Pain associated with cancer is of widespread concern. We conducted a systematic review to evaluate the best available evidence on the efficacy of treatments of cancer-related pain. The sources used were MEDLINE, CancerLit, and the Cochrane Library from 1966 through April 2001, as well as bibliographies of meta-analyses and review articles. We selected randomized controlled trials (RCTs) reporting on cancer pain treatment. We recorded the study characteristics, patient and disease characteristics, treatment comparisons, outcome measures, and results. The methodological quality, applicability, and magnitude of treatment effect for each study were graded. We screened 24 822 titles and selected 213 RCTs to address specific questions. RCTs of cancer pain control often enroll few subjects, have low methodological quality, offer little detail about pain characteristics and mechanisms, and involve heterogeneous interventions and outcomes. Nonsteroidal anti-inflammatory drugs (NSAIDs), opioids, selected adjuvant medications, bisphosphonates, radionuclides, external radiation, palliative chemotherapy, and neurolytic celiac plexus block are each efficacious in relieving cancer pain. However, the retrieved RCTs indicate no difference in the analgesic efficacies of NSAIDs versus other NSAIDs, NSAIDs plus opioids versus NSAIDs alone, or NSAIDs versus opioids. Studies of adjuvant medications and behavioral therapies are too few and varied to synthesize. RCTs of the analgesic effects of corticosteroids were not retrieved in our review, although we did conduct supplemental evidence reviews concerning pain control in oral mucositis, acute herpes zoster, or postherpetic neuralgia. RCTs confirm the efficacy of diverse interventions in relieving cancer pain. The optimal initial and subsequent sequence of choices among analgesic drug types cannot be inferred from the retrieved RCTs. Patient preferences, the relative efficacy of different routes of drug administration, the side effects of analgesics, and the relation of pain control to quality of life have not been studied comprehensively. The quantity and quality of scientific evidence on cancer pain relief compare unfavorably with evidence related to treatment of other high-impact conditions, including cancer itself. One contributor to this gap is the heterogeneity of outcomes instruments employed: of 218 retrieved trials, there were 125 distinct pain outcomes assessed. In the current era of patient-centered care, improving the quality and combinability of trials on cancer pain relief should be a high research priority.