BACKGROUND: The EQ-5D and the SF-6D are examples of commonly used generic preference-based instruments for assessing health-related quality of life (HRQoL). However, their suitability for mental disorders has been repeatedly questioned. OBJECTIVE: To assess the responsiveness and convergent validity of the EQ-5D-3L and SF-6D in patients with depressive symptoms. METHODS: The data analyzed were from cardiac patients with depressive symptoms and were collected as part of the SPIRR-CAD (Stepwise Psychotherapy Intervention for Reducing Risk in Coronary Artery Disease) trial. The EQ-5D-3L and SF-6D were compared with the HADS (Hospital Anxiety and Depression Scale) and PHQ-9 (Patient Health Questionnaire) as disease-specific instruments. Convergent validity was assessed using Spearman's rank correlation. Effect sizes were calculated and ROC analysis was performed to determine responsiveness. RESULTS: Data from 566 patients were analysed. The SF-6D correlated considerably better with the disease-specific instruments (|r|= 0.63-0.68) than the EQ-5D-3L (|r|= 0.51-0.56). The internal responsiveness of the SF-6D was in the upper range of a small effect (ES: - 0.44 and - 0.47), while no effect could be determined for the EQ-5D-3L. Neither the SF-6D nor the EQ-5D-3L showed acceptable external responsiveness for classifying patients' depressive symptoms as improved or not improved. The ability to detect patients whose condition has deteriorated was only acceptable for the EQ-5D-3L. CONCLUSION: Overall, both the convergent validity and responsiveness of the SF-6D are better than those of the EQ-5D-3L in patients with depressive symptoms. The SF-6D appears, therefore, more recommendable for use in studies to evaluate interventions for this population.