OBJECTIVES: Type 2 diabetes (T2DM) may increase the risk of Parkinson's disease (PD). We evaluated the role of oral anti-hyperglycemic agents (OAA) in any diabetes-PD linkage. METHODS: From the Taiwan National Health Insurance database on 01-01-2000, a representative cohort of 800,000 was obtained between 1996-01-01 and 2007-12-31. Those ≥ 20 years were classified by presence (n = 64,166) or absence (n = 698,587) of T2DM, and whether any OAA (n = 41,003) or not (n = 23,163) was used. Those with T2DM were matched with those diabetes-free by birth-date and gender for the comparison of PD incidence. We considered those ≥ 50 years and matched PD-free diabetes patients with and without OAAs by age, gender, locality, health service, Charlson comorbidity index and T2DM diagnosis-date to avoid 'immortal time bias'. PD incidence densities (PID, per 10,000 person-years) and hazard ratios (HRs) were calculated. RESULTS: HRs (95% confidence interval, CI), related to diabetes-free, were 2.18 (1.27-3.73) and 1.30 (0.77-2.19) for T2DM without and with OAAs. For sulfonylurea alone, PID (95% CI) increased from 58.3 (46.6-70.1) to 83.2 (68.6-97.7), with similar findings by gender, but little difference if metformin was used. The metformin-alone HR (95% CI) was 0.95 (0.53-1.71), sulfonylurea-alone 1.57 (1.15-2.13), and combined therapy 0.78 (0.61-1.01) and these differences persisted when incident PD was excluded for 4 years after T2DM diagnosis. The use of metformin first, in those without insulin, provided an HR of 0.40 (0.17-0.94). CONCLUSIONS: Incident PD risk in T2DM increases 2.2-fold. Sulfonylureas further increase risk by 57%, which is avoided by combination with metformin.