Trimetazidine in heart failure with preserved ejection fraction: a randomized controlled cross‐over trial

Abstract

Aims

Impaired myocardial energy homeostasis plays an import role in the pathophysiology of heart failure with preserved ejection fraction (HFpEF). Left ventricular relaxation has a high energy demand, and left ventricular diastolic dysfunction has been related to impaired energy homeostasis. This study investigated whether trimetazidine, a fatty acid oxidation inhibitor, could improve myocardial energy homeostasis and consequently improve exercise haemodynamics in patients with HFpEF.

Methods and results

The DoPING‐HFpEF trial was a phase II single‐centre, double‐blind, placebo‐controlled, randomized cross‐over trial. Patients were randomized to trimetazidine treatment or placebo for 3 months and switched after a 2‐week wash‐out period. The primary endpoint was change in pulmonary capillary wedge pressure, measured with right heart catheterization at multiple stages of bicycling exercise. Secondary endpoint was change in myocardial phosphocreatine/adenosine triphosphate, an index of the myocardial energy status, measured with phosphorus‐31 magnetic resonance spectroscopy. The study included 25 patients (10/15 males/females; mean (standard deviation) age, 66 (10) years; body mass index, 29.8 (4.5) kg/m²); with the diagnosis of HFpEF confirmed with (exercise) right heart catheterization either before or during the trial. There was no effect of trimetazidine on the primary outcome pulmonary capillary wedge pressure at multiple levels of exercise (mean change 0 [95% confidence interval, 95% CI −2, 2] mmHg over multiple levels of exercise, P = 0.60). Myocardial phosphocreatine/adenosine triphosphate in the trimetazidine arm was similar to placebo (1.08 [0.76, 1.76] vs. 1.30 [0.95, 1.86], P = 0.08). There was no change by trimetazidine compared with placebo in the exploratory parameters: 6‐min walking distance (mean change of −6 [95% CI −18, 7] m vs. −5 [95% CI −22, 22] m, respectively, P = 0.93), N‐terminal pro‐B‐type natriuretic peptide (5 (−156, 166) ng/L vs. −13 (−172, 147) ng/L, P = 0.70), overall quality‐of‐life (KCCQ and EQ‐5D‐5L, P = 0.78 and P = 0.51, respectively), parameters for diastolic function measured with echocardiography and cardiac magnetic resonance, or metabolic parameters.

Conclusions

Trimetazidine did not improve myocardial energy homeostasis and did not improve exercise haemodynamics in patients with HFpEF.

Introduction

Therapeutic options for heart failure with preserved ejection fraction (HFpEF) are currently limited. ¹ Recently, mitochondrial function was identified as a potential therapeutic target. ² , ³ Several studies have demonstrated that intrinsic mitochondrial dysfunction and/or compromised myocardial energy homeostasis is common in HFpEF. ³ , ⁴ , ⁵ Diastole is an energy‐demanding process, stemming from cross‐bridge detachment and active transport of calcium ions back into the sarcoplasmic reticulum of the cardiomyocytes during diastole. ² Indeed, the high energy demand of left ventricular (LV) relaxation emerges as an apparent association of diastolic dysfunction with impaired myocardial energy metabolism. ² , ⁶

Trimetazidine, a drug currently used for treatment of angina, can improve a compromised myocardial energy homeostasis. ⁷ , ⁸ Trimetazidine's main mechanism of action is attributed to partial inhibition of fatty acid oxidation, shifting oxidative metabolism from fatty acid oxidation to glucose oxidation. ⁷ As a result, more units of adenosine triphosphate (ATP) can be produced per mole of oxygen, improving mitochondrial efficiency. ⁷ In addition, the increase of glucose oxidation leads to a reduction of (excessive) anaerobic glycolysis and consequently a reduction of cell acidosis. ⁹ Furthermore, small studies provide evidence that trimetazidine reduces inflammation, myocardial interstitial fibrosis, and the production of reactive oxygen species, while improving whole‐body insulin sensitivity and endothelial function. ⁷ , ⁹ , ¹⁰

In heart failure with reduced ejection fraction (HFrEF), trimetazidine was shown to improve myocardial energy homeostasis, and systolic and diastolic function. ⁸ , ⁹ , ¹¹ , ¹² , ¹³ It is therefore conceivable that a trimetazidine intervention may have beneficial effects on myocardial energy metabolism and diastolic function in HFpEF. Currently, the clinical effect of trimetazidine in HFpEF is unknown. Here, we investigated whether trimetazidine in HFpEF improves LV diastolic function and exercise haemodynamics in particular, by improving myocardial energy homeostasis, using exercise right heart catheterization and phosphorus‐31 MR spectroscopy measurements. ¹⁴

Methods

Study design and randomization

The DoPING‐HFpEF trial is a phase II single‐centre, double‐blind, placebo‐controlled, cross‐over trial. Details on study design have been described previously ¹⁴ and can also be found in the supporting information. In short, the study consisted of two treatment periods of 3 months separated by a 2‐week wash‐out period (Figure 1 ). Patients were randomized 2:1 to either placebo‐first or trimetazidine‐first to minimize potential carry‐over effects. Randomization was carried out in block sizes of 6 and 3, respectively, using the web‐based application Castor EDC. The pharmacists received the randomization status via the web‐based application and dispensed the drug. Both the drug (trimetazidine or placebo), and the casings were indistinguishable from each other. Patients, physicians, and study personnel were blinded, with exception of the study pharmacist for safety reasons.

