Benzodiazepines (BZs) are commonly used in conjunction with cognitive behavioral therapy (CBT) in the treatment of panic disorder with agoraphobia (PDA). However, empirical evidence provides little support for the utility of this combined treatment approach over CBT alone. Westra and Stewart [Clin. Psychol. Rev. 18 (1998) 307] have proposed that pm or as-needed use of BZs may inhibit positive CBT outcome to a greater extent than regularly scheduled BZ use. Using a naturalistic design, the present study investigated the impact of manner of BZ use on treatment outcome from CBT in 43 patients with PDA. Among various BZ parameters (chronicity, frequency, dose, and frequency of prn use), pm use of BZs for coping with anxiety symptoms was a significant negative predictor of degree of change in both anxiety sensitivity and anxious arousal from pre- to post-CBT. Although no significant between-group differences were evident in pre-treatment symptomatology, unmedicated subjects demonstrated the most positive overall CBT outcome, while pm BZ users evidenced the fewest gains. Regular BZ users were generally not significantly differentiated from unmedicated subjects in CBT outcome and both tended to obtain post-treatment scores in the nonclinical range. Implications of these findings for clinical management of BZ use throughout CBT for PDA are discussed.