contact Us | Français|
The effect size for exercise therapy in the treatment of chronic non-specific low back pain (cLBP) is only modest. This review aims to analyse the specificity of the effect by examining the relationship between the changes in clinical outcome (pain, disability) and the changes in the targeted aspects of physical function (muscle strength, mobility, muscular endurance) after exercise therapy.
We searched for exercise therapy trials for cLBP published up to 15 April 2010 in Medline, Embase, Cochrane Library, Cinahl, and PEDro. Two independent reviewers selected studies according to the inclusion criteria. Data extraction: one author extracted the data of the articles.
Data synthesis: 16 studies with a total of 1,476 participants met the inclusion criteria. There was little evidence supporting a relationship between the changes in pain or physical function and the changes in performance for the following measures: mobility (no correlation in 9 studies, weak correlation in 1 study), trunk extension strength (7 and 2, respectively), trunk flexion strength (4 and 1, respectively) and back muscle endurance (7 and 0, respectively). Changes in disability showed no correlation with changes in mobility in three studies and a weak correlation in two; for strength, the numbers were four (no correlation) and two (weak correlation), respectively.
The findings do not support the notion that the treatment effects of exercise therapy in cLBP are directly attributable to changes in the musculoskeletal system. Future research aimed at increasing the effectiveness of exercise therapy in cLBP should explore the coincidental factors influencing symptom improvement.
Studies examining the effects of exercise therapy in the treatment of chronic non-specific low back pain (cLBP) show in general only moderate effectiveness [1–3]. This is often explained by the contradiction between the heterogeneity of cLBP patients and the uniformity of the exercise therapy approach . In line with this approach lies the recommendation to sub-group patients and to develop relevant exercise programs for each group. However, there is little evidence that individually tailored or specific exercise programs show better success, which tends to question this approach . Most exercise therapy trials report the changes in key outcome variables such as patient-rated pain, disability and global improvement, but they rarely examine these in relation to improvements in the targeted aspect of performance, e.g. strength or mobility. Hence, it cannot be assumed that the observed positive clinical outcome is a direct result of improvements in the specific functional deficit targeted by the treatment.
Recently, alternative theories have been proposed that aim to explain the lack of specificity of exercise therapy in cLBP. One suggests that the treatment effects of many cLBP therapies may be attributable to changes within the brain of cLBP patients rather than specific changes in the musculoskeletal system [4, 5]. Evidence supporting the involvement of cortical reorganisation in cLBP  comes from the finding of central nervous system changes [7–11] proportional to the severity and duration of the cLBP [4, 11, 12] and alterations (grey matter density loss)  in the brain of cLBP patients [4, 12, 14]. It is conceivable that other changes elicited by exercise therapy, e.g. improvements in self-efficacy, coping strategies and fear-avoidance [14–20], modification of motor control patterns as a consequence of a re-weighting of sensory input , changes in cortical organisation [22–25] or simply a positive therapist–patient interaction/relationship  may be responsible for the improvements in self-reported pain and disability.
Current treatments for cLBP may be ineffective because they are based on the unsubstantiated assumption that the problem is located in the lower back itself and is the result of a specific functional deficit that might be remedied by a specific type of exercise. The treatments applied to target these dysfunctions can be expensive, requiring individual treatment, specially trained therapists, and/or specialised equipment; if there is no evidence that specific exercises are actually required, then they represent an unnecessary drain on our limited health-care resources. If the observed alterations in the periphery, such as increased movement asymmetry and variability [27–29], reduced movement speed , increased muscle co-contraction [31, 32], and decreased back muscle endurance [33–35], strength [33, 35] and mobility  are compensatory rather than causative, then future research could be directed towards alternative (and perhaps less costly) intervention models with new approaches, e.g. strategies for re-training the cortical function [14, 37–42], hopefully generating more effective results in the treatment of cLBP.
The European guidelines for the management of cLBP  briefly reported on the relationship between changes in physical performance and changes in clinical outcome, and found that the associations were at best tenuous. The aim of this study was to perform a systematic review of the studies that have examined this phenomenon. Specifically, we evaluate the reported correlations between changes in clinical outcome(s) (pain, disability) and changes in physical function (range of motion, strength, and muscular endurance) as a result of physical therapy and exercise interventions in patients with cLBP.