Figure 1.

DoPING‐HFpEF study design. 6‐MWD, six‐minute walking distance; Echo, echocardiogram; Lab, laboratory assays; MRI, magnetic resonance imaging; MRS, magnetic resonance spectroscopy; QoL, quality of life; RHC, right heart catheterization.

Participants

The key inclusion criteria were the diagnosis of clinically stable HFpEF with New York Heart Association (NYHA) functional class II or higher, despite optimal medical treatment (including complete revascularization and cardiac rehabilitation). HFpEF was diagnosed based on symptoms of heart failure, LV ejection fraction ≥50% and evidence of LV diastolic dysfunction, based on either elevated filling pressures at rest or exercise during right heart catheterization, or the combination of diastolic dysfunction grade II on echocardiography with a N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP) >125 pg/mL. A complete list of in‐ and exclusion criteria can be found in Table S1 . Patients were recruited from the Amsterdam University Medical Centers outpatient dyspnoea/HFpEF clinic and by referral from satellite hospitals.

Intervention

Patients were treated with either trimetazidine 20 mg or placebo (microcrystalline cellulose powder) three times per day, or twice per day if eGFR 30–60 mL/min/1.73 m², similar to other studies. ⁹ Participants were instructed not to change their diet and exercise regimens and minimize changes of other medications. Compliance was checked by pill count at every visit. In addition, trimetazidine plasma levels were assessed after 3 months of treatment.

Study procedures and endpoints

Right heart catheterization with exercise, ¹⁵ MR examinations, ¹⁶ and echocardiography were performed in an outpatient clinic setting at the end of each treatment period (Figure 1 ). Additional tests (6‐min walking distance [6‐MWD], laboratory assays, Kansas City Cardiomyopathy Questionnaire [KCCQ], and the EQ‐5D‐5L questionnaire) were conducted at the start and end of each treatment period. The primary endpoint was change in pulmonary capillary wedge pressures (PCWPs) measured at multiple intensities of exercise between the placebo and trimetazidine treatment arms. Exercise was performed supine during right heart catheterization on a bicycle ergometer. Participants were required to cycle at 60–65 rotations per minute, starting at 20 W resistance, increasing resistance with 20 W every 3 min. PCWP and other haemodynamic parameters were measured at rest and at the end of every exercise stage. The secondary endpoint was the change in the myocardial phosphocreatine (PCr) to ATP ratio (PCr/ATP) as a non‐invasive measure of the myocardial energy status, ⁵ , ⁶ , ⁸ quantified with single‐voxel 3D image‐selected in vivo spectroscopy (3D‐ISIS) ³¹P‐MR spectroscopy of the LV at 3 Tesla. ¹⁶ All analyses were performed by analysts blinded for randomization status. Study procedures are detailed in supporting information.

Statistical analyses

Normally distributed data are presented as mean (standard deviation), and non‐normal distributions are presented as median [interquartile range]. Data with a non‐normal distribution were log‐transformed (if applicable) before performing statistical analyses. If applicable, values were indexed to baseline body surface area (BSA) using the Haycock formula.

Continuous variables at the end of both study periods were compared using two‐sided paired t‐tests or Wilcoxon signed‐rank tests, as appropriate. The primary endpoint and other pressures measured at multiple levels (20–100 W) of exercise were compared using mixed‐model analyses of repeated measures at subject level. We did not remove any of the additional measurements in those cases where a subject achieved a higher workload in one of the study arms (Table  S4 ), but incorporated them in the mixed‐model analyses with adjustment for the level of exercise that were captured as a fixed effect. Data with measurements at start and end of both treatment periods (e.g., laboratory data, QoL questionnaires, and 6‐MWD) were compared using ANCOVA. The EQ‐5D‐5L questionnaire was indexed for the preferences of the Dutch population using the established Dutch Tariff (Model 3: Tobit with constraints). ¹⁷ Any carry‐over effect in primary and secondary outcomes, and 6‐MWD was tested by addition of the allocation order to the mixed linear model. Exploratory subgroup analysis of the primary endpoint was performed based on the change in PCWP (averaged over all performed levels of exercise) and visualized with a forest plot.

