Stepped-care (SC) models for anxiety disorders are implemented on a large scale and are assumed to be as effective for the greater majority of patients as more intensive treatment schemes. To compare the outcomes of SC and international guideline-based treatment (Treatment as Usual: TAU) for panic disorder, a total of 128 patients were randomized to either SC or TAU (ratio 2: 1, respectively) using a computer generated algorithm. They were treated in four mental health care centres in the Netherlands after therapists had been trained in SC by a senior expert therapist. SC comprised 10-week guided self-help (penand- paper version) followed, if indicated, by 13-week manualized face-to-face cognitive behavioural therapy (CBT), with medication- if prescribed- kept constant. TAU consisted of 23-week regular face-to-face CBT (RCBT) with medication -when prescribed- also kept constant. The means of the attended sessions in the SC condition was 5.9 (SD = 4.8) for ITT and 9.6 (SD = 9.6) for the RCBT condition. The difference in the number of attended sessions between the conditions was significant (t(126) = -3.87, p < .001). Remission rates between treatment conditions did not differ significantly (SC: 44.5%; RCBT: 53.3%) and symptom reduction was similar. Stepping up SC treatment to face-to-face CBT showed a minimal additional effect. Importantly, drop-out rates differed significantly for the two conditions (SC: 48.2%; RCBT: 26.7%). SC was effective in the treatment of panic disorder in terms of symptom reduction and remission rate, but dropout rates were twice as high as those seen in RCBT, with the second phase of SC not substantially improving treatment response. However, SC required significantly less therapist contact time compared to RCBT, and more research is needed to explore predictors of success for guided self-help interventions to allow treatment intensity to be tailored to patients' needs and preferences.