This systematic review protocol was registered on PROSPERO (registration number: CRD42017080795) and adhered to the PRISMA recommendations and followed the recommendations from the Cochrane Collaboration.27,28

Electronic searches were performed using the following databases: Medline (via PubMed), Embase, Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science, Web of Knowledge, SCOPUS, CINAHL (via EBSCO), ProQuest, and PEDro. Additionally, manual searches on reference lists of previously published systematic reviews and relevant trials were conducted. The reference list from the McKenzie International Institute (www.mckenzieinstitute.org/clinicians/research-and-resources/references-list) was also screened. No time limit was set, and all studies that existed until March 2018 were considered for inclusion.

The search strategy for each database was developed by combining the relevant terms from each of the following three conceptual areas: MDT, non-specific CLBP, and clinical trial. The search strategy used in MEDLINE is available as supplemental online material.

Articles were eligible for inclusion if they were randomized controlled trials; included participants with CLBP (i.e., defined as chronic pain in the lumbar spine lasting more than 12 weeks) with or without radiation to the lower extremities; the experimental group received MDT-based exercises only (trial specifies treatment according to the MDT principles) and the comparison group received a form of physical therapy intervention other than MDT-based exercises. Lastly, the studies had to be written in English, French, Spanish, Portuguese, Italian, Bulgarian, or Russian. Trials were excluded if the recruited participants had any specific spinal pathology (such as lumbar radiculopathy), if the participants’ symptoms were acute or subacute, or if it was not clearly indicated in the study that the participants’ symptoms were chronic.

Two reviewers (E.S.S and S.G.G) conducted the study selection process, defined prior to the beginning of the review, which included a checklist with the eligibility criteria. The results of the searches were transferred into an electronic reference manager (Mendeley Desktop, version 1.17.9), which eliminated duplicate entries. Two reviewers independently screened the titles, abstracts, and full-texts of potentially eligible studies for inclusion into this review. Disagreements or discrepancies were resolved by a third reviewer (J.M.B).

Prior to the quantitative synthesis, the results were extracted and categorized based on the outcomes of interest. The primary outcomes were pain, as measured with the McGill Pain Questionnaire,29 disability, and lumbar range of motion (ROM). Secondary outcomes of interest included quality of life, recurrence, return to work/sick leave, medical visit, medication use, or global perceived effect.

The PEDro scale was used to evaluate the methodological quality of the included studies. Studies were categorized into high (score 3/10 or lower), moderate (score 4/10 or 5/10), or low (score 6/10 or higher) risk of bias. Agreement between the two reviewers (E.S.S and S.G.G) was evaluated using Cohen's kappa coefficients and any disagreement or discrepancies were resolved through consensus by a third reviewer (J.M.B).

First, a descriptive synthesis was performed. To summarize the characteristics of the included studies, we extracted the following data: study design, sample characteristics (i.e., number, mean age, and sex of participants), control and intervention groups, and scoring systems used to assess the predefined outcome measures. Subsequently, we performed a quantitative synthesis with meta-analysis.30 To do so, we summarized the evidence according to the type of experimental and control interventions used in the trials, and also to the time points used for outcome assessment (follow-up points). An intervention was classified as experimental when it included treatment based on MDT as the differentiating therapy. The control interventions were divided into two groups: active and passive treatment approaches. In the active treatment approach, participants played an active role as they performed some form of active exercise (e.g., Pilates), while in the passive treatment approach, participants received the treatment from the physical therapist (e.g., manual therapy). For trials testing more than two interventions, the control group included in the meta-analysis was the one that did not include MDT as part of the treatment or data from the control groups were combined. The Review Manager 5 software, from the Cochrane Collaboration, was used to conduct the meta-analysis. Time points of assessment were categorized as follows: short-term (less than three months after randomization), intermediate-term (at least three months but less than 12 months after randomization), and long-term (12 months or more after randomization). When there were multiple time points that fell within the same category we used the one that was closer to the end of the treatment, six months, and 12 months. Mean, standard deviation, and sample size from each group were extracted and used in the data analysis to estimate effect sizes. If these data were not available, estimations were based on t- and p-values. Because different scoring systems were used to assess the same outcome measures, effect sizes were estimated and presented using standardized mean difference (SMD). Cut-off points for interpreting the result are usually dependent on the type of intervention. However, the arbitrary thresholds established are generally accepted and allow interpretation of the magnitude of the pooled effects. Accordingly, the pooled effects were interpreted as small (SMD<0.2), moderate (0.2<SMD<0.8), or large (SMD>0.8).31 The meta-analysis was performed using the random effect model, assuming that the true effect may vary due to heterogeneity in the study design. Heterogeneity assessment was based on I2 statistics, and considered moderate to high when greater than 50%.32 The summary value for each study was represented in forest plots. Confidence intervals (CI) were set at 95%.