With an effect size set to a mean PCWP change of −3.2 (5.0) mmHg at exercise, the predefined sample size was 18 participants as previously described. ¹⁴ Due to uncertainty of the effect size and to mitigate potentially poor data quality for some parameters, sample size was raised to 25 participants. Drop‐outs were replaced to reach the required sample size. In HFrEF, trimetazidine increased myocardial PCr/ATP with 0.45 (0.6), that is, 34%. ⁸ Power calculation (two‐sided, α = 0.05, 1‐β = 0.80) predicted a required sample size of n = 17 to detect a similar change in myocardial PCr/ATP as the secondary endpoint of this study. Given the heterogeneity of the HFpEF population and to fully utilize our cross‐over design, only patients without loss to follow‐up were included in the analyses, except for the evaluation of (serious) adverse events.

IBM SPSS 28 was used for all statistical analyses. Missing data were not imputed. Graphs were made with RStudio 3.6.1 and GraphPad Prism 9 (Dotmatics, Boston MA).

Ethical approval

The investigation conforms with the principles outlined in the Declaration of Helsinki and was approved by the Medical Ethics Review Committee of the VU University Medical Center, Amsterdam, The Netherlands (NL66242.029.18). All participants provided written informed consent. This study is registered in the European Union Clinical Trials Register (2018–002170‐52) and the Netherlands Trial Register (NL7830). There were no major amendments during the duration of the study.

Results

Baseline characteristics

Between May 2019 and February 2021, 261 patients were screened, 30 patients were included and randomized, of whom 25 patients completed the trial (Figure 1 , Figure 2  and Table S2 ). Drop‐outs are discussed below. Inclusion was closed once the pre‐defined sample size of 25 participants was reached. The last patient's last follow‐up visit was November 2021. Patients (n = 25; male/female, 10/15) had a mean age of 66 (10) years and a body mass index of 29.8 (4.5) kg/m². Baseline characteristics are reported in Tables 1 and S3 .

Figure 2.

Flow diagram of the DoPING‐HFpEF Trial. *Reasons for exclusion are listed in Table S2 . †Reasons for exclusion from analyses are discussed in the supporting information.

Table 1.

Baseline characteristics of patients that completed the trial

	Placebo‐first (n = 17)	Trimetazidine‐first (n = 8)	All patients (n = 25)
Sex (female), n (%)	10 (59%)	5 (63%)	15 (60%)
Age, years	67 (10)	64 (11)	66 (10)
Caucasian ethnicity	17 (100%)	8 (100%)	25 (100%)
BMI, kg/m2	29.2 (3.0)	30.9 (6.2)	29.8 (4.2)
BSA, m2	2.06 (0.20)	2.18 (0.37)	2.10 (0.27)
Waist‐to‐hip ratio
Female	0.90 (0.09)	0.81 (0.03)	0.87 (0.09)
Male	0.97 (0.12)	1.03 (0.04)	0.99 (0.11)
NYHA class, n (%)
II	10 (59%)	5 (63%)	15 (60%)
III	7 (41%)	3 (38%)	10 (40%)
Systolic BP, mmHg	129 (17)	125 (18)	128 (17)
Diastolic BP, mmHg	75 (10)	70 (12)	74 (11)
Heart rate, beats/min	68 (10)	62 (13)	66 (12)
Haemoglobin, mmol/L	8.5 (0.8)	8.5 (0.9)	8.5 (0.8)
NT‐proBNP, pg/mL	135 [80, 609]	89 [58, 342]	95 [72, 490]
LVEF at baseline, %	57 (4)	61 (4)	58 (5)
H2FPEF score*	5 (3)	4 (3)	4 (3)
Comorbidities, n (%)
Hypertension	13 (77%)	7 (88%)	20 (80%)
(Pre) type II diabetes mellitus†	5 (29%)	4 (50%)	9 (36%)
Atrial fibrillation
Paroxysmal	4 (24%)	0 (0%)	4 (16%)
Permanent	3 (18%)	1 (13%)	4 (16%)
Renal dysfunction‡	5 (29%)	1 (13%)	6 (24%)
Hypercholesterolemia	8 (47%)	3 (38%)	11 (37%)
History of coronary artery disease	3 (18%)	3 (38%)	6 (24%)
COPD	3 (18%)	0 (0%)	3 (12%)
Obesity	6 (35%)	5 (63%)	11 (44%)
OSAS	4 (24%)	3 (38%)	7 (28%)
Medication
Diuretics	12 (71%)	7 (88%)	19 (76%)
ACE‐I/ARB	10 (59%)	6 (75%)	16 (64%)
Beta‐blocker	5 (29%)	5 (63%)	10 (40%)
Calcium antagonist	10 (59%)	4 (50%)	14 (56%)
Digoxin	1 (6%)	0 (0%)	1 (4%)
Statins	8 (47%)	5 (63%)	13 (54%)
Anti‐diabetic medication	2 (12%)	1 (13%)	3 (12%)

Data are presented as mean (SD), median [interquartile range], or n (%), as appropriate.

