This meta-analysis examined whether effects of psychosocial interventions on psychological distress in cancer patients are conditional upon pre-intervention distress levels. Published articles and unpublished dissertations between 1980 and 2005 were searched for interventions reporting the Hospital Anxiety and Depression Scale (HADS) or the Spielberger State-Trait Anxiety Inventory (STAI). Multilevel mixed-effects modeling was used to meta-analyze effect-sizes separately for the HADS (27 trials, 2,424 patients) and STAI (34 trials, 2,029 patients). Pre-intervention distress significantly moderated intervention effects, explaining up to 50% of the between-study effect-size variance: effects on anxiety and depression were generally negligible when pre-intervention distress was low and pronounced when it was high. These results could not be explained by differences in intervention type, setting, dose, and whether intervention was targeted at distressed patients. Psychosocial interventions may be most beneficial for cancer patients with elevated distress. Future research should identify which treatment components are most effective for these patients to facilitate optimal treatment tailoring and cost-effective health care.