A cross-national randomised trial of alprazolam for chronic panic disorder with agoraphobia was run. Compared with previous trials it had three new features: an exposure therapy contrast group, a six-month treatment-free follow-up, and a low rate of early placebo drop-outs ('non-evaluables'). The dose of alprazolam was high (5 mg/day). The 154 patients had eight weeks of: alprazolam and exposure (combined treatment); or alprazolam and relaxation (a psychological placebo); or placebo and exposure; or placebo and relaxation (double placebo). Drug taper was from weeks 8 to 16. Follow-up was to week 43. Results were similar at both sites. Treatment integrity was good. All four treatment groups, including double placebo, improved well on panic throughout. On non-panic measures, by the end of treatment, both alprazolam and exposure were effective, but exposure had twice the effect size of alprazolam. During taper and follow-up, gains after alprazolam were lost, while gains after exposure were maintained. Combining alprazolam with exposure marginally enhanced gains during treatment, but impaired improvement thereafter. The new features put previous trails in a fresh light. By the end of treatment, though gains on alprazolam were largely as in previous studies, on phobias and disability they were half those with exposure. Relapse was usual after alprazolam was stopped, whereas gains persisted to six-month follow-up after exposure ceased. Panic improved as much with placebo as with alprazolam or exposure.