Randomized controlled trials support the antidepressant efficacy of transcranial magnetic stimulation (TMS); however, there is individual variability in the magnitude of response. Examination of response predictors has been hampered by methodological limitations such as small sample sizes and single-site study designs. Data from a multisite sham-controlled trial of the antidepressant efficacy of TMS provided an opportunity to examine predictors of acute outcome. An open-label extension for patients who failed to improve provided the opportunity for confirmatory analysis. Treatment was administered to the left dorsolateral prefrontal cortex at 10 pulses per second, 120% of motor threshold, for a total of 3000 pulses per day. Change on the Montgomery-Asberg Depression Rating Scale after 4 weeks was the primary efficacy outcome. A total of 301 patients with nonpsychotic unipolar major depression at 23 centers were randomized to active or sham TMS. Univariate predictor analyses showed that the degree of prior treatment resistance in the current episode was a predictor of positive treatment outcome in both the controlled study and the open-label extension trial. In the randomized trial, shorter duration of current episode was also associated with a better outcome. In the open-label extension study, absence of anxiety disorder comorbidity was associated with an improved outcome, but duration of current episode was not. The number of prior treatment failures was the strongest predictor for positive response to acute treatment with TMS. Shorter duration of current illness and lack of anxiety comorbidity may also confer an increased likelihood of good antidepressant response to TMS.