^*H₂FPEF score diagnostic and prognostic tool for HFpEF.

^†Pre‐type II diabetes mellitus was defined as the presence of impaired fasting glucose (≥6.1 mmol/L).

^‡Defined by an estimated glomerular filtration rate <60 mL/min/1.73 m².

ACE‐I, angiotensin‐converting enzyme inhibitors; ARB, angiotensin receptor blocker; BMI, body mass index; BSA, body surface area; COPD, chronic obstructive pulmonary disease; LVEF, left ventricular ejection fraction; NT‐proBNP, N‐terminal‐pro‐brain natriuretic peptide; NYHA, New York Heart Association; OSAS, obstructive sleep apnoea syndrome.

Compliance

Compliance to treatment (pill count) was 98% [97, 100] in the placebo arm and 100% [97, 100] in the trimetazidine arm. The mean plasma trimetazidine concentration was 50.7 (14.5) ng/mL in samples obtained 1.8 [1.4, 2.8] hours after morning intake. Trimetazidine could not be detected in plasma of any of the patients in the placebo arm nor after the wash‐out period.

Right heart catheterization and exercise pulmonary capillary wedge pressures

There was no effect of trimetazidine treatment on the primary outcome PCWP at (multiple levels of) exercise (P = 0.60; Figure 3A ), with a change in PCWP of 0 (95% CI −2, 2) mmHg over multiple levels of exercise. In addition, we did not detect an effect on PCWP at rest, during passive leg raise, or after recovery (P = 0.63, P = 0.85, and P = 0.53, respectively). Exploratory analyses established that the absence of effect was consistent over sex, LV mass index, NYHA class, beta‐blocker usage, and the absence or presence of co‐morbidities: atrial fibrillation, kidney dysfunction, obesity, and (pre) type 2 diabetes mellitus (Figure 3B ). There was no carry‐over effect (P = 0.38). Individual PCWP data are presented in Figure S1 , showing small intra‐individual variation. Furthermore, pressures in the right atrium, right ventricle, or pulmonary artery at rest (Table  2 ), or during exercise (Table  S4 ) were similar with placebo or trimetazidine. In addition, there were no differences between placebo and trimetazidine in cardiac index, PCWP/cardiac output slope or in mixed venous oxygen saturation at rest or during exercise (Table  2 ).

Figure 3.

Effect on pulmonary capillary wedge pressure at multiple levels of exercise. (A) PCWP at multiple levels of exercise (primary endpoint). Point and error bar depict the mean and standard error of the mean. (B) Forest plot with subgroup analyses of change in PCWP at multiple levels of exercise. *eGFR <60 mL/min/1.73 m² during the trimetazidine treatment period. †LV hypertrophy was defined as LVMI ≥55 g/m² for women and ≥72 g/m² for men. LV, left ventricular, LVMI, left ventricular mass index; NYHA, New York Heart Association class; PCWP, pulmonary capillary wedge pressure; PLR, passive leg raise.

Table 2.

Haemodynamic parameters assessed by right heart catheterization

Parameter	Placebo	Trimetazidine	Mean difference (95% CI)	P‐value	n
Heart rate, b.p.m.	73 (12)	70 (11)	−2 (−5, 0)	0.10	25
Mean RAP, mmHg	8 (3)	7 (3)	‐0 (−1, 1)	0.74	24
Systolic RVP, mmHg	34 [30, 43]	35 [32, 42]	1 (−2, 5)	0.63	22
Diastolic RVP, mmHg	9 (4)	10 (4)	1 (−1, 3)	0.42	18
Systolic PAP, mmHg	33 [28, 43]	32 [28, 43]	1 (−2, 4)	0.89	25
Diastolic PAP, mmHg	14 [12, 18]	15 [11, 20]	‐0 (−2, 1)	0.78	25
Mean PAP, mmHg	22 [20, 28]	21 [19, 30]	‐0 (−2, 2)	0.57	25
PCWP rest, mmHg	14 (4)	14 (5)	0 (−1, 2)	0.63	25
V‐wave, mmHg	18 (7)	20 (9)	2 (−1, 4)	0.23	24
SVR, dynes/s/cm5	1160 (369)	1158 (410)	‐2 (−130, 127)	0.98	25
PVR, dynes/s/cm5	96 [76, 179]	128 [62, 165]	4 (−18, 26)	0.65	25
Other parameters
Mixed venous oxygen saturation, %
Rest	71 (5)	70 (5)	−1 (−2, 1)	0.32	25
Submaximal exercise	53 (9)	54 (12)	0 (−4, 4)	0.85	17
Peak exercise	39 (11)	38 (10)	−1 (−3, 1)	0.39	24
Cardiac index, mL/min/m2
Rest	3.2 (0.8)	3.1 (0.7)	−0.1 (−0.3, 0.2)	0.61	25
Submaximal exercise	4.6 (1.1)	4.4 (1.1)	−0.1 (−0.5, 0.2)	0.40	20
PCWP/cardiac output slope at 20 W exercise
ΔPCWP/ΔCardiac Output, mmHg/L/min	3.5 (2.0)	3.0 (1.7)	−0.5 (−1.3, 0.3)	0.22	19

Data are presented as mean (SD) or median [interquartile range], as appropriate. Submaximal exercise at 20 W.

