Cardiovascular Effects of Treatment With the Ketone Body 3-Hydroxybutyrate in Chronic Heart Failure Patients

We studied 24 HFrEF patients and 10 age-matched volunteers. For HFrEF patients, the inclusion criteria were (a) left ventricular ejection fraction (LVEF) ≤40% and optimal, stable HFrEF treatment according to guidelines; (b) New York Heart Association (NYHA) class II-III; and (c) ability to give informed consent. Exclusion criteria were uncorrected cardiac valvular disease, signs or symptoms of myocardial infarction within the previous 3 months, and treatment for diabetes or HbA1c level >53 mmol/mol. Age-matched volunteers had normal LVEF (defined as >50%), ECG, physical examination, systolic blood pressure <160 mm Hg, and blood samples (including NT-proBNP [N-terminal pro-B-type natriuretic peptide], cholesterol, creatinine, transaminases, electrolytes, and blood cell count) and were not allowed to have any cardiac symptoms (angina, dyspnea, palpitations and edema). Treatment with statins, antihypertensive, and antiplatelet medication was allowed. All patients were recruited from the outpatient HF clinic at Aarhus University Hospital, Denmark, and age-matched volunteers by advertisements.

Four studies addressed each of the aims. In all studies, the participants fasted overnight and avoided strenuous exercise for 48 hours prior to the investigations. At 7:15 am, venous catheters were inserted into participants’ cubital veins. The local pharmacy prepared the dissolved sodium (Na)-3-OHB at a 7.5% concentration. Placebo and 3-OHB were provided in red bags for concealment. Potassium chloride (KCl) was added to a concentration of 60 mM. To minimize endogenous production of 3-OHB during prolonged fasting throughout the study and to mimic a fed postprandial condition, a low-dose insulinemic euglycemic clamp (0.3 IE insulin · kg^-1 · h^-1) was administered to all participants. Glucose (20% solution, 60 mM KCl) was infused to maintain euglycemia.

Study 1: 16 HFrEF patients were investigated once for 6 hours and received either 3-OHB infusion (0.18 g · kg^-1 · h^-1; Na-3-OHB, Gold Biotechnology Inc., USA [racemic mixture 50/50 D and L]) or placebo (isotonic saline) at an equivalent volume for 3 hours in a randomized crossover design. The participants were blinded to the order of the infusion and were continuously monitored with pulmonary artery catheter, noninvasive blood pressure measurements, ECG, repeated blood sampling, and echocardiography. They received an oral load of 60 mmol KCl when randomized to initiate with 3-OHB and 20 mmol when randomized to initiate with placebo.

Study 2: After Study 1 was completed, 8 additional HFrEF patients were enrolled consecutively in the dose-response study. These participants were studied once for 6 hours and received placebo infusion (isotonic saline 1.2 mL · kg^-1 · h^-1) followed by 3-OHB 0.045 g · kg^-1 · h^-1 with coadministration of saline 0.6 mL · kg^-1 · h^-1 and finally 3-OHB 0.09 g · kg^-1 · h^-1. Each dosage level lasted 2 hours and equivalent volumes were infused. Participants were monitored as described in Study 1 and received 30 mmol KCl orally.

Study 3: 10 of the HFrEF patients enrolled in Study 1 were consecutively requested to participate in additional 11C-acetate PET examinations on a separate day. These subjects then received either 3-OHB infusion or placebo in a randomized single-blinded crossover design, as well as KCl administered in the same way as in Study 1. 11C-acetate PET was conducted to evaluate MEE, MVO₂, and MBF.

Study 4: 10 age-matched subjects underwent interventions and PET examinations similar to those performed in Study 3 (Figure 1).

