Both noradrenergic and serotonergic systems have been implicated in the pathophysiology of panic disorder. The advent of selective serotonin (5-HT) reuptake inhibitors (SSRIs) (e.g. citalopram) and, more recently, selective noradrenergic (NA) reuptake inhibitors (NRIs) (e.g. reboxetine) has provided potentially important avenues of treatment for the disorder. To date, the comparative efficacy of selective NA and 5-HT reuptake inhibitors for panic disorder remains unresolved. Nineteen patients with panic disorder were randomized in a single-blind, cross-over design to either citalopram or reboxetine for 8 weeks and after a 2-week washout were switched to the other study drug. At week 18, seven of 13 patients (54%) in the intent-to-treat sample responded to reboxetine and nine of 11 patients responded to citalopram (82%). Both citalopram and reboxetine led to significant improvements in panic attack severity with no apparent between-drug differences in efficacy. However, citalopram demonstrated superior efficacy in treating depressive symptoms. One non-responder to citalopram responded to reboxetine and three non-responders to reboxetine responded to citalopram. Although SSRIs are viewed as a first-line treatment for panic disorder, these results suggest that a NA agent such as reboxetine may also have a role. These data also suggest an advantage for citalopram in treating comorbid depressive symptoms, although some patients may respond preferentially to an SSRI and other patients to an NRI.