BP, blood pressure; CI, confidence interval; PAP, pulmonary artery pressure; RAP, right artery pressure; RVP, right ventricular pressure; SVR, systemic vascular resistance; PVR, pulmonary vascular resistance.

Phosphorus‐31 magnetic resonance spectroscopy and myocardial energy status

Median myocardial PCr/ATP after placebo was 1.30 [0.95, 1.86]. Trimetazidine did not increase PCr/ATP (1.08 [0.76, 1.76], P = 0.08; Figure 4 , Figure S2 , and Table 3 ), indicating that the myocardial energy status had not improved. This was consistent for patients with or without LV hypertrophy, diabetes, obesity, or renal dysfunction. There was no carry‐over effect (P = 0.24). Furthermore, there was no change in myocardial total creatine content (P = 0.30) or myocardial triglyceride levels (P = 0.26) as measured with localized proton MR spectroscopy (Table 3 and Figure S3 ).

Figure 4.

Mean phosphorus‐31 MR spectra of the heart during placebo and trimetazidine treatment. Localized ³¹P‐MR spectroscopy at 3 Tesla. Placebo group mean spectrum (blue) and trimetazidine group mean spectrum (red) are overlaid on individual spectra in grey (n = 25). Myocardial PCr/ATP was quantified as the ratio of the PCr signal amplitude at 0.00 ppm over the γ‐ATP signal amplitude at −2.48 ppm, corrected for partial saturation. α‐, β‐ and γ‐ATP, α‐, β‐ and γ‐phosphate groups in adenosine triphosphate; MR, magnetic resonance; PCr, phosphocreatine; P_i, inorganic phosphate; 2,3‐DPG, 2,3‐diphosphoglycerate.

Table 3.

Magnetic resonance and echocardiography measurements

Parameter	Placebo	Trimetazidine	Mean difference (95% CI)	P‐value	n
Magnetic resonance at 3 Tesla
MR spectroscopy
PCr/ATP, −	1.30 [0.95, 1.86]	1.08 [0.76, 1.76]	−0.15 (−0.41, 0.11)	0.08	25
Myocardial total creatine, % of total water signal	0.09 [0.08, 0.11]	0.10 [0.08, 0.12]	0.01 (−0.01, 0.03)	0.30	22
Myocardial triglyceride, % of total water signal	0.56 [0.36, 1.07]	0.76 [0.49, 1.10]	0.07 (−0.14, 0.28)	0.26	22
Left ventricle
LVEDVi, mL/m2	72.2 (19.1)	69.5 (18.4)	−2.8 (−5.7, 0.2)	0.06	25
LVESVi, mL/m2	28.9 (10.7)	26.3 (9.9)	−2.6 (−4.9, −0.3)	0.03	25
LVSVi, mL/m2	43.4 (11.0)	43.2 (10.9)	−0.2 (−2.3, 2.0)	0.86	25
LVEF, %	60.5 (6.9)	62.7 (6.8)	2.2 (−0.3, 4.8)	0.09	25
Cardiac index, mL/min/m2	2.8 (0.6)	2.8 (0.6)	0.0 (−0.2, 0.2)	0.96	25
LV GLS, %	−15.0 (2.9)	−15.5 (1.7)	−0.4 (−1.7, 0.9)	0.26	21
Peak torsion, rad	−0.19 (0.06)	−0.17 (0.06)	0.01 (−0.00, 0.03)	0.07	23
Endocardial circumferential strain, %	−28.0 (7.8)	−26.7 (8.0)	1.3 (−0.7, 3.3)	0.21	23
Torsion/shortening ratio, rad	0.63 (0.18)	0.60 (0.21)	−0.03 (−0.09, 0.04)	0.39	22
LV mass index, g/m2	55.6 (15.3)	55.9 (15.2)	0.2 (−1.9, 2.4)	0.83	25
Myocardial native T1 (ms)	1264 (59)	1259 (38)	−6 (−34, 23)	0.69	23
Myocardial T2 (ms)	48 (7)	46 (4)	−2 (−5, 1)	0.22	20
Right ventricle
RVEDVi, mL/m2	75.1 (21.4)	73.4 (19.8)	−1.7 (−4.6, 1.1)	0.22	24
RVESVi, mL/m2	32.3 [26.8, 45.4]	34.1 [27.2, 46.7]	−0.8 (−3.8, 2.2)	0.59	24
RVSVi, mL/m2	38.5 (10.1)	37.6 (10.9)	−0.9 (−3.8, 1.9)	0.45	24
RVEF, %	54.6 (6.8)	53.8 (8.6)	−0.8 (−3.6, 2.0)	0.58	24
RV GLS, %	−22.6 (3.8).	−21.4 (3.4)	1.3 (0.0, 2.5)	0.048	23
Atria
LAVimin, mL/m2	21.2 [18.7, 41.2]	21.0 [15.6, 37.0]	0.2 (−2.8, 3.3)	0.42	24
LAVimax, mL/m2	47.4 (16.0)	46.3 (19.9)	−1.0 (−4.5, 2.5)	0.55	24
LA strain, % (excl. AF)	−20.3 (4.1)	−20.4 (4.0)	−0.1 (−2.5, 2.3)	0.93	18
RAVimin, mL/m2	22.9 [19.8, 34.2]	25.6 [21.6, 35.3]	−0.3 (−5.0, 4.4)	0.81	23
RAVimax, mL/m2	39.7 [36.2, 55.1]	38.1 [30.5, 54.7]	−1.0 (−6.4, 4.4)	0.93	24
Echocardiography
E/A	0.99 (0.35)	0.98 (0.38)	−0.0 (−0.1, 0.1)	0.89	18
e′ septal, cm/s	7.7 (1.8)	7.4 (1.5)	−0.2 (−0.9, 0.4)	0.45	23
E/e′ septal	10.0 (3.0)	10.3 (3.4)	0.3 (−0.4, 1.0)	0.38	22
DT, ms	206 (56)	218 (62)	12 (−3.4, 27)	0.12	23
IVRT, ms	79 (25)	84 (12)	5 (−11, 22)	0.52	15
TRV, cm/s	288 (66)	297 (74)	9 (−12, 29)	0.35	9
PV S/D	1.40 (0.26)	1.30 (0.27)	−0.10 (−0.25, 0.05)	0.17	14
TAPSE, mm	2.19 (0.46)	2.28 (0.46)	0.09 (−0.4, 0.23)	0.17	21
RV S′, cm/s	11.9 (3.0)	12.3 (2.8)	0.4 (−0.9, 1.6)	0.56	20

