Policy makers require cost-effectiveness studies to help decide which health care programmes to prioritise. The best data for evaluating group-level interventions often come from cluster randomised trials (CRTs), where randomisation is at the level of the cluster, for example the primary care provider, rather than the individual. A fundamental issue raised by the cluster design is that individuals within a cluster tend to be more similar in their characteristics and the care they receive than those in different clusters. However, most published cost-effectiveness analyses (CEA) alongside CRTs ignore the clustering. Unless appropriate methods are used, these studies will report misleading cost-effectiveness results which may lead to inappropriate decisions on resource allocation. We developed and assessed alternative statistical methods for CEA that use CRTs across a wide range of circumstances, including settings with systematic imbalance in baseline covariates and missing data. We have compared the different methods using several case-studies and statistical simulations. This work illustrates the importance of accounting for clustering in CEA that use CRTs, and provides methodological guidance and practical recommendations on the use of alternative appropriate methods.