Objective: Study findings of the relationship of each arterial stiffness index with incident heart failure (HF) are conflicting. We aimed to compare the association between the indices of arterial stiffness and the risk of HF.Methods: We analysed 3,034 patients from a prospective cohort that enrolled patients with high cardiovascular risk. They underwent brachial-ankle pulse wave velocity (baPWV), brachial pulse pressure (PP), carotid-femoral pulse wave velocity (cfPWV), and central PP measurements.Results: Over a median follow-up of 4.7 years (interquartile range, 3.4–5.8 years), 65 HF events occurred. The incidence rate of HF was 4.7 per 1,000 person-years [95% confidence interval (CI), 3.7–6.0]. There was no difference in baPWV in those with and without HF events (1,561 ± 401 and 1,520 ± 321 cm/s, respectively, P = 0.415); however, there was a significant difference in brachial PP (63.2 ± 16.9 vs. 52.3 ± 11.5 mmHg, P < 0.001), cfPWV (11.0 ± 3.1 vs. 9.4 ± 2.4 m/s, P < 0.001) and central PP (56.6 ± 19.9 vs. 42.9 ± 13.8 mmHg, P < 0.001). In the multivariable-adjusted model, brachial PP [hazards ratio (HR) per standard deviation unit (SDU), 1.48; 95% CI, 1.19–1.84, P < 0.001], cfPWV (HR per SDU, 1.29; 95% CI, 1.02–1.63, P = 0.032) and central PP (HR per SDU, 1.44; 95% CI, 1.17–1.78; P < 0.001) were associated with incident HF, but baPWV was not (HR per SDU, 0.83; 95% CI, 0.63–1.10; P = 0.198). In the receiver operating characteristic analysis, the area under the curve (AUC) of brachial PP (P < 0.001), cfPWV (P = 0.003) or central PP (P = 0.001) was larger than that of baPWV, and there was no difference in the AUCs of brachial PP, cfPWV and central PP.Conclusion: Among arterial stiffness indices, brachial PWV was less associated with the risk of heart failure, and brachial PP and measures representing central hemodynamics were highly associated with incident HF.