OBJECTIVES: 09tl6829page Results of intervention studies in patients with type 2 diabetes have led to concerns about the safety of aiming for normal blood glucose concentrations. We assessed survival as a function of HbA1c in people with type 2 diabetes. METHODS: 09tl6829page Two cohorts of patients aged 40 years and older with type 2 diabetes were generated from the THIN database of UK general practice records from 1986 to 2008. Those with diabetes secondary to other causes were excluded. We identified 37,626 and 22,769 patients commenced on metformin or sulphonylurea monotherapy respectively; 24,784 intensified from oral monotherapy to combination therapy with oral blood-glucose lowering agents; and 15,855 changed to regimens that included insulin. All-cause mortality was the primary outcome. Age, sex, smoking status, cholesterol, cardiovascular risk, and general morbidity were identified as important confounding factors, and Cox survival models were adjusted for these factors accordingly. RESULTS: 09tl6829page Compared to the HbA1c response category 7.0 to ≤7.5, those achieving the tightest glycaemic control (mean post-index HbA1c <6.1%) had a substantially elevated adjusted hazard ratio for all-cause mortality (HR=1.52 [95%CI 1.39 to 1.66]), as did those with the poorest glycaemic control (HbA1c response >10%, HR 1.45 [1.30 to 1.63]).  A ‘U-shaped’ association between glycaemic control and survival emerged in all treatment cohorts for which the adjusted HR in those achieving tightest glycaemic control (cf. referent) ranged from 1.44 (1.20 to 1.72) for metformin monotherapy to 2.61 [2.05 to 3.32] for the insulin-treated cohort. CONCLUSIONS: 09tl6829page Low and high mean HbA1c values were associated with increased all-cause mortality corroborating the findings of the ACCORD trial.  Diabetes guidelines and current health economic models may need revision to reflect the safety implications of very tight glycaemic control in Type 2 diabetes.