OBJECTIVE：To compare the cost-effectiveness of insulin analogues and human insulin in the treatment of type 2 diabetes mellitus ，to reduce economic burden of patients ，and to provide new ideas for non-randomized pharmacoeconomic research. METHODS ：Retrospective data of a Third-grade Class-A hospital were adopted and 533 patients with type 2 diabetes were taken as object to balance the difference between groups through propensity score matching ，and construct net-benefit regression framework so as to analyze the cost-effectiveness of insulin analogues and human insulin in the treatment of type 2 diabetes mellitus. RESULTS ：There was a positive relationship between the efficacy and cost of the two therapeutic schemes ，i.e. insulin analogues were more effective than human insulin ，and the difference of total effective rate between the two schemes was 14.5％. The corresponding treatment cost of insulin analogues was higher than that of human insulin ，and the average total cost difference was 960.3 yuan. The cost of insulin analogues was 66.23 yuan more than that of human insulin for each additional unit of clinical effect （total effective rate ）. Results of net-benefit regression analysis showed that critical value of willingness to pay of therapy plan was 16 947.5 yuan. CONCLUSIONS ：When the willingness to pay is more than 16 947.5 yuan，the insulin analogue scheme is suitable ；on the contrary ，the human insulin analogue scheme is suitable. The propensity scoring technology used for the net-benefit regression analysis can broaden the source of avaliable data for non-randomized pharmacoeconomic evaluation.