Individualised search strategies for Medline and Pre-Medline with Ovid, Cochrane library with Wiley, and Embase, Cinahl and PEDro databases (Appendix 1) were developed in collaboration with a librarian from the local university library. No limits were applied for the publication date of the articles. We used medical sub-headings as search terms, including low back pain, chronic disease, chronic low back pain, backache, treatment outcome, perception, pain measurement, pain, exercise therapy, physical therapy modalities, exercise movement techniques, and the free text words exercise therapy, physical therapy modalities, physiotherap*, exercise, global improvement, global impression, physical function, functional * restoration. We also reviewed the bibliographies of retrieved articles and relevant conference proceedings. The final search in all databases was performed on 15 April 2010.
A study was considered eligible for inclusion in the review if it was a randomised controlled trial (RCT) or a non-randomised controlled trial (non-RCT) in English or German, examining the results of a physical activity intervention on patient reported outcome in cLBP. The studies had to investigate the effects of exercise and/or a physical therapy intervention of any type. Studies that evaluated the effectiveness of drugs, transcutaneous electrical stimulation and other non-exercise therapy modalities were excluded. The outcome measures of interest were: (1) outcomes related to physical function/performance (strength, mobility, muscular endurance) and (2) clinical outcomes (pain, disability). All participants with cLBP regardless of age were included. The definition of cLBP was the same as that given in the European guidelines for the management of cLBP .
On the basis of the abstracts of the articles, studies were firstly eliminated if they did not focus on cLBP, exercise therapy or clinical and physical outcomes. Full text copies of the studies that were potentially suitable were then obtained and were independently assessed for inclusion by two of the authors (F.S. and E.D.B.) on the basis of the eligibility criteria. Studies were excluded if the individuals examined were not suffering from cLBP or if the studied patient groups were inhomogeneous due to the inclusion of a mixture of acute, subacute and chronic LBP patients. Further reasons for exclusion were (as per the European Guidelines ): a specific and uniform pathology (e.g. spondylolysis/spondylolisthesis, post-operative pain) and grouping of patients with mixed complaints (cLBP together with another complaint). In the event of disagreement between the raters, a third reviewer (B.W.) was available for consultation.
One author (F.S.) independently extracted the following information from each study selected for inclusion: (1) characteristics of the study participants; (2) type of intervention (including type, duration, frequency of training); (3) type of outcome measure (pain scores, disability scores, strength measurements, mobility scores, and muscular endurance measures); (4) statements concerning correlations and/or correlation coefficients for the relationship between clinical variables and performance outcomes. If a study reported both immediate post-intervention and follow-up data, we used the post-intervention data.
The risk of bias and methodological quality of the RCT studies was assessed by the PEDro quality assessment tool, which is a ten-point checklist  that assesses randomisation, blinding of patients and therapists, follow-up, group baseline comparability and statistical analysis. All the identified RCTs had been already included in the PEDro database and we adopted their published rating (after double checking them ourselves and liaising with PEDro where discrepancies arose). The methodological quality of the non-RCT studies was assessed by means of the Downs and Black checklist . Percentage agreement and Cohen’s kappa for the reviewers’ ratings were calculated with GRAPHPAD software (Version 2002–2005; GRAPHPAD Software Inc, San Diego, CA), and were interpreted in accordance with Landis and Koch’s  benchmarks for assessing the agreement between raters: poor (< 0), slight (0.0–0.20), fair (0.21–0.40), moderate (0.41–0.60), substantial (0.61–0.80), and almost perfect (0.81–1.0). The PRISMA-statement  was followed for reporting items of this systematic review and meta-analyses.
The study results, i.e. correlation coefficients, were pooled using a random effects model. Appropriateness of pooling was checked through evaluating heterogeneity. Heterogeneity of the study findings was assessed with the I-squared statistic, where a value greater than 50% is considered to indicate substantial heterogeneity . All other information was summarised and analysed qualitatively.
The database searches returned 1,217 articles of which 277 were duplicates. After reading the abstracts and applying the inclusion and exclusion criteria, 58 studies were considered potentially suitable and ordered for full text reading. After consulting the full texts, 16 studies, and 2 studies [16, 18] identified from a secondary analysis, with a total of 1,476 participants were identified as suitable for inclusion in the review (Fig. 1; Appendix 1).
