In many countries, prescribing guidelines recommend the use of atypical or second-generation antipsychotics (SGAs) in the first-line treatment of individuals with newly diagnosed schizophrenia. This recommendation has increased the utilisation of these agents and, consequently, produced a progressive increase in the proportion of total direct costs in schizophrenia accounted for by drug therapy. In this still-evolving context of care, it becomes relevant to critically investigate the literature base on the relative cost effectiveness of each SGA in comparison with the others, the purpose being to ascertain whether the data reveal any one agent to be truly more cost effective than the others.A systematic search of economic evaluations comparing two or more SGAs yielded 19 studies meeting the inclusion criteria. Of these, 11 were retrospective database or chart review analyses, six were observational prospective or mirror-image studies, and two were randomised clinical trials. Olanzapine and risperidone were compared in 16 studies, two studies compared clozapine, olanzapine and risperidone, and one compared clozapine and risperidone. While experimental studies indicated an absence of differences among the SGAs in terms of total expenditure, database analyses found contrasting evidence. These latter studies, although susceptible to bias and confounding, should theoretically provide an added dimension, in that they are based on observations from 'real world' practice. However, there were too many potential threats to the validity of these analyses to draw a firm conclusion that any one agent is truly more cost effective than the others. In this uncertain situation, clinicians and policy makers should be aware that indirect evidence from independent randomised controlled trials comparing individual SGAs with haloperidol suggested similar cost effectiveness. As healthcare providers in different settings are ultimately the ones who pay for new innovations, it seems appropriate that they commission research into the cost effectiveness of SGAs.