Cardiovascular adverse effects are amongst the most serious observed with antidepressant drugs and are often due to effects on cardiac conduction and refractoriness. However, such electrophysiologic effects may not be evident when using conventional electrocardiographic measures. Forty patients with major depressive disorder (according to DSM-III-R criteria) were enrolled in a 6-week double-blind parallel group study of fluoxetine (N = 20) or doxepin (N = 20). Cardiac conduction (QRS duration) and repolarization (corrected QT interval, QTc), were measured using signal-averaged electrocardiograms and 12-lead electrocardiogram at baseline and after 2, 4, and 6 weeks of treatment. Patients taking doxepin (mean daily dosage at 6 weeks 169 +/- 42 mg) were similar to those taking fluoxetine (37 +/- 18 mg) for demographic variables and improvement in depression scores but volunteered more side effects (p = 0.011), especially dry mouth (p < 0.001) and dizziness/lightheadedness (p = 0.005). After 6 weeks, doxepin increased heart rate (69 +/- 12 to 81 +/- 13 beats per minute; p = 0.0003) and prolonged QTc (from 417 +/- 36 to 439 +/- 28 msec; p < 0.03); overall QRS duration was not prolonged but was correlated with serum doxepin concentrations (r = 0.78, p < 0.0001). Fluoxetine had no effect on QTc (428 +/- 24 msec at baseline vs. 430 +/- 24 msec at 6 weeks) or QRS duration (97 +/- 12 msec at baseline vs. 94 +/- 12 msec at 6 weeks). The standard 12-lead electrocardiogram showed no significant change in QRS or QTc for either drug. Using a sensitive measure of electrocardiographic effects, doxepin prolongs repolarization and may slow cardiac conduction. Fluoxetine has no measurable electrocardiographic effects, which suggests an increased safety margin for cardiac adverse effects. The ability of the signal-averaged electrocardiogram to resolve small changes in the electrocardiogram is useful in the assessment of drugs with subtle electrophysiologic effects.