The results of the methodological quality assessment are presented in Tables 1 (non-RCTs) and and22 (RCTs). For all the RCTs, PEDro scores were available in the PEDro database and were therefore not analysed in Graphpad. For the non-RCTs, the reviewers agreed on 114 of 130 methodological ratings (87.7%) (item 27 in the Downs and Black checklist was not rated, because this item was not clearly explained). The remaining disagreements were resolved after discussion between the reviewers. The inter-reviewer κ statistic was 0.752 (95% CI 0.638–0.866). The median criteria score on the PEDro list (range 1–10) was 5.0 (Table 2).
All 16 studies explicitly stated the eligibility criteria employed, 12 studies reported using an appropriate method to generate the random allocation sequence, reported group similarity at baseline for the most important prognostic indicators, were successful in obtaining at least 85% of the data for the primary outcome(s), performed an intention-to-treat analysis, provided between-group comparisons and provided point estimates and measures of variability for the primary outcome(s). 2 of 16 studies reported using an appropriate method for concealment of treatment allocation. The outcome assessors were blinded in 12 of 16 studies.
Characteristics of the patients, duration of exercises, interventions and main findings for each study are summarised in Table 3.
Figure 2 provides a summary of the study results describing the correlations between changes in physical function and changes in pain.
Ten studies were found (of which 3 gave the actual correlation coefficient data [49–51] and 7 did not [52–58]) that focused on the correlation between changes in pain and changes in sagittal mobility (flexion or flexion and extension). Nine studies reported that there was no correlation while one reported a low, but significant correlation . We performed a meta-analysis using the data from the three studies reporting their correlation coefficients. The total correlation was very low (−0.009). However, the I-squared factor was high (68.4%) indicating high heterogeneity (Fig. 3).
Two studies examined the relationship between changes in pain and changes in rotational and lateral mobility. One study  found weak significant negative correlations between changes in pain and in rotational and lateral mobility (r = −0.35 and r = −0.35, respectively). The second study  found a weak significant correlation between pain and rotational total mobility (r = −0.22) and pain and total mobility in lateral flexion (r = −0.12).
Nine studies performed regression analyses to determine the relationship between changes in pain and changes in extension strength, of which four gave the actual correlation coefficient data [16, 18, 51, 59] and five did not [52–54, 56, 60]. Seven studies reported that there was no correlation between these attributes but five of these had provided no actual correlation coefficients. Among the four studies that reported the actual coefficients, two found no significant correlation [16, 51] (r = −0.4 and r = 0.2) and two [18, 59] reported a significant correlation (r = 0.56 and r = 0.55). The meta-analysis resulted in a total correlation of 0.262. Again, the I-squared factor was high (90.70%), indicating high heterogeneity (Fig. 4).
Five studies addressed the relationship between changes in pain and changes in trunk flexion strength. Four [52, 55, 57, 59] reported that there was no correlation (but did not give the actual correlation coefficients). The only study  that reported correlation coefficients showed a weak non-significant correlation (r = 0.01).
Seven studies [52, 54, 56, 57, 61–63] examined the relationship between changes in pain and changes in muscular endurance, but none of them reported any specific correlation coefficients to substantiate their statements that there was no significant correlation between the variables.
Figure 5 provides a summary of the study results describing the correlations between changes in physical function and changes in disability.
Five studies (three providing the actual correlation coefficients and two without such data) focused on the relationship between changes in disability and changes in spinal mobility. Three of the five studies [36, 53, 57] reported that they found no correlation (one reporting r = −0.02, p = 0.86 ). Of the other two, one  found a significant correlation, but only in women (beta coefficient in multiple regression = 0.29, p < 0.05), while the second study  found a weak, but significant correlation (r = 0.18, p = 0.04). Due to the limited number of studies and the heterogeneity of the reported data, we refrained from performing a meta-analysis.
Six studies (two providing the actual correlation coefficients and four without such data) investigated the relationship between changes in disability and changes in strength. Four reported that there was no correlation [53, 58, 60, 65] while two studies reported significant correlations (r = 0.57 , r = 0.40 ).
No studies were found reporting the correlation between changes in disability and changes in muscular endurance.
