Background The combination of three cycles of sleep deprivation (SD), light therapy (LT), and lithium has recently been proposed as a possible first-line treatment for bipolar depression. However, it is unclear whether early improvement predicts final response/remission in bipolar depression treated with this regimen. Method We studied 220 consecutively admitted inpatients with a major depressive episode in the course of bipolar disorder. The relation between response to first SD and response/remission at the end of the treatment (day 6) was analyzed using logistic regression analysis. Severity of depression was rated using the Hamilton Depression Rating Scale (HDRS). Clinical response was defined as a ≥50% reduction in HDRS scores, and remission was defined as an HDRS score of ≤7. Results Among the 217 completers, 67.7% showed response and 54.4% reached remission at the end of the treatment. Multiple logistic regression analysis revealed that response after first recovery sleep (day 2) predicted final response and remission at the end of the treatment with high odds ratios (10.9 for response and 8.2 for remission); however, response immediately after the first SD (day 1) did not predict final response or remission. Limitations Whether our results can be generalized to unipolar depression remains uncertain. Conclusion Clinical status after first recovery sleep is a strong predictor of successful final outcome in patients with bipolar depression treated with the combination of repeated SD, LT, and lithium. Recovery sleep may play a role in inducing the antidepressant effect associated with the success of treatment. © 2018 Elsevier B.V.