This paper reviews the role of patient preferences within the framework of cost-effectiveness analysis (CEA). CEA typically adopts a system-wide perspective by focusing upon efficiency across groups in the allocation of scarce healthcare resources, whereas treatment decisions are made over individuals. However, patient preferences have been shown to have a direct impact on the outcome of an intervention via psychological factors or indirectly via patient adherence/compliance rates. Patient values may also be in conflict with the results of CEA through the valuation of benefits. CEA relies heavily on the QALY model to reflect individual preferences, although the healthy year equivalent offers an alternative measure that may be better at taking individual preferences into account. However, both measures typically use mean general population or mean patient values and therefore create conflict with individual-level preferences. For CEA to reflect practice, it must take into account the impact of individual patient preferences even where general population preferences are used to value the benefits of interventions. Patient preferences have implications for cost effectiveness through costs and outcomes, and it is important that cost-effectiveness models incorporate these through its structure (e.g. allowing for differing compliance rates) and parameter values, including clinical effectiveness. It will also be necessary to try to predict patient preferences in order to estimate any impact on cost effectiveness through analyses of revealed and stated preference data. It is recognized that policy makers are concerned with making interventions available to patients and not forcing them to consume healthcare. One way of moving towards this would be to adopt a two-part decision process: the identification of the most cost-effective therapy using mean general population values (i.e. the current rule), then also making available those treatments that are cheaper than the most cost-effective therapy.