Background: Blood pressure (BP) control has been shown to reduce the risk of heart failure (HF) among diabetic patients; however, it is not known whether the lowest clinical BP achieved ultimately results in the lowest risk of HF in diabetic patients. Methods: We performed a prospective cohort study which included 17,181 African American and 12,446 white diabetic patients without established coronary heart disease and HF at diabetes diagnosis. Cox proportional hazards regression models were used to estimate the association of different levels of BP stratification with incident HF. Results: During a mean follow up of 6.5 years, 5,089 incident HF cases were identified. The multivariable-adjusted hazard ratios of HF associated with different levels of systolic/diastolic BP (<110/65, 110-119/65-69, 120-129/70-80, 130-139/80-90 [reference group], 140-159/90-100, and >= 160/100 mmHg) were 1.79 (95% confidence interval [CI] 1.53-2.11), 1.34 (95% CI 1.16-1.53), 1.02 (95% CI 0.92-1.13), 1.00, 1.04 (95% CI 0.95-1.12), and 1.26 (95% CI 1.16-1.37) using baseline BP measurements, and 2.63 (95% CI 2.02-3.41), 1.84 (95% CI 1.59-2.13), 1.25 (95% CI 1.14-1.37), 1.00, 1.11 (95% CI 1.03-1.19), and 1.32 (95% CI 1.20-1.44) using an updated mean value of BP during follow-up, respectively. The U-shaped association was confirmed in both patients who were and were not taking antihypertensive drugs, and in incident systolic HF (ejection fraction <= 40%) and incident HF with a preserved ejection fraction (ejection fraction >40%). Conclusions: The current study suggests a U-shaped association between observed BP and the risk of HF among diabetic patients. (C) 2014 Elsevier Ireland Ltd. All rights reserved.