Our aim was to describe the clinical features of 35% CO2-induced panic attacks in patients with panic disorder (PD) (Diagnostic and Statistical Manual and Mental Disorders, Fourth Edition) and compare them with the last spontaneous panic attack in patients with PD who had not had a panic attack after the 35% CO2 challenge test. We examined 91 patients with PD submitted to the CO2 challenge test. The test consisted of exhaling as fully as possible, took a fast vital capacity breath, held their breath for 8 seconds, exhaled, and then repeated the fast vital capacity breath, again holding for 8 seconds. The patients inhaled the 35% CO2/65% O2 mixture or atmospheric compressed air, randomly selected in a double-blind design. Scales were applied before and after the test. A total of 68.1% (n = 62) patients with PD had a panic attack (responders) after the CO2 test (chi2(1) = 25.87, P = .031). The last spontaneous panic attack and the symptom profile from the patients with PD who had not had a panic attack after the test (n = 29, 31.9%) were described to compare. The responders had more respiratory symptoms (chi2(1) = 19.21, P < .001), fulfilling the criteria for respiratory PD subtype (80.6%); the disorder started earlier (Mann-Whitney, P < .001), had a higher familial prevalence of PD (chi2(1) = 20.45, P = .028), and had more previous depressive episodes (chi2(1) = 27.98, P < .001). Our data suggest that there is an association between respiratory PD subtype and hyperreactivity to a CO2 respiratory challenge test. The responders may be a subgroup of respiratory PD subtype with future diagnostic and therapeutic implications.