The clinical response to antidepressant treatment in late-life depression is often delayed and highly variable. Better indicators of antidepressant efficacy are needed early in the course of treatment, so that augmentation strategies or alternative treatments may be initiated. The goal of this study was to evaluate whether the change in the Hamilton depression rating scale (HDRS) after 36 h of total sleep deprivation (TSD) and recovery sleep predicted clinical outcome after 12 weeks of antidepressant treatment, and whether greater predictive value was observed in certain aspects of depressive symptomology. Fifteen elderly patients diagnosed with major depression underwent combined treatment with an initial 36 hours of TSD and a 12-week trial with the antidepressant paroxetine. Six HDRS subscores were evaluated with respect to how the changes after TSD and after one night of recovery sleep correlated with HDRS scores after 12 weeks of treatment. A significant correlation was obtained between the change in the core depressive symptomology subscale from baseline to recovery sleep and the HDRS score at 12 weeks, but the correlation was not significant when evaluating the change from baseline to TSD. These results indicate that the decrease in symptoms after recovery sleep compared with baseline levels (indicating the persistence of the antidepressant response), rather than the symptom reduction after TSD, has greater predictive value with respect to treatment outcome.