OBJECTIVES:: The use of cost-minimisation analysis, based on a negative result from a clinical trial, has recently been questioned. Instead, it has been argued that cost-minimisation analysis should only be undertaken as part of an equivalence trial, designed to confirm the absence of a meaningful difference between treatments. The ‘non-inferiority' trial is becoming a popular method of comparing a new treatment to an existing standard as non-inferiority may be an acceptable outcome for licensing purposes. Non-inferiority trials are designed to show that a new treatment is no less effective than an existing treatment – in reality, it may be more effective or have a similar effect (within a margin of clinical equivalence). METHODS: In this paper, the role of the non-inferiority trial design is critically appraised as a vehicle for economic evaluation. RESULTS: A number of issues are noted in relation to non-inferiority designs that limit their usefulness for making judgements concerning the value for money of treatments. The margin of clinical equivalence is essentially arbitrary and is rendered meaningless once costs (both in terms of resources and the health effects of any adverse events) are considered. From an economic perspective, acceptance of a new treatment can only be recommended if the new treatment is both ‘non-inferior' in both clinical and cost terms, however, the role of the margin of clinical equivalence (and potentially a similar margin of cost equivalence) serves to cloud the comparison with the traditional concept of ‘dominance' in economic studies. CONCLUSION: The design of trial in terms of ‘superiority' or ‘non-inferiority' does not change the fact that separate and sequential tests of hypothesis for costs and effects independently are to be avoided in economic evaluation. The recommendation remains that estimation and not hypothesis testing should be the key to interpreting cost-effectiveness studies of new treatments.