The data analysis was mainly done qualitatively. Examination of just two relationships (pain and spinal mobility in the sagittal plane; pain and trunk extensor strength) resulted in sufficient studies reporting correlation coefficients to allow a meta-analysis to be carried out. These analyses resulted in high I-square values indicating large heterogeneity of the pooled data due to differences in intervention, control groups, duration of follow-up, outcome measures and study population.
The aim of this systematic review was to study the relationship between changes in clinical outcome (pain, disability) and changes in physical function (range of motion, strength, muscular endurance) as a result of physical therapy and exercise interventions in cLBP. The majority of the 16 studies reviewed indicated that no such relationship exists. Changes in pain showed predominantly no significant correlation with changes in mobility (9 studies reported no significant correlation and just 1 reported a correlation), or trunk extensor strength (7 and 2 studies, respectively) or trunk flexor strength (4 studies and 1 study, respectively), and no correlation with changes in back muscle endurance (7 and 0 studies, respectively). The meta-analysis for the associations between changes in pain and mobility also supported this conclusion, although the I-squared coefficient of greater than 60% reduced the explanatory power of the pooled data. Overall, we conclude that there is not convincing evidence that changes in pain are strongly associated with changes in physical function/performance.
Similarly, for disability, a predominance of studies showed no significant correlation with changes in mobility (3 reported no significant correlation and 2 a significant correlation) and changes in trunk extensor strength (4 and 2 studies, respectively), although these findings were less consistent than for pain.
In general, these findings concur with those of other systematic reviews and individual studies [43, 66–69]. As highlighted before , if specific types of exercise therapy are to be advocated—especially those that aim to target specific functional deficits—it is important to be able to establish that improvements in the clinical complaint after therapy are in some way associated with the specific changes in function elicited. It is often not clear whether changes in performance are responsible for improvements in pain/disability or whether these two simply occur coincidentally and are actually mediated by a common third factor. If a correlation between the changes in two variables (e.g., muscle strength and disability) is established, this does not necessarily prove the existence of a causal relationship; the converse, however, i.e. a reduction in disability/pain in the absence of any significant change in the performance dimension under investigation or vice versa (i.e., no correlation), would certainly imply that the two were unrelated. The latter appears to be emerging as the overarching conclusion of the studies conducted on this theme to date, and might also explain why no particular type of exercise therapy is presently considered to be superior to any other [1, 70], i.e., because the exercise therapy is not actually eliciting its effects by improving specific aspects of (dys)function. The assumption that the reversal of deficits in physical function—believed to either predispose to LBP or to arise due to physical deconditioning subsequent to cLBP—results in a decrease in pain/disability was hence not substantiated by this review. Instead, our findings appear more congruent with reports showing that patients with cLBP do not necessarily show marked deficits in function [71, 72]. Recently, the popular intervention of core-strengthening exercises (focusing on strengthening the rectus abdominus, internal and external obliques, and erector spinae muscles) was questioned in a study that sought to compare this type of exercise with a general non-specific strengthening programme . The outcomes were similar in the two treatment groups, and the authors concluded that focusing specifically on core exercises might be a potential mistake in the rehabilitation of cLBP . Furthermore, it was shown in other studies that even stretching exercises appeared to improve strength , which is difficult to explain on any physiological basis. A noteworthy feature of the trials included in this review was the large variability in exercise interventions. The diversity in the activities prescribed (e.g. strength and endurance training, interventions, with or without counselling) reflects the absence of consensus on the optimal activity programme for cLBP. Guidelines report that exercises may include aerobic activity, movement instruction, muscle strengthening, posture control and stretching, but at the same time provide no information about the required intensity, frequency, loading, progression, etc. for the chosen training programme. However, it is conceivable that these same factors—that undoubtedly influence the prescription of exercise in relation to medical conditions such as hypertension or obesity—are of less relevance when prescribing exercise for cLBP. Indeed, if the main aim of exercise therapy in cLBP is to get patients moving again and be able to confront their fears and anxieties about physical activity and movement, then the method used to do this may be immaterial. And if this were indeed the case, it may have the fortuitous side-effect that it would open up the array of potential options for the type of exercise to be carried out, allowing consideration of the all-important issues of cost, access to facilities and patient-preferences.