Central venous pressure (CVP), mean pulmonary arterial pressure (mPAP), and mean pulmonary capillary wedge pressure (PCWP) were measured at end-expiration. Mixed venous saturation (SVO₂) was measured continuously, and cardiac output (CO) was assessed by thermodilution technique and averaged over 3 measurements. Heart rate (HR) and blood pressure were measured noninvasively. Stroke volume (SV) (SV=CO/HR), mean arterial pressure (MAP), and systemic (SVR=[MAP-CVP]/CO) and pulmonary (PVR=[mPAP-PCWP]/CO) vascular resistance were calculated. All hemodynamic measurements were performed with the patient in the supine position.

End diastolic and systolic volumes were measured by Simpson biplane methods to assess LVEF. Numeric global longitudinal strain (GLS), mitral inflow velocities (E and A), and mitral plane velocities in the lateral mitral annulus (e’ and s’) were measured. Velocity time integral of the left ventricular outflow track was assessed. A GE Vivid E9 (GE Healthcare, USA) was used for echocardiographic acquisition and EchoPAC 13 (GE-Vingmed Ultrasound, Norway) for analysis. All measures were averaged over 3 heart beats, and the assessor was blinded to allocation and study.¹⁰

A total of 400 MBq 11C-acetate was injected, followed by list-mode PET recording for 27 minutes on a Biograph TruePoint TrueV 64 PET/computed tomography scanner (Siemens, Germany). Blood pressure and HR were measured at 1, 5, 10, and 20 minutes after injection. The examination was performed to obtain the global clearance rate (k₂) and to calculate MVO₂ as previously described:¹¹

Dynamic data sets were analyzed to estimate SV by use of aQuant Research.¹² HR and MAP were measured noninvasively. Left ventricular mass and CO were obtained from the PET dataset, and MEE was calculated as:¹³

MBF (mL · g^-1 · min^-1) was estimated by measuring k₁ and correcting for rate pressure product as previously described.¹⁴ The assessor of the PET examinations was blinded to treatment allocation and group.

Blood samples were drawn before and during the examinations. Glucose, electrolytes, lactate, and pH were analyzed using YSI STAT 2100 (YSI Inc., Netherlands) immediately after sampling. All other samples were stored at −20°C and analyzed in batch. Plasma (P)-3-OHB was measured using hydrophilic interaction liquid chromatography tandem mass spectrometry. Insulin was measured with ELISA (Mercodia, Sweden) and free fatty acids (FFA) quantified by in vitro enzymatic colorimetric method assay (Trichem, Denmark). Brain natriuretic peptide (BNP) was analyzed by chemiluminescent microparticle immunoassay on an Abbott Architect i2000SR (Abbott, Germany), as described by the manufacturer, and catecholamines (adrenaline and noradrenaline) by an in-house time-resolved high-performance liquid chromatography method. High sensitive cardiac troponin T was analyzed by an immunoassay kit according to the manufacture (Cobas; Roche Diagnostics GmbH, Germany).

The primary end point for Studies 1 and 2 was CO measured by thermodilution technique. Secondary end points were SV, SVO₂, PCWP, PVR, SVR, HR, MAP, mPAP, CVP, LVEF, GLS, catecholamines, P-potassium, pH, and P-lactate levels.

In Studies 3 and 4, the primary end point was MEE by 11C-acetate PET. Secondary end points were MVO₂, MBF, CO and SV, catecholamines, P-potassium, pH, and P-lactate levels.

In all studies, the end point measurements were compared at the end of each intervention period.

The coefficient of variation for CO is 4%.¹⁵ Hence, we expected to be able to detect relative differences of 3% (Study 1, N=16) and 4% (Study 2, N=8) with a power of 80% and an alpha of 5%. At our institution, coefficient of variation for MEE assessed by 11C-acetate PET is approximately 9%.¹³ Therefore, we expected to be able to detect relative differences of 9% in each study group and of 6% for both groups combined (Studies 3 and 4).

