This article examines the ways in which changes in the treatment environment and in measurement perspectives can affect the evaluation of cost-effectiveness of new medications. In three studies we reexamined data from a clinical trial of haloperidol and clozapine conducted from 1993 to 1996. The results of the studies are as follows: Study 1 found that clozapine treatment was associated with significantly reduced inpatient costs, and increased outpatient costs, suggesting that as systems use less inpatient care and more outpatient care, more effective medications may increase, rather than decrease, costs in sicker patients. Study 2 found that while provider assessments and standard measures favored clozapine over haloperidol, patient responses showed little evidence of a clinical advantage for clozapine and a less favorable side-effect profile. Study 3 found that while annual drug costs in the published trial were estimated to be dollars 4,545 for a full year of clozapine treatment, atypical antipsychotic costs in 2000 were estimated to range from dollars 1,254 to dollars 3,016 in the Department of Veterans Affairs system, and from dollars 2,221 to dollars 8,147 in the private sector. In conclusion, cost-effectiveness, as evaluated in studies like CATIE, will increasingly need to be tied to service system contingencies, environments, and evaluation perspectives.