The ankle brachial index (ABI) is regarded as a predictor of future cardiovascular events. However, the relationship between ABI and incident heart failure (HF) in patients without previous HF is poorly understood. This study aimed to assess the prognostic value of ABI for incident HF in patients without previous HF. The IMPACT–ABI study was a retrospective, single-center, cohort study that enrolled and measured ABI in 3131 patients hospitalized for cardiovascular disease between January 2005 and December 2012. From this cohort, 307 patients were excluded because of previous HF and high (>1.4) ABI. The remaining 2824 patients were stratified into three groups: low ABI (≤0.9), borderline ABI (0.91–0.99), and normal ABI (1.0–1.4). The primary endpoint was hospitalization for HF. Over a mean 4.8-year follow-up, 105 cases of HF occurred. The cumulative incidence of HF was significantly higher in patients with low and borderline ABIs than in those with normal ABI (19.3 vs. 21.0 vs. 10.4 %, log rank P <0.001). In multivariate Cox proportional hazard analysis, low ABI and borderline ABI were independent predictors of incident HF [hazard ratio (HR) 3.00; 95 % confidence interval (CI) 1.70–5.28; P < 0.001 and HR 2.68; 95 % CI 1.35–5.34; P = 0.005, respectively]. In conclusion, low and borderline ABI were strong predictors for future incident HF in patients without previous HF.