OBJECTIVES: Opioid dependence is a serious and costly medical condition that can occur with regular opioid use. We conducted a systematic review of published model-based economic evaluations of drug substitution therapy in treating non-medical opioid dependence. METHODS: Literature searches were conducted in March 2015 in 8 electronic databases and supplemented by hand-searching reference lists and searches on 6 health technology assessment (HTA) agency websites. The selection criteria included: A population dependent on opioids and receiving opioid substitution therapy or maintenance therapy. The intervention included any pharmacological maintenance therapy and the comparator included any pharmacological maintenance regimen, including placebo or no treatment. The outcomes and study types included health economic models of any type. RESULTS: After removal of duplicates, 2,163 citations were retrieved, of which 63 progressed to full-text review. Of these, 19 publications of 18 unique models were included in the review. These 18 models used a wide range of modelling approaches, including Markov models (n=4), decision tree with Monte Carlo simulations (n=4), decision analysis (n=3), dynamic transmission models (n=3), decision tree (n=1), cohort simulation (n=1), Bayesian (n=1), and Monte Carlo simulations for sensitivity analysis (n=1). Time horizons ranged from 6 months to a lifetime. The most common evaluation was cost-utility analysis reporting cost per quality-adjusted life-year (n=11), followed by cost-effectiveness analysis (n=4), budget impact analysis/cost comparison (n=2) and cost-benefit analysis (n=1). Countries modelled were the US (n=11), UK (n=4), Spain (n=1), Vietnam (n=1) and New Zealand (n=1). A range of perspectives were modelled, including societal and healthcare systems. CONCLUSIONS: This review identified 8 different modelling structures with a range of perspectives, time horizons and inputs, illustrating that there is no single preferred approach. Further research is needed into the advantages and disadvantages of the different modelling approaches in this disease area.