Aims: In patients with chronic heart failure with reduced ejection fraction (HFrEF), myocardial ketone metabolism is increased and short-term treatment with the ketone body 3-hydroxy butyrate (3-OHB) has beneficial haemodynamic effects. In patients with HFrEF, we investigated whether the level of circulating 3-OHB predicted all-cause mortality and sought to identify correlations between patient characteristics and circulating 3-OHB levels. Methods and results: We conducted a cohort study in 218 patients with HFrEF. Plasma 3-OHB levels were measured using high-performance liquid chromatography tandem mass spectrometry. Data on all-cause mortality were obtained by reviewing the patients' medical records, which are linked to the national ‘Central Person Registry’ that registers the timing of all deaths in the country. Mean left ventricular ejection fraction was 35 ± 8.6%, mean age was 67 ± 10 years, 54% were New York Heart Association II, and 27% had type 2 diabetes mellitus. Median follow-up time was 7.3 (interquartile range 6.3–8.4) years. We observed large variations in 3-OHB levels between patients (median 59 μM, range: 14–694 μM). Patients with 3-OHB levels above the median displayed a markedly increased risk of death compared with those with low levels {hazard ratio [HR]: 2.1 [95% confidence interval (CI): 1.3–3.5], P = 0.003}. In a multivariate analysis, 3-OHB predicted mortality independently of known chronic heart failure risk factors [HR: 1.004 (95% CI: 1.001–1.007), P = 0.02] and with a similar statistical strength as N-terminal pro-brain natriuretic peptide (NT-proBNP) [HR: 1.0005 (95% CI: 1.000–1.001), P = 0.02]. For every 100 μmol increase in plasma 3-OHB, the hazard of death increased by 49%. The following factors significantly predicted 3-OHB levels in the univariate analysis: free fatty acids (FFAs) [β: 238 (95% CI: 185–292), P < 0.0001], age [β: 2.43 (95% CI: 1.14–3.72), P < 0.0001], plasma insulin {β: −0.28 [95% CI: −0.54–(−0.02)], P = 0.036}, body mass index {β: −3.15 [95% CI: −5.26–(−0.05)], P = 0.046}, diabetes [β: 44.49 (95% CI: 14.84–74.14), P = 0.003], glycosylated haemoglobin [β: 1.92 (95% CI: 0.24–3.59), P = 0.025], New York Heart Association class [β: 26.86 (95% CI: 5.99–47.72), P = 0.012], and NT-proBNP [β: 0.03 (95% CI: 0.01–0.04), P = 0.001]. In a multivariate analysis, only FFAs predicted 3-OHB levels [β: 216 (95% CI: 165–268), P > 0.001]. Conclusions: In patients with HFrEF, circulating 3-OHB was a strong predictor of all-cause mortality independently of NT-proBNP. Circulating FFAs were the best predictor of 3-OHB levels.