Data are presented as mean (SD) or median [interquartile range], as appropriate.

A, late diastolic transmitral flow velocity; AF, atrial fibrillation; ATP, adenosine triphosphate; CI, confidence interval; DT, mitral flow deceleration time; E, early diastolic transmitral flow velocity; e′, early diastolic mitral annular tissue velocity; EDVi, indexed end diastolic volume; EF, ejection fraction; ESVi, indexed end systolic volume; GLS, global longitudinal strain; IVRT, isovolumetric relaxation time; LAVi_max, indexed maximal left atrial volume; LAVi_min, indexed minimal left atrial volume; LV, left ventricular; PCr, phosphocreatine; PV S/D, ratio between early systolic and early diastolic pulmonary vein velocity; RAVi_max, indexed maximal right atrial volume; RAVi_min, indexed minimal right atrial volume; RV, right ventricular; SVi, indexed stroke volume. TRV, tricuspid valve regurgitation velocity.

Exploratory outcomes

Structural and functional parameters as assessed by echocardiography and MRI

Diastolic function as assessed by echocardiography did not differ between placebo or trimetazidine (Table  3 ). Likewise, there were no differences in LA volume, LA strain, LV global longitudinal strain (GLS) or LV mass index (Table  3 ). Absolute right ventricular longitudinal strain was larger in the placebo arm, although this difference was very small (−22.6 (3.8) vs. −21.4 (3.4), P = 0.048). Other right ventricular functional parameters (as assessed by right heart catheterization, MRI or echocardiography) were similar after placebo and trimetazidine treatment.

Exercise capacity, symptoms, and quality of life

There was no difference in change in 6‐MWD between placebo or trimetazidine (mean change of −5 m [95% CI −22, 22 m] vs. −6 m [95% CI −18, 7 m], respectively, P = 0.93; Table S5 ).

There was no difference between reported overall change in symptoms in both treatment arms (Table  S5 ). Furthermore, there was no difference between placebo and trimetazidine in overall QoL, as assessed by KCCQ and EQ‐5D‐5L (P = 0.78 and P = 0.51, respectively; Figure S4 and Table S6 ). However, there was an improvement in the self‐care domain of the EQ‐5D‐5L after trimetazidine treatment (P = 0.04). All other domains in both questionnaires did not differ between placebo and trimetazidine.

Metabolic and other laboratory parameters

Trimetazidine had no effect on fasting glucose, insulin resistance (HOMA‐IR or HOMA‐β), or on body weight (Table  S5 ). Trimetazidine did not result in an accumulation of circulating long‐chain acylcarnitines (all P > 0.05; Figure S5 ). Markers of myocardial stress and injury, that is, NT‐proBNP, high sensitive Troponin‐T, and C‐reactive protein did not differ between placebo and trimetazidine (Table  S5 ).

