OBJECTIVES: Drug-eluting stents (DES) and bare-metal stents (BMS) are both used widely in percutaneous coronary interventions. However, cost-effectiveness analyses of DES versus BMS conflict about whether the reduction in repeat revascularizations of DES versus BMS offsets the initial higher treatment costs of DES. A systematic review was performed to examine whether modelling methods influenced the cost-effectiveness of DES versus BMS. METHODS: We reviewed modelling studies published until January 2012 that compared the costs and consequences of DES versus BMS. General information (e.g. funding) and modelling methods used, related to the framing of the economic evaluation (e.g. population and intervention characteristics, time horizon) and parameterisation of the models were extracted from the relevant studies for each of the individual analyses performed in the studies. Associations between these characteristics and the incremental costs and effectiveness were explored using regression analysis. We also examined whether the results were associated with the quality of the models based on the Philips et al. (2006) checklist. RESULTS: 15 eligible studies accounted for 498 separate analyses, in which the incremental cost-effectiveness ratios ranged from DES being dominated by BMS to DES being dominant. The most important predictors significantly associated with these differences were several population and procedure characteristics, funding and assumptions concerning stent efficacy. The results and conclusions of individual studies corresponded with the findings of this meta-level systematic review. Overall quality of the models was moderate (55%±17%) and significantly negatively associated with repeat revascularizations avoided.  CONCLUSIONS: Models are important to obtain valid estimates of the cost-effectiveness of DES versus BMS, and framing decisions (e.g. time horizon) and quality of the models both influence incremental costs and effects. The most influential parameters are identified with this systematic review and we showed the need of examining those parameters and of performing a quality check when interpreting the results.