We conducted an exploratory post hoc study that compared the cost effe
ctiveness of five treatments for bulimia nervosa: 15 weeks of cognitiv
e behavioral therapy (CB) followed by three monthly sessions, 16 weeks
(Med(16)) and 24 weeks (Med(24)) of desipramine (less than or equal t
o 300 mg/day), and CB combined with dysproteinemia for those durations
(Combo(16) and Combo(24)). We illustrate how a treatment's cost effec
tiveness varies according to when evaluation is done and how effective
ness and cost are defined. At 32 weeks, Med(16) appears the most cost-
effective treatment, and Combo(16) appears the least. At 1 year, Med(2
4) appears the most cost-effective treatment, and Combo(16) appears th
e least. Using this post hoc analysis as an example, we discuss the pi
tfalls and limitations of cost-effectiveness analysis of psychiatric t
reatments.