OBJECTIVE: To prospectively examine the magnitude and predictors of diabetes-attributable non-blood glucose-lowering (non-BGL) medication costs in type 2 diabetes. RESEARCH DESIGN AND METHODS: Detailed data from 593 community-dwelling patients were available over 4.3 +/- 0.4 years. Diabetes-attributable costs (in year 2000 Australian dollars [A$]) were calculated by applying a range of attributable proportions for each complication for which medication was prescribed. RESULTS: Non-BGL medications accounted for 75% of all prescription medication costs over the study period, and one-third were attributable to diabetes. The median annual cost (in A$) of non-BGL medications per patient increased from A$220 to A$429 over 4 years (P < 0.001), whereas the diabetes-attributable contribution increased from A$31 (range 15-40) to A$159 (range 95-219) per patient (P < 0.001). Diabetes-attributable hospital costs remained stable during the study. Diabetes-attributable non-BGL costs were skewed and, therefore, square root transformed before regression analysis. Independent baseline determinants of square root cost/year were coronary heart disease, systolic blood pressure, total serum cholesterol, ln(serum triglycerides), ln(albumin-to-creatinine ratio), serum creatinine, education, and, negatively, male sex and fasting plasma glucose (P </= 0.043; R(2) = 29%). Projected to the Australian population, diabetes-attributable non-BGL medication costs for patients with type 2 diabetes totaled A$79 million/year. CONCLUSIONS: The median annual cost of diabetes-attributable non-BGL medications increased fivefold over 4 years. This increase was predicted by vascular risk factors and complications at baseline. Better-educated patients had higher costs, probably reflecting improved health care access. Men and patients with higher fasting plasma glucose levels had lower costs, suggesting barriers to health care and/or poor self-care. The contemporaneous containment of hospital costs may be due to the beneficial effect of increased medication use.