Data were investigated for normal distribution by qq-plots and transformed when appropriate. Unpaired t tests were used for comparison between groups (baseline characteristics). Changes in data are presented as mean±SEM and between-group differences as mean±SD, if not stated otherwise. When appropriate, 2-way repeated measures ANOVA (Study 1: terms were intervention and intervention sequence), 1-way repeated measures ANOVA (Study 2), and 3-way repeated measures ANOVA tests were applied (Studies 3 and 4: terms were intervention, group [age-matched versus HFrEF], and intervention sequence). Pairwise comparisons were performed if the overall ANOVA analysis demonstrated significant difference. P values <0.05 were considered statistically significant. All data analyses were performed using STATA version 14 (StataCorp, USA).

All patients gave informed consent, and the local Ethics Committee of the Central Denmark Region and the Danish Data Protection Agency approved the study. The study is registered at clinicaltrials.gov (NCT03073356).

Thirty-three stable chronic HFrEF patients were screened. Two withdrew consent and 7 had LVEF >40%; hence, 24 were enrolled. Twelve age-matched volunteers were screened, of whom 1 withdrew consent and another was excluded because of asymptomatic left bundle-branch block (Figure I in the online-only Data Supplement).

HFrEF patients enrolled in Study 1 had lower body mass index and higher NT-proBNP levels than those enrolled in Study 2, but were otherwise comparable on baseline variables. As scheduled, the age-matched volunteers enrolled in Study 4 had normal LVEF and received no treatment for HFrEF, as opposed to those enrolled in Study 3 (Table 1). All participants were white.

Compared with placebo, 3-OHB infusion (0.18 g · kg^-1 · h^-1) increased circulating P-3-OHB levels from 0.4±0.3 mM to 3.3±0.4 mM after 3 hours (P<0.001; Table 2, Figure 2A). The increment was associated with an increase in CO of 2.0±0.3 L/min (P<0.001) ie, approximately 40% (Figure 2B, Figure II in the online-only Data Supplement), and with an increase in SVO₂ from 72±3 to 79±4% (P<0.001; Table 3). The changes in CO were caused by increases in SV of 20±2 mL (P<0.001) and HR of 7±2 bpm (P<0.001), however, differences in CO between study arms were already present within the first 30 minutes (0.9 L/min, P=0.01) and preceded the changes in HR within the same period (P=0.99). We observed no significant changes in MAP or mPAP, but a minor decrease in CVP and PCWP (approximately 1 mm Hg; P=0.05 and P=0.03, respectively). Thus, SVR decreased by 30% (P<0.001) and PVR by 21% (P=0.003) (Figure III in the online-only Data Supplement). The administered volumes did not differ between study arms (Table 2). However, the increase in CO was most pronounced in those who were randomized to placebo followed by 3-OHB infusion (interaction between randomization sequence and treatment P=0.03).

Infusion of 3-OHB increased systolic left ventricular echocardiographic variables: left ventricular outflow track velocity time integral (4.0±0.6 cm; P<0.001), GLS (2.0±0.3% [absolute improvement]; P<0.001), S’max (0.79±0.22 cm/sec; P=0.003) and LVEF (8±1% [absolute increase]; P<0.001) compared with placebo (Table 3). We observed no association between LVEF at baseline and changes in either CO (P=0.72) or SV (P=0.49). There were no changes in end diastolic volumes (P=0.09), whereas end systolic volumes decreased significantly by 17±3 mL (P<0.001). The increase in GLS was most pronounced in those who were randomized to receive placebo followed by 3-OHB (P<0.001), whereas there was no interaction between randomization sequence and treatment in other echocardiographic measures (Table 3).

In the dose-response study, circulating P-3-OHB increased from <0.1 mM to 0.7±0.1 mM (P<0.001) at an infusion rate of 0.045 g · kg^-1 · h^-1, and to 1.6±0.3 mM at an infusion rate of 0.09 g · kg^-1 · h^-1, P-3-OHB (P<0.01; Table 2). The increase in P-3-OHB was associated with a significant, gradual increase in CO (Table 3, Figure IV in the online-only Data Supplement). The maximal increase in CO (1.2±0.1 L/min; P<0.001) was lower than the increase from placebo to 3-OHB infusion in Study 1 (P=0.007), where a higher infusion rate was applied (0.18 g · kg^-1 · min^-1) (Figure 2C). The changes in regard to SV, HR and SVO₂ were similar to those observed in Study 1, whereas CVP increased approximately 1 mm Hg (P=0.02).

