Background: Exercise therapy encompasses a heterogeneous group of interventions. There continues to be uncertainty about the most effective exercise approach in chronic low back pain. Purpose: To identify particular exercise intervention characteristics that decrease pain and improve function in adults with nonspecific chronic low back pain. Data Sources: MEDLINE, EMBASE, Psychinfo, CINAHL, and Cochrane Library databases to October 2004 and citation searches and bibliographic reviews of previous systematic reviews. Study Selection: Randomized, controlled trials evaluating exercise therapy in populations with chronic (> 12 weeks duration) low back pain. Data Extraction: Two reviewers independently extracted data on exercise intervention characteristics: program design (individually designed or standard program), delivery type (independent home exercises, group, or individual supervision), dose or intensity (hours of intervention time), and inclusion of additional conservative interventions. Data Synthesis: 43 trials of 72 exercise treatment and 31 comparison groups were included. Bayesian multivariable randomeffects meta-regression found improved pain scores for individually designed programs (5.4 points [95% credible interval (Cri), 1.3 to 9.5 points]), supervised home exercise (6.1 points [Crl, -0.2 to 12.4 points]), group (4.8 points [Crl, 0.2 to 9.4 points]), and individually supervised programs (5.9 points [Crl, 2.1 to 9.8 points]) compared with home exercises only. High-dose exercise programs fared better than low-dose exercise programs (1.8 points [Cri, -2.1 to 5.5 points]). Interventions that included additional conservative care were better (5.1 points [Crl, 1.8 to 8.4 points]). A model including these most effective intervention characteristics would be expected to demonstrate important improvement in pain (18.1 points [Crl, 11.1 to 25.0 points] compared with no treatment and 13.0 points [CrI, 6.0 to 19.9 points] compared with other conservative treatment) and small improvement in function (5.5 points [CrI, 0.5 to 10.5 points] compared with no treatment and 2.7 points [Cri, -1.7 to 7.1 points] compared with other conservative treatment). Stretching and strengthening demonstrated the largest improvement over comparisons. Limitations: Limitations of the literature, including low-quality studies with heterogeneous outcome measures and inconsistent and poor reporting; publication bias. Conclusions: Exercise therapy that consists of individually designed programs, including stretching or strengthening, and is delivered with supervision may improve pain and function in chronic nonspecific low back pain. Strategies should be used to encourage adherence. Future studies should test this multivariable model and further assess specific patient-level characteristics and exercise types.
