Background: Predicting rTMS nonresponse could be helpful in sparing patients from futile treatment, and in improving use of limited rTMS resources. While several predictive biomarkers have been proposed, few are accurate for individual-level prediction; none have entered routine use. An alternative approach in pharmacotherapy predicts outcome from early response; patients showing minimal (e.g., ≤20%) improvement at 2 weeks can be predicted as nonresponders with negative predictive values (NPV) > 80–90%. This approach has recently been extended to ECT, but never before to rTMS. Objective: To assess the accuracy of 2-week clinical response in predicting rTMS treatment outcome. Methods: We reviewed clinical symptom scores for 101 patients who underwent 20 sessions of dorsomedial prefrontal rTMS for unipolar major depression in a naturalistic retrospective case series, defining nonresponders both at the conventional