OBJECTIVES To identify and assess mathematical models predicting the relative effectiveness of drug treatments in “real world” populations, based on data from randomized control trials and other sources of evidence. METHODS Systematic review of mathematical modelling studies addressing the step from relative efficacy to relative effectiveness. We identified eligible studies through electronic and manual searches of MEDLINE and EMBASE databases, selected websites and reference lists of relevant papers. Two reviewers screened the articles independently and extracted study characteristics such as model type, disease area, validation and software used via an extraction form. RESULTS Eight papers met the inclusion criteria covering four broad modelling approaches: multi-state models, simulation-based approaches, mechanistic models, and classical regression based models. The multi-state models were the predominant class of models. These models are defined as time-dependent stochastic processes with discrete event space. Most examples belonged to the special case of Markov multi-state models. Multi-state models were applied at the level of population groups or at the individual patient level. The other approaches we identified were less frequent. Discrete event simulation was used in one paper. This approach is entirely based on simulations. One article described a mechanistic model based on ordinary differential equations, which are typically derived from biological knowledge and first principles. Finally, more classical regression techniques from survival analysis were used in two papers. Six articles included models built for cardiovascular indications, the remaining ones covered oncology and neurosciences. Internal or external model validation was presented in six papers, while two papers considered only sensitivity analysis to evaluate the model performance. CONCLUSIONS This review shows the range of models currently used for predicting the relative effectiveness of drug interventions in real world patient populations. They complement the available tools for evidence synthesis in comparative effectiveness research.