Safety and protocol deviations

Throughout the duration of the study, eight serious adverse events (SAE) had occurred in six patients, of which three in the trimetazidine arm, and five in the placebo arm (Table  4 ). Based on the medical histories and clinical course of the adverse events, we concluded that there was no relationship between the SAEs and the study medication or study procedures. Five of these patients were withdrawn from the study: one due to withdrawal of consent and four by study personnel because the potential SAE impact on the endpoints was considered substantial (see Supplemental Protocol Deviations in the supporting information for considerations) and were replaced by new study participants as per protocol. Although not significant, adverse events potentially related to therapy were more common in the trimetazidine arm than in the placebo arm (n = 33 vs. n = 21, P = 0.40; Table 4 ), particularly regarding neuromuscular symptoms (n = 20 vs. n = 9, P = 0.09). No patients stopped medication due to side effects. In two cases, neurological symptoms (spasms, tremors, and Raynaud's phenomenon) were classified as severe, and continuation of trimetazidine after the 3‐month study period was not recommended.

Table 4.

(Serious) adverse events

	Placebo (n = 30)	Trimetazidine (n = 29)
Serious adverse events	n = 5	n = 3
	Pneumonia	Tachycardia‐bradycardia syndrome complicated by pneumonia
	Appendicitis	Worsening of heart failure + tachycardia‐bradycardia syndrome
	COVID‐19 hospitalization (2×)	Airway infection with recurrence of fast atrial fibrillation a
	Death due to ischemic cerebrovascular accident
Adverse events	n = 34	n = 47
Patients with any adverse event	23 (77%)	23 (79%)
Patients with adverse event potentially related to study	16 (53%)	16 (55%)
Potentially related to medication	n = 21	n = 33
Neuro‐muscular	Dizziness (2×)	Dizziness (2×)
	Headache (3×)	Spasm/cramp in limbs (6×)
	Tingling on lips	Muscle pain (3×)
	Tingling arms	Tingling fingers
	Cramp in legs (short‐duration)	Severe tremor
	Formication (short‐duration)	Formication (short‐duration)
		Unstable tred (short‐duration)
		Raynaud's phenomenon (2×)
		Photophobia
		Hot flashes
		Increase in migraine frequency
Gastro‐intestinal symptoms	Gastro‐intestinal symptoms (short‐duration) (6×)	Gastro‐intestinal symptoms (short‐duration) (5×)
		Reduced appetite
Cardiovascular	Worsening of decompensation cordis requiring temporary increase of diuretics (2×)	Worsening of decompensation cordis requiring temporary increase of diuretics (3×)
	Pain on the chest (2×)	Worsening of fatigue (2×)
	Collapse	Mild decline in renal function
Other	Negative thoughts	Lethargy
Study procedure related	n = 0	n = 2
		Shoulder pain after MRI
		Neck pain (1 month) after puncture of the arteria carotis during right heart catheterization
No suspected relation to study	n = 13	n = 11
Inflammatory	COVID‐19 without hospitalization	COVID‐19 with de novo AF without hospitalization
	Urinary tract infection	Urinary tract infection (2×)
	Common cold (2×)	Common cold (2×)
	Gingivitis
Cardiovascular	Orthostasis (2×)	Hypertensive crisis
Neuromuscular	Backache	Backache
		Polyneuropathy
Other	Type II diabetes mellitus	Sun‐induced skin rash
	Keratoconjunctivitis sicca	Xerostomia
	Hyponatremia	Iron deficiency anaemia related to gynaecology problem
	Delirium (short‐duration)
	Ankle fracture

^a Patient experienced COVID‐19 hospitalization the following study period.

Discussion

This phase II double‐blind, placebo‐controlled, cross‐over trial investigated the effect of trimetazidine on cardiac performance and myocardial energy homeostasis in HFpEF. In contrast to our hypothesis, and despite our extensive and state‐of‐the‐art battery of non‐invasive imaging and invasive haemodynamic testing, essentially no effect of trimetazidine could be established. In particular, there was no effect on the primary endpoint: PCWP at multiple levels of exercise. Likewise, we did not detect any metabolic effect with ³¹P‐MRS or in plasma markers. In absence of any metabolic improvement through trimetazidine, we could not establish whether LV diastolic function and exercise haemodynamics improve with an improved myocardial energy homeostasis. Taken together, our findings indicate that trimetazidine may not be beneficial in HFpEF.

