Purpose -This paper aims to examine how direct-to-consumer advertising (DTCA) in terms of endorser selection and message tonality affect patients' self-efficacy, response efficacy, and compliance. Design/methodology/approach -This study employed a 3 (Endorser: physician, patient, or celebrity) £ 2 (Tonality: supportive vs threatening) experimental design. Subjects were 1,211 people with diabetes from Germany. Findings -First, the study shows that the interaction between message sender and tonality significantly affects all dependent variables such as self-efficacy, response efficacy, and patient compliance. Second, physicians as endorsers work best when they use unfavorable, threatening arguments. The results are significant for all dependent variables such as self-efficacy, response efficacy, and patient compliance. Most surprisingly, patients judge attitude significantly higher if physicians use threatening instead of supportive argumentation. Third, tonality does not play a dominant role for patients as person-based testimonials. Fourth, a celebrity performs best by using a supportive message. With regard to stimulating health outcome (self-efficacy, response efficacy, and patient compliance) the celebrity has a significant impact in the supportive rather than in the threatening condition. Originality/value -This is the first empirical study that examines the effects of DTCA in terms of endorser selection and message tonality on patients' self-efficacy, response efficacy, and compliance.