In line with Study 1, we observed a gradual increase in left ventricular outflow track velocity time integral, GLS, LVEF and S’max during 3-OHB infusion. The increments were not statistically significant until the infusion rate was increased to 0.09 g · kg^-1 · h^-1 (Table 3).

Circulating P-3-OHB concentrations increased from 0.4±0.3 to 3.4±0.6 mM (P<0.001; Table I in the online-only Data Supplement) after infusion of 3-OHB, with no difference between HFrEF patients and control subjects (test for interaction P=0.43). There was no difference between the administered 3-OHB and placebo volumes, whereas slightly more glucose was administered during 3-OHB infusion (P=0.04; Table I in the online-only Data Supplement). As opposed to Study 1, there was no interaction between infusion sequence and CO (P=0.40).

In Study 3 (HFrEF patients), 3-OHB infusion increased LV external work by 38% (P=0.002) because of an increase in SV of 20±3 mL (P=0.001) and an increase in HR of 8±2 bpm (P=0.003) compared with placebo (Figure V in the online-only Data Supplement). We observed a proportional increase in MVO₂. Hence, MEE did not change (P=0.87; Table 4, Figure 3).

MEE was higher in age-matched volunteers (Study 4) than in HFrEF patients (22±5% versus 14±2%; P<0.001). As in HFrEF patients, CO, SV, left ventricular external work, and MVO₂ increased during 3-OHB compared with placebo infusion in age-matched subjects. Consequently, MEE was unaffected by 3-OHB infusion (P=0.21; Table 4, Figure 3), and we observed no difference between HFrEF patients and age-matched volunteers (P=0.42, test for interaction). MBF increased in both study groups during 3-OHB infusion. The increment was slightly higher in age-matched volunteers than in HFrEF patients (P=0.03, test for interaction). Infusion sequence did not interact with MEE (P=0.11) or any other PET-derived findings (Table 4).

Relative changes in key measurements for Study 1 and MEE for Studies 3 and 4 are summarized in Figure 4. Examples of time-activity curves during the PET examinations are demonstrated in Figure VI in the online-only Data Supplement.

In Study 1, circulating P-lactate concentrations and pH increased, whereas P-potassium concentrations decreased during 3-OHB infusion despite potassium substitution (Table 2, Figure VII in the online-only Data Supplement). In Study 2, pH increased slightly during 3-OHB infusion, whereas P-lactate and P-potassium concentrations were stable (Table 2, Figure VIII in the online-only Data Supplement). In the PET studies, the observed changes were similar to those observed in Study 1 (Table I and Figure IX in the online-only Data Supplement). Circulating BNP, high sensitive cardiac troponinT, noradrenalin, and insulin levels did not change during 3-OHB infusion, whereas FFA levels were slightly lower during 3-OHB than during placebo infusion (Table 2, Table I in the online-only Data Supplement). Adrenaline decreased in the healthy volunteers during 3-OHB infusion (P=0.003). In Studies 3 and 4, insulin sensitivity assessed by glucose infusion rate during the last 30 minutes of each infusion periods did not differ between 3-OHB and placebo infusion (P=0.11, Figure IXb and Table I in the online-only Data Supplement). Glucose infusion rate was not analyzed in Studies 1 and 2, as additional glucose was infused to measure CO by thermodilution. For those parameters affected by treatment in Study 1, there was an interaction between infusion sequence and pH and FFA levels (Table 2, Figures VIIB and VIIF in the online-only Data Supplement). Similarly, for those parameters affected by treatment in Studies 3 and 4, infusion sequence interacted with pH, potassium, and FFA levels (Table I and Figures IXC, IXG, and IXH in the online-only Data Supplement).

