In 2006, the ADA’s Clinical Practice Recommendations introduced a new recommendation that unless contraindicated metformin be initiated immediately at diagnosis for patients with new-onset type 2 diabetes (T2DM), in addition to nutrition and lifestyle modification. We describe 2006-2008 trends in metformin initiation among new-onset T2DM patients and the characteristics of these metformin users across four geographically diverse integrated delivery systems. The cohort included 31,383 adults aged 18-80 years with new-onset T2DM on the basis of 2 fasting plasma glucose (FPG)>126 mg/dL or random glucose>200mg/dL values (or 1 of each), or 1 glycated hemoglobinA1c reading.6.5%. Subjects were followed for 6-months to determine medication use and subsequent clinical values. Mean A1c (7.8%) and FPG (157 mg/dL) at T2DM detection were relatively stable across time. The proportion of patients starting metformin within 6-months of detection was low and rose only modestly (32% in 2006; 38% in 2008). The median A1c measured closest to metformin initiation changed little over time (8.2% in 2006; 8.4% in 2008). Early initiation of metformin varied across the 4 plans (range 27%-42% in 2008). Metformin accounted for most initial diabetes pharmacotherapy dispensed within 6-months (78% in 2006; 84% in 2008). Early metformin use was much higher in subjects with a physician-assigned diagnosis of diabetes noted after cohort entry (e.g.,48% vs 10% in patients with 2 FPGs >126 mg/dL in 2008). In Cox-proportional hazard models, metformin initiation within 6 months was slightly more likely in later years (HR 1.2, CI 1.1-1.2, 2008 vs. 2006). Other associations included Asian race (HR=0.9, 95% CI:0.8-0.9 vs. whites), BMI>25 kg/m2 (HR=1.3, CI:1.2-1.4), A1c (7-8% HR=4.4, CI:4.1-4.6; >8% HR=11.4, CI:10.8-12.1 vs. A1c<7%). We observed significant but modest changes in practice following modifications to treatment recommendations regarding metformin initiation. For the most part, A1c levels were quite high by the time metformin was initiated. Whether this represents “clinical inertia”, failure to diagnose T2DM based on lab values alone, or lack of conviction that available evidence supports this recommendation is a question for future research.