The biological mechanisms explaining the effects of exercise therapy are not yet clear , but the findings of the present review suggest that the improvements in clinical outcome do not result from local (muscle, joint, etc.) changes. Other possible explanations are that they derive from more central effects [14–20], perhaps as a correction of a distorted “body schema” [4, 14] or altered cortical representation of the back [22–25], from modification of motor control patterns as a consequence of a reweighting of sensory input , or simply from a positive therapist–patient interaction/relationship . Several studies have reported a correlation between psychological status and low back pain or pain tolerance [15, 58, 75–78]. The efficacy of treatments that solely focus on psychological targets has, however, been shown to be small . These psychological phenomena, similar to the peripheral physical deficits, may also be responses to an altered body schema in the sense of sensory–motor incongruence that causes fear . Exercise therapy seems to positively influence psychological variables such as fear-avoidance beliefs, catastrophising and self-efficacy regarding pain-control , in addition to providing physical benefits. Possibly by experiencing no harm in completing exercises, patients gain trust in the function of their back and thereby adjust their irrational cognitions and appraisals , whilst simultaneously improving their physical function.
Based on the findings of our review and on similar information from other systematic reviews and studies [43, 66–69], we suggest that changes in physical function are largely unable to explain changes in the clinical condition in cLBP patients, and that the important “side effects” of exercise therapy (including, amongst other things, changes in psychological variables such as fear-avoidance beliefs, catastrophising and self-efficacy regarding pain-control) should be more specifically emphasised and investigated in future rehabilitation programs.
We used a structured study protocol to guide our search strategy, study selection, extraction of data and statistical analysis. However, a number of possible limitations of this review should be noted. First, the search strategy was limited to published studies identified through the selected search engines. Second, as noted, a publication bias may have been present, as well as a language bias, given that we restricted our search to English and German language publications. Third, as there were only 12 randomised trials, we also included several observational studies, the results of which may be affected by confounding bias due to the absence of random assignment. However, as the focus of our analysis was not the relative efficacy of different treatments, this was expected to be of little consequence. The literature search for this review revealed 58 studies that potentially could have been included, but more than half of them had not conducted any correlation analyses. We tried to obtain the original data by contacting the authors of the studies that had failed to report actual correlation data, either by email, telephone or both. Unfortunately, the few who responded either no longer had access to the data or were not interested in providing their data. This undoubtedly resulted in a loss of potential information. A further problem was that most studies that did conduct correlation analyses, did not report any corresponding data (correlation coefficients) substantiating their reported non-significant correlations that would otherwise have allowed for quantitative data analysis with meta-analyses. Finally, the interventions were heterogeneous in their design and of variable quality.
Intervention strategies that focus solely on the symptom area in the lower back should be extended to apply a more global treatment approach. Both psychological and psychosocial interventions in addition to conventional exercise therapy may have a more positive effect on treatment outcome [80–90]. The targeted effect of such an approach would be the development of a sense of control over pain and the elimination of pain-avoidance mechanisms, whilst simultaneously improving overall physical fitness/function. Emphasis would shift from the “reversal of specific performance deficits” to the “adoption of enjoyable health-promoting physical activity” and this would potentially be associated with a wider choice and reduced cost. The availability of and access to such treatments might also be broadened by offering, e.g. group treatment sessions in community-based (rather than medical) settings. The exercise programs might include the training of proprioception, sensorimotor control and postural balance , in addition to the more conventional aspects of performance (strength, mobility, etc.). Lastly, the beneficial psychological effects of exercise should be investigated in greater detail. A better knowledge of the psychological changes induced by physical activity and training, and any accompanying “placebo” effects or educational effects due to the therapist–patient interaction, has the potential for enhancing the efficacy of exercise as a treatment for cLBP.
We conclude that the available literature does not appear to support a convincing association between changes in clinical outcome and changes in physical function after exercise therapy for cLBP. We hypothesise that the beneficial effects of exercise are more “central” than local, perhaps involving psychological, cognitive or neurophysiological (cortical organisation) adaptations. Thus, instead of trying to subdivide cLBP patients into further subgroups on the basis of specific functional deficits, future therapy approaches might better focus on influencing these central factors in cLBP patients.
We acknowledge the financial support of the Institute of Human Movement Sciences and Sport, Swiss Federal Institute of Technology, Zürich, Switzerland. We would like to thank the librarians of the Swiss Federal Institute of Technology Library, Zürich, for their assistance with the literature search.