All participants completed the study program. One HFrEF patient (LVEF: 23%, NYHA 3) experienced asymptomatic, sustained ventricular tachycardia episodes (frequency: 100 bpm) during both study arms. These episodes occurred 2 hours after 3-OHB infusion had been initiated and continued for approximately 1 hour into the placebo infusion arm. The patient was known to have such episodes. No other arrhythmias were recorded. Another HFrEF patient (randomized to placebo followed by 3-OHB) reported a brief episode of nausea during the placebo infusion and another volunteer had an episode of nausea at the very end of the study after the PET examination and termination of the 3-OHB–infusion.

In early experimental studies, CO decreased in the isolated perfused rat heart at high concentrations of 3-OHB when ketone bodies were administered as the sole energy substrate.¹⁶ Adding glucose to the perfusate reversed this detrimental effect. Coadministration of 3-OHB (4 mM) and glucose actually increased cardiac work¹⁷ and output compared with glucose alone in an isolated rat heart model.¹⁸ Knock-out of the gene responsible for control of flux through the enzymes of cardiac ketone metabolism resulted in pathological remodeling and cardiac dysfunction in the failing heart,¹⁹ suggesting a pivotal role for sustained ketone oxidation in HF.²⁰ Moreover, increased myocardial ketone utilization has been demonstrated in explanted hearts from patients with severe HF.⁵

The observed improvement of cardiac function was not caused by increments in preload, as PCWP decreased and BNP remained unchanged. Lactate levels increased with the rise in ketone levels, presumably because of energy substrate competition with increased lactate production from pyruvate,^21–23 in combination with reduced lactate oxidation in peripheral tissues.²⁴ The changes in circulating lactate levels were minor (1.1 versus 1.5 mM) and well below the levels between 2.4 to 4.8 mM that affects hemodynamics.²⁵

The observed increase in CO and LVEF may partly be caused by reduced SVR. In fact, other studies are supportive of vasodilatory effects of ketones.²⁶ However, in the present study, blood pressure did not change definitively, despite a decline in SVR of 30%. In addition, we observed no major change in preload during 3-OHB infusion. Consequently, the Frank–Starling relationship shifted upward because of vasodilatation, increased contractility, or a combination thereof. Invasive pressure–volume measurements were not performed since the main scope of the present study was to investigate the clinical, cardiovascular and hemodynamic effects of 3-OHB. Therefore, we cannot definitively determine whether 3-OHB increases CO by a direct effect on the myocardium, by vasodilation or by a combination thereof. Since 3-OHB interacts with G-protein coupled receptors there is a plausible mechanism for peripheral, vascular effects. Finally, 3-OHB may also work as an agonist for cellular receptors known to increase HR,²⁷ which could be involved in the observed increase in HR.

Irrespective of the mechanism, our data demonstrate that 3-OHB has beneficial cardiovascular effects in HF as it improves cardiac function and hemodynamics, causes vasodilatation, and thus, improves organ perfusion.

In healthy subjects, circulating P-3-OHB levels range from 0.1 to 0.3 mM following an overnight fast. Levels increase to 2 to 3 mM after 48 to 72 hours of fasting and may reach 5 to 10 mM after prolonged fasting.^28,29 Thus, in the dose-response study, circulating P-3-OHB concentrations were in the physiological range from 0.7 to 1.6 mM, and we observed a dose-dependent cardiovascular response to 3-OHB infusion (Table 3 and Figure 2C). At the lowest infusion rate of 0.045 g · kg^-1 · min^-1, P-3-OHB increased from 0.3 mM at baseline to 0.7 mM, and this increase was associated with significant increases in CO and SVO₂. The hemodynamic response increased at higher infusion rates and concentrations. In HFrEF patients, circulating 3-OHB concentrations are increased in proportion to the severity of cardiac dysfunction.^30,31 This may be caused partly by increased FFA mobilization in response to stimulation by stress hormones and insulin resistance.³¹ Our observations suggest that increased levels of ketones bodies may be adaptive and act as endogenous mediators of cardiovascular function in HFrEF patients. This finding may be relevant in the context of sodium-glucose cotransporter 2 inhibitors. These glucose-lowering drugs increase ketogenesis and circulating P-3-OHB concentrations to approximately 0.6 mM.³² Hence, it is conceivable that the associated reduction in HF hospitalizations during sodium-glucose cotransporter 2 treatment⁷ is partly attributed to the properties of 3-OHB, as demonstrated in the current study. Thus, the current study demonstrates that the hemodynamic response to 3-OHB infusion is dose-dependent, observed in the physiological concentration range, and can be adjusted by changing the infusion rate (Figure 2C).

