BACKGROUND: Depression has substantial effects on cancer patients' quality of life. Estimates of its prevalence vary widely. We aimed to systematically review published studies to obtain the best estimate of the prevalence of depression in clinically meaningful subgroups of cancer patients. DESIGN: Systematic review that addressed the limitations of previous reviews by (i) including only studies that used diagnostic interviews; (ii) including only studies that met basic quality criteria (random or consecutive sampling, ≥70% response rate, clear definition of depression caseness, sample size ≥100); (iii) grouping studies into clinically meaningful subgroups; (iv) describing the effect on prevalence estimates of different methods of diagnosing depression. RESULTS: Of 66 relevant studies, only 15 (23%) met quality criteria. The estimated prevalence of depression in the defined subgroups was as follows: 5% to 16% in outpatients, 4% to 14% in inpatients, 4% to 11% in mixed outpatient and inpatient samples and 7% to 49% in palliative care. Studies which used expert interviewers (psychiatrists or clinical psychologists) reported lower prevalence estimates. CONCLUSIONS: Of the large number of relevant studies, few met our inclusion criteria, and prevalence estimates are consequently imprecise. We propose that future studies should be designed to meet basic quality criteria and employ expert interviewers.