Objective: Estimates of treatment effects reported in placebo-controlled randomized trials are less subject to bias than those estimates provided by other study designs. The objective of this meta-analysis was to estimate the analgesic effects of treatments for non-specific low back pain reported in placebo-controlled randomized trials.
Methods: Medline, Embase, Cinahl, PsychInfo and Cochrane Central Register of Controlled Trials databases were searched for eligible trials from earliest records to November 2006. Continuous pain outcomes were converted to a common 0-100 scale and pooled using a random effects model.
Results: A total of 76 trials reporting on 34 treatments were included. Fifty percent of the investigated treatments had statistically significant effects, but for most the effects were small or moderate: 47% had point estimates of effects of <10 points on the 100-point scale, 38% had point estimates from 10 to 20 points and 15% had point estimates of >20 points. Treatments reported to have large effects (>20 points) had been investigated only in a single trial.
Conclusions: This meta-analysis revealed that the analgesic effects of many treatments for non-specific low back pain are small and that they do not differ in populations with acute or chronic symptoms.