Reduced MEE is believed to be causally involved in the development of HF³³ and is a prognostic factor in HF patients.³⁴ Ketone bodies improve myocardial efficiency of the working rat heart by 25%.¹⁸ However, the results of animal experiments are inconsistent,^5,16,18 and the relationship between myocardial ketone body oxidation, cardiac work, and myocardial efficiency in patients with HF remains undetermined.^4,6,35 We observed that 3-OHB infusion increased MVO₂ and cardiac work proportionally, leaving MEE unaltered. Thus, short-term treatment with 3-OHB did not impair the coupling between myocardial energy generation and cardiac work although hemodynamics were improved. This is clinically important as reduced MEE could worsen the prognosis.³⁴ Our experiments were conducted during a low-dose insulinemic euglycemic clamp, partly to mimic ordinary, fed insulin-stimulated conditions of every-day life and partly to suppress endogenous ketone production and impose clear-cut differences in 3-OHB levels between the two infusion periods. Thus, the findings cannot be extrapolated to a setting of prolonged fasting where FFA levels are higher and myocardial FFA uptake is increased;³⁶ under such conditions, a metabolic shift in substrate oxidation from FFA to 3-OHB oxidation could theoretically improve oxygen efficiency and MEE.

The effect of inotropic agents on MEE have been equivocal in previous studies.^37–39 The present study demonstrates that despite the hemodynamic effects of 3-OHB infusion, there was no detrimental effects on MEE. Consequently, 3-OHB improves CO and organ perfusion without MEE deterioration.

Like others,⁴⁰ we observed that MEE was approximately 50% lower in HFrEF patients than in age-matched volunteers (Figure 3). We have previously shown that in healthy subjects a short-term increase in circulating 3-OHB levels reduces myocardial glucose uptake by 50% without affecting FFA oxidation,⁴¹ suggesting that the healthy human heart rapidly increases 3-OHB utilization for mitochondrial ATP generation. We now extend our findings and demonstrate that infusion of 3-OHB increases CO and SV in subjects without heart disease. MBF was increased, most likely by autoregulatory mechanisms in the coronary vasculature, in response to increased external work and MVO₂.⁴² Both cardiac work and MVO₂ increased during 3-OHB infusion. Thus, MEE remained unchanged in both HFrEF patients and age-matched volunteers. In highly trained endurance athletes, ingestion of a ketone ester modestly increases exercise capacity.⁴³ Although this effect may have been caused by an increased muscular capacity to oxidize fatty-acid–derived carbon moieties, it remains possible that increased cardiac performance may have contributed.

Overall, the effects of modulation of circulating 3-OHB in age-matched controls did not differ from the effects observed in HFrEF patients. These findings suggest that the cardiovascular response to 3-OHB infusion is universal, and that the metabolic shift toward increased ketone metabolism⁵ during metabolic stress in evolutionary terms may be viewed as a beneficial species-preserving mechanism.

