OBJECTIVE: To identify utility weights for use in cost-utility analyses of antidepressants. METHODS: Systematic search of MEDLINE using search terms for depression and utility/preference weights. Utility weights by depression severity and changes in utility by response to treatment were abstracted. Methods used to derive utilities were cross-tabulated with the values obtained. RESULTS: Six published studies were reviewed. Three studies obtained utility weights using the standard gamble (SG) method, one study used the SG and time trade-off (TTO) methods, one study used the EQ5D TTO weights, and one study used the quality of well being (QWB) scale weights. One of the SG studies compared utility for those with and without depression over a 10-year time horizon (0.942 (standard deviation (SD) 0.159) versus 0.963 (SD 0.144)). The other three SG studies compared utility for different depression severity and, depending on the whether the SG lottery was presented for temporary or lifetime health states, estimated utilities for severe depression were between 0.09 (SD 0.02) and 0.813 (SD 0.209) and for mild depression were between 0.59 (SD 0.02) and 0.871 (SD 0.184). Three of the six reviewed studies compared the gain in utility for those who responded to treatment to those who did not. The gain in utility for responders compared to non-responders in these studies was: 0.053 at 1 year using the SG lifetime method; 0.180 at 4 months using the QWB weights; and 0.220 at 2 months using the EQ5D TTO weights. CONCLUSIONS: Published estimates of utility weights for people with depression and of the gains in utility in people recovering from depression vary considerably depending on the method of assessment. We recommend that utility gains for antidepressant treatments be estimated using SG for temporary health states along with sensitivity analyses using alternative methods of utility assessment.