Reviews conflict regarding the effectiveness of psycho-oncological therapies at reducing patient anxiety, depression and general distress, suggesting that unknown factors are moderating trial results. This meta-analysis investigates the moderating impact of a large range of socio-demographic, psychological, medical and therapy factors using published and unpublished data from 146 prospective controlled trials, including non-random designs. Preliminary analyses of trial design quality features exposed 2 moderators: recruit screening for psychological distress or history, and the nature of the control condition. These structured a 2 x 2 matrix used to conduct substantive analyses. Admitting only recruits with established baseline distress was found to predict greater effect size, as was excluding patients with a history of distress. Main effects for patients with baseline distress compared with untreated controls, were medium-strong at g = 0.52 - 0.70. Evidence of varying strengths indicated that patients who were older, of lower income, male, single, or suffering from cancer sited elsewhere than breast produced higher effect sizes. Data also highlighted particular stages in the cancer journey: re-entry to normal life at the end of medical treatment, recurrence, and distant disease spread. Findings suggest that risk and distress screening should be employed by both clinicians and researchers. Researchers should also re-direct attention away from unscreened middle class early stage breast cancer patients, towards more vulnerable socio-demographic and medical groups. The potential of using survivors and indirect therapies to effectively and efficiently reach vulnerable groups deserves exploration. Reviewers need to take into account the 2 trial design moderators discovered, and should include non-random controlled trials which may have more access to particularly vulnerable groups because some past conclusions were confounded by the co-variation of study design with sampling characteristics.