BACKGROUND: The prevalence, efficacy, and risk for addiction for persons receiving opioids for chronic back pain are unclear. PURPOSE: To determine the prevalence of opioid treatment, whether opioid medications are effective, and the prevalence of substance use disorders among patients receiving opioid medications for chronic back pain. DATA SOURCES: English-language studies from MEDLINE (1966-March 2005), EMBASE (1966-March 2005), Cochrane Central Register of Controlled Clinical Trials (to 4th quarter 2004), PsychInfo (1966-March 2005), and retrieved references. STUDY SELECTION: Articles that studied an adult, nonobstetric sample; used oral, topical, or transdermal opioids; and focused on treatment for chronic back pain. DATA EXTRACTION: Two investigators independently extracted data and determined study quality. DATA SYNTHESIS: Opioid prescribing varied by treatment setting (range, 3% to 66%). Meta-analysis of the 4 studies assessing the efficacy of opioids compared with placebo or a nonopioid control did not show reduced pain with opioids (g, -0.199 composite standardized mean difference [95% CI, -0.49 to 0.11]; P = 0.136). Meta-analysis of the 5 studies directly comparing the efficacy of different opioids demonstrated a nonsignificant reduction in pain from baseline (g, -0.93 composite standardized mean difference [CI, -1.89 to -0.03]; P = 0.055). The prevalence of lifetime substance use disorders ranged from 36% to 56%, and the estimates of the prevalence of current substance use disorders were as high as 43%. Aberrant medication-taking behaviors ranged from 5% to 24%. LIMITATIONS: Retrieval and publication biases and poor study quality. No trial evaluating the efficacy of opioids was longer than 16 weeks. CONCLUSIONS: Opioids are commonly prescribed for chronic back pain and may be efficacious for short-term pain relief. Long-term efficacy (> or =16 weeks) is unclear. Substance use disorders are common in patients taking opioids for back pain, and aberrant medication-taking behaviors occur in up to 24% of cases.