Pain is very common in patients with cancer. Opioid analgesics, including codeine, play a significant role in major guidelines on the management of cancer pain, particularly for mild to moderate pain. Codeine is widely available and inexpensive, which may make it a good choice, especially in low-resource settings. Its use is controversial, in part because codeine is not effective in a minority of patients who cannot convert it to its active metabolite (morphine), and also because of concerns about potential abuse, and safety in children. To determine the efficacy and safety of codeine used alone or in combination with paracetamol for relieving cancer pain. We searched the Cochrane Central Register of Controlled Trials (CENTRAL; The Cochrane Library 2014, Issue 2), MEDLINE and EMBASE from inception to 5 March 2014, supplemented by searches of clinical trial registries and screening of the reference lists of the identified studies and reviews in the field. We sought randomised, double-blind, controlled trials using single or multiple doses of codeine, with or without paracetamol, for the treatment of cancer pain. Trials could have either parallel or cross-over design, with at least 10 participants per treatment group. Studies in children or adults reporting on any type, grade, and stage of cancer were eligible. We accepted any formulation, dosage regimen, and route of administration of codeine, and both placebo and active controls. Two review authors independently read the titles and abstracts of all studies identified by the searches and excluded those that clearly did not meet the inclusion criteria. For the remaining studies, two authors read the full manuscripts and assessed them for inclusion. We resolved discrepancies between review authors by discussion. Included studies were described qualitatively, since no meta-analysis was possible because of the small amount of data identified, and clinical and methodological between-study heterogeneity. We included 15 studies including 721 participants with cancer pain due to diverse types of malignancy. All studies were performed on adults; there were no studies on children. The included studies were of adequate methodological quality, but all except for one were judged to be at a high risk of bias because of small study size, and six because of methods used to deal with missing data or high withdrawal rates. Three studies used a parallel group design; the remainder were cross-over trials in which there was an adequate washout period, but only one reported results for treatment periods separately.Twelve studies used codeine as a single agent and three combined it with paracetamol. Ten studies included a placebo arm, and 14 included one or more of 16 different active drug comparators or compared different routes of administration. Most studies investigated the effect of a single dose of medication, while five used treatment periods of one, seven or 21 days. Most studies used codeine at doses of 30 mg to 120 mg.There were insufficient data for any pooled analysis. Only two studies reported our preferred responder outcome of 'participants with at least 50% reduction in pain' and two reported 'participants with no worse than mild pain'. Eleven studies reported treatment group mean measures of pain intensity or pain relief; overall for these outcome measures, codeine or codeine plus paracetamol was numerically superior to placebo and equivalent to the active comparators.Adverse event reporting was poor: only two studies reported the number of participants with any adverse event specified by treatment group and only one reported the number of participants with any serious adverse event. In multiple-dose studies nausea, vomiting and constipation were common, with somnolence and dizziness frequent in the 21-day study. Withdrawal from the studies, where reported, was less than 10% except in two studies. There were three deaths, in all cases due to the underlying cancer. We identified only a small amount of data in studies that were both randomised and double-blind. Studies were small, of short duration, and most had significant shortcomings in reporting. The available evidence indicates that codeine is more effective against cancer pain than placebo, but with increased risk of nausea, vomiting, and constipation. Uncertainty remains as to the magnitude and time-course of the analgesic effect and the safety and tolerability in longer-term use. There were no data for children.