Objective: Electroconvulsive therapy (ECT) is a valuable treatment option in severely depressed elderly patients. Structural abnormalities in the brain, such as white matter hyperintensities, medial temporal lobe atrophy (MTA), or global cortical atrophy, may influence therapeutic response. The respective value of these factors in response prediction is unclear. Method: In a naturalistic clinical cohort of 81 elderly patients diagnosed with DSM-IV major depressive disorder, magnetic resonance imaging (MRI) was recorded and rated before ECT treatment. The study was conducted at the clinic for Geriatric Psychiatry of the VU University Medical Center/Stichting Buitenamstel Geestgronden, Amsterdam, The Netherlands, over a 5-year period (2001-2006). Severity of depressive symptoms was measured by using the Montgomery-Asberg Depression Rating Scale (MADRS). Response to ECT was defined as a decrease of at least 50 percent on the MADRS, and remission was defined as a score below 10 points on the MADRS. Results: Patients with moderate or severe MTA had a lower mean percentage decrease in MADRS scores after ECT (37.9% in those with MTA, compared to 66.2% in those without MTA, P = .008). Patients without MTA had a 3 times greater chance of remitting from their depression compared to patients with moderate or severe MTA, ie, the hazard ratio for remission was 3.22 (95% CI, 1.30 to 7.69, P = .01). In contrast, no differences in change in MADRS scores were found for white matter hyperintensities or global cortical atrophy. Conclusions: Medial temporal lobe atrophy - not white matter hyperintensities or global cortical atrophy - contributes to poor response to ECT in severely depressed elderly patients. These findings suggest that assessment of MTA in severely depressed elderly patients may be useful in the prediction of potential ECT response. © Copyright 2010 Physicians Postgraduate Press, Inc.