The aim of this study was to determine the efficacy of palliative oxygen for relief of dyspnoea in cancer patients. MEDLINE and EMBASE were searched for randomised controlled trials, comparing oxygen and medical air in cancer patients not qualifying for home oxygen therapy. Abstracts were reviewed and studies were selected using Cochrane methodology. The included studies provided oxygen at rest or during a 6-min walk. The primary outcome was dyspnoea. Standardised mean differences (SMDs) were used to combine scores. Five studies were identified; one was excluded from meta-analysis due to data presentation. Individual patient data were obtained from the authors of the three of the four remaining studies (one each from England, Australia, and the United States). A total of 134 patients were included in the meta-analysis. Oxygen failed to improve dyspnoea in mildly- or non-hypoxaemic cancer patients (SMD=-0.09, 95% confidence interval -0.22 to 0.04; P=0.16). Results were stable to a sensitivity analysis, excluding studies requiring the use of imputed quantities. In this small meta-analysis, oxygen did not provide symptomatic benefit for cancer patients with refractory dyspnoea, who would not normally qualify for home oxygen therapy. Further study of the use of oxygen in this population is warranted given its widespread use.