To determine the quality of evidence, the Grade of Recommendation Assessment, Development, and Evaluation (GRADE) approach was applied, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions32 and adapted in the updated Cochrane Back Review Group method guidelines.33 The GRADE scale shows the overall certainty of the evidence for the outcome being reviewed and following assessment of five domains: risk of bias (downgraded if more than 25% of the participants were from studies with low methodological quality), inconsistency of results (downgraded if significant heterogeneity was present by visual inspection or if the I2 value was greater than 50%), indirectness (generalizability of the findings, downgraded if more than 50% of the participants were outside the target group), imprecision (downgraded if fewer than 400 participants were included in the comparison), and other factors such as reporting bias or publication bias. The GRADE approach involves downgrading the evidence from high quality to moderate, to low, or to very low, based on the cumulative assessment of each of these criteria. The GRADE scale was applied to assess the evidence regarding the outcome measures “pain” and “disability”. Two authors independently reviewed and scored each study, with disagreements settled through discussion and consensus.

Fig. 1 shows the flow diagram of the studies considered in the review. The initial database search resulted in a total of 2633 articles. After removing duplicates and screening titles and abstracts, 14 studies34–47 were included. Of these, 11 studies provided data to be included in the meta-analyses. Three studies36,41,44 were not included in the meta-analyses because MDT was used in all arms of the trial36,44 or the control group used a combination of active and passive treatments (Fig. 2).41

Figure 1.

Flow diagram of search strategy and results.

(0.58MB).

Figure 2.

Forest plots for MDT compared with active treatments (A) and compared with passive treatments (B).

(0.66MB).

The studies characteristics are shown in Table 1. The following information was extracted from each study: study design, participants, groups, outcome measures, and follow-up points.

All studies included in this review compared MDT with different intervention strategies. Four studies40,42,43,47 compared MDT with passive treatments, such as massage and reflexology43 (data from both groups were combined in the meta-analysis), mobilization,40 manipulation,47 or electrophysical agents.42 Another seven studies34,35,37–39,45,46 compared MDT with active interventions such as back school education,46 stabilization exercises,39,45 motor control exercises,35 repeated lumbar flexion movements,37,38 or Pilates.34 In the study by Long et al.,38 the effects of MDT were compared with two different active interventions, so data from both controls were pooled for analysis. Finally, the remaining three studies,36,41,44 not included in the meta-analysis, combined MDT with other interventions such as endurance exercises, passive traditional interventions, and dynamic back exercises. In all included studies,34–47 the MDT directional preference was lumbar extension and participants were instructed on how to perform the exercises independently.

All studies34–47 assessed pain intensity over time. Eleven studies35,36,38–43,45–47 evaluated disability, six36,37,40–42,46 evaluated lumbar ROM, and four34,44,46,47 evaluated general health and quality of life. Treatment lasted from 10 days to 12 weeks, and all participants were adults, with an age ranging from 40 to 60 years. Regarding the time point of assessment, 11 studies34–40,42–45 assessed short-term, two studies41–46 intermediate, and one study47 long-term effect.

The quality assessment and risk of bias is shown in Table 2. The kappa index for reliability between the two reviewers was high (k=0.76; 95% CI: 0.70, 0.82), indicating good reliability. The main disagreements between the two reviewers were related to the selection of the participants, the blinding of participants and assessors, the concealment of allocation, and the statistical analyses.

From the 14 studies,34–47 six of them35,37,42,43,46,47 had a low risk of bias, and the remaining eight 34,36,38–41,44,45 had a moderate risk of bias. Incomplete outcome data or a non-specific description of the type of pain, as well as lack of blinding of participants due to the nature of clinical trials, were commonly reported.

A total of seven trials34,35,37–39,45,46 compared MDT with active treatments with follow-up assessment up to three months. The MDT approach was no more effective than active treatments for reducing pain (SMD=0.01; 95% CI: −0.44, 0.46) and disability (SMD=0.08; 95% CI: −0.53, 0.68). For pain intensity, based on seven trials,34,35,37–39,45,46 and for disability, based on four trials,38,39,45,46 the quality of evidence was downgraded due to inconsistency and risk of bias (Table 3). Therefore, the overall quality of evidence for both outcomes was considered to be low.

A total of four trials40,42,43,47 compared MDT with passive treatments with follow-up assessment up to three months. The MDT approach was no more effective than passive treatments for reducing pain (SMD=−0.39; 95% CI: −0.90, 0.11) or disability (SMD=−0.13; 95% CI: −0.29, 0.03). For pain intensity, the quality of evidence (downgraded due to risk of bias and inconsistency) was low. However, for disability, the overall quality of evidence (downgraded due to risk of bias) was considered to be moderate (Table 3).

Six studies36,37,40–42,46 assessed improvements in lumbar ROM. However, the characteristics of the included studies precluded meta-analysis. For two studies36,41 the interventions were not comparable, while for the remaining four studies,37,40,42,46 the authors used different approaches for assessing lumbar mobility. Two studies37,46 compared MDT with an active intervention by assessing sagittal, coronal, and transverse lumbar mobility versus overall trunk flexion; the other two40,42 compared MDT with a passive intervention, by assessing lumbar flexion versus fingertip to floor distance test. Therefore, a meta-analysis was not performed for this outcome.

Among the studies included in the meta-analysis, two46,47 assessed the outcomes three months or longer after randomization. But, one46 compared MDT versus an active intervention and the other one47 compared MDT versus a passive intervention; hence, it was not possible to perform meta-analysis for the immediate and long-term follow-up time points of assessment.