The changes in pH and potassium by 3-OHB infusion have previously been shown.^41,44 The dissolved Na-3-OHB dissociates and allows 3-OHB to act as a conjugated base, which increases pH.²⁸ In the present study, pH increased only marginally and unlikely influenced the results. Major changes in P-potassium were avoided by administering a potassium supplement. In future clinical trials, changes in pH and potassium levels should be taken into consideration.

HR increased by 7 bpm during 3-OHB infusion at a rate of 0.18 g · kg^-1 · min^-1. Since increased HR is associated with a worse prognosis in chronic HFrEF,⁴⁵ the increase might raise concerns for future use of 3-OHB in HFrEF patients. Importantly, the increase in HR was only 2 to 4 bpm at an infusion rate of 3-OHB 0.045 g · kg^-1 · h^-1 and 3-OHB 0.09g · kg^-1 · h^-1, corresponding to 3-OHB levels in the range of 0.7 to 1.6 mM. These findings may be useful for establishing the optimal dosage in clinical trials.

One episode of asymptomatic idioventricular arrhythmia was observed in a patient with advanced HFrEF. The patient was known to present with such episodes. Whether the arrhythmia was related to cellular hyperpolarization due to increased ketone metabolism,⁴⁶ placement of the Swan–Ganz catheter or merely a coincidence is unknown. Episodes of arrhythmias were not recorded in the remaining HFrEF patients.

Fluid load is of importance in HFrEF. In the present study, no patients had acute HFrEF, and a Na-3-OHB load for 3 hours seems to be acceptable as we observed no increase in PCWP, BNP, or high sensitive cardiac troponinT levels. However, the Na⁺ conjugated to the 3-OHB may have detrimental effects in patients with severe, acute HF. The used Na-3-OHB consisted of a racemic mixture of both L- and D-isomer in which only the D-isomer was metabolized. Hence, a pure D-isomer of 3-OHB conjugated with another low-molecular–weight combustible carbon moiety could decrease the infused volume and avoid Na⁺ load.^21,43

Studies 1, 3, and 4 were crossover studies which is a design with potential carry-over effect. Patients who received 3-OHB infusion initially had slightly elevated P-3-OHB levels at the end of the placebo period (Figure 2), as well as persisting elevated pH (Figure VIIF in the online-only Data Supplement). However, there was no interaction between infusion order and changes in SV and SVO₂ (Study 1), nor was there interaction between infusion order and changes in MEE and CO in the PET studies.

In Studies 1 and 2, the investigator performing the cardiac output assessment by thermodilution was not blinded to the randomization. Echocardiographic measurements were blinded during analyses and confirmed the invasive findings (Table 3). CO measurements by PET in Studies 3 and 4 also confirmed the increase in cardiac output (Table 4) during 3-OHB infusion. The investigator performing the PET analyses was not involved in the experiments and was blinded to randomization.

The decrease in filling pressures (CVP, PCWP) were small despite improvement in cardiac performance. This is in consistence with previous studies concerning dobutamine in stable HFrEF patients.⁴⁷ It is also conceivable that the volume loading given during the study outbalanced the decrease in filling pressures induced by vasodilation.

Insulin was coadministered during the interventions. However, it seems unlikely that insulin caused the observed effects as insulin levels did not differ when end point measurements were assessed in any of the studies (Table 2, Table I in the online-only Data Supplement). In terms of relative differences, FFA levels were approximately 60% lower during 3-OHB infusion. However, absolute changes in FFA levels were minimal because of suppression by the low-dose insulinemic euglycemic clamp (medians between 0.04 and 0.27 mM; Figures VIIB, VIIIC, and IXC in the online-only Data Supplement). Therefore, it seems unlikely that the small absolute differences in FFA levels had any impact on the hemodynamic findings.

3-OHB demonstrated dose-dependent beneficial cardiac and hemodynamic effects in HFrEF patients without deteriorating mechanoenergetic coupling and without causing any safety issues. We observed similar response to 3-OHB infusion in age-matched test subjects and HFrEF patients. Thus, modulation of circulating 3-OHB may represent a novel treatment principle in patients with HF.