Can We Explain Heterogeneity Among Randomized Clinical Trials of Exercise for Chronic Back Pain? A Meta-Regression Analysis of Randomized Controlled Trials

Exercise programs may vary in terms of duration, frequency, and dosage; whether they are supervised; and whether they include a home-based program. Uncritical pooling of heterogeneous exercise trials may result in misleading conclusions regarding the effects of exercise on chronic low back pain (CLBP). The purpose of this study was to establish the effect of exercise on pain and disability in patients with CLBP, with a major aim of explaining between-trial heterogeneity. Six databases were searched up to August 2008 using a computerized search strategy. Eligible studies needed to be randomized clinical trials evaluating the effects of exercise for nonspecific CLBP. Outcomes of interest were pain and disability measured on a continuous scale. Baseline demographic data, exercise features, and outcome data were extracted from all included trials. Univariate meta-regressions were conducted to assess the associations between exercise effect sizes and 8 study-level variables: baseline severity of symptoms, number of exercise hours and sessions, supervision, individual tailoring, cognitive-behavioral component, intention-to-treat analysis, and concealment of allocation. Only study-level characteristics were included in the meta-regression analyses. Therefore, the implications of the findings should not be used to differentiate the likelihood of the effect of exercise based on patient characteristics. The results show that, in general, when all types of exercise are analyzed, small but significant reductions in pain and disability are observed compared with minimal care or no treatment. Despite many possible sources of heterogeneity in exercise trials, only dosage was found to be significantly associated with effect sizes.