No metabolic improvement by trimetazidine in HFpEF

The lack of improvement in myocardial energy status in the present study is in apparent contrast with animal studies investigating the effect of trimetazidine in obesity‐ or diabetes‐induced cardiomyopathy ¹⁸ , ¹⁹ , ²⁰ and with the majority of the clinical studies that investigated trimetazidine in HFrEF. ⁸ , ⁹ , ¹⁰ , ¹¹

Whereas trimetazidine has been shown to improve the myocardial energy status and systemic insulin sensitivity in HFrEF studies, ⁸ , ⁹ , ¹⁰ we found no such beneficial effect of trimetazidine in HFpEF. Compared with the healthy heart, both HFrEF and HFpEF have a reduced mitochondrial oxidative capacity, reduced myocardial glucose oxidation and an increased myocardial glycolysis rate. ⁴ However, potential differences between HFpEF and HFrEF in any of the links in the complex metabolic chain may result in a less pronounced increase in glucose oxidation and consequentially a blunted improvement of the myocardial energy status upon trimetazidine treatment in HFpEF.

For example, the metabolism‐shifting potential of trimetazidine may be limited in moderate to severe LV hypertrophy where fatty acid oxidation may already be blunted. ²¹ Moderate LV hypertrophy was only present in a small subset of patients in this study, and sub‐analyses based on the presence of LV hypertrophy showed no differences in treatment effect. Furthermore, differences in circulating metabolic substrates may impact the potential effect of trimetazidine. Trimetazidine's inhibition of the last enzyme in the fatty acid oxidation pathway, 3‐ketoacyl coenzyme A (CoA) thiolase, may be relieved or reversed by competing higher levels of its substrate, 3‐keto‐hexadecanoyl CoA, in HFpEF. ²²

Indeed, circulating free fatty acids are often elevated in HFpEF, similar to obesity and type 2 diabetes. ⁴ , ²³ Elevated free fatty acids enhance peroxisome proliferator‐activated receptor‐mediated expression of pyruvate dehydrogenase kinase 4, inactivating the pyruvate dehydrogenase complex and reducing glucose oxidation. ²⁴ Consequently, trimetazidine might be less effective in shifting metabolism from fatty acid oxidation to glucose oxidation under conditions of high circulating levels of free fatty acids. Note that myocardial lipid levels, measured with localized proton MR spectroscopy, were indeed higher in our cohort of HFpEF patients compared with healthy volunteers (0.35% ± 0.13% of total water signal). ²⁵ Yet beneficial effects of trimetazidine have been reported in HFrEF patients independent of their diabetic status, as well as in obese subjects without heart failure. ⁹ , ²⁶

Recently, Hahn et al. used a metabolomics approach to demonstrate that in HFpEF, but not in HFrEF, the myocardial flexibility to use substrate alternatives to fatty acids, such as amino acids, ketones, and glucose, is lacking. ²³ That important finding in endomyocardial biopsy samples supports our current observations in patients. Indeed, the putative shift from fatty acid oxidation to glucose oxidation by trimetazidine did not improve myocardial energy homeostasis in our cohort. Instead, PCr/ATP trended to have decreased by 17% after trimetazidine treatment, suggesting that inhibiting fatty acid oxidation in HFpEF may worsen myocardial energy homeostasis. Moreover, recent studies with sodium‐glucose cotransporter 2 inhibitors (SGLT2i) in heart failure suggest a shift from glucose oxidation to fatty acid oxidation and particularly ketone body metabolism as one of the potential mechanisms for the beneficial effect of SGTL2i observed in HFpEF. ²⁷ , ²⁸ The effect of such inhibitors on the myocardial energy status is, however, still unclear. ²⁷

Although we consider it unlikely, several factors of the study design and execution could have further contributed to finding neutral trial results: (i) patient selection, (ii) dosage and duration, (iii) cross‐over design, (iv) outcome measures. We address these factors below.

Patient selection

All patients met the diagnostic criteria for HFpEF as confirmed by the gold standard. ¹ That is, all patients had elevated exercise PCWP measured either prior to inclusion or during the trial. Other potential causes (e.g., valve disease or ischemia) for their dyspnoea were extensively examined and excluded prior to inclusion. As such, we are convinced that HFpEF patients with clinically relevant LV diastolic dysfunction were selected.

The myocardial PCr/ATP in our cohort was lower than the value that we previously established for healthy volunteers (1.57 ± 0.17). ¹⁶ This low myocardial energy status is indicative of a compromised myocardial energy homeostasis in our cohort of HFpEF patients, and matches previously reported findings in HFpEF. ⁵ , ⁶ As such, these patients could potentially benefit from a metabolism‐modulating therapy that improves myocardial energy homeostasis. Moreover, it is likely that the patients included in this study were in a relatively early stage of HFpEF, with filling pressures at rest and NT‐proBNP values at near normal values in the majority of our patients. It could be argued that patients with more advanced HFpEF, that is, those who have a more pronounced exercise limitation and more reduced diastolic function, have more to gain from a new therapeutic heart failure drug. However, early HFpEF patients may be more susceptible to metabolism‐modulating therapy than patients with an advanced stage of heart failure, as advanced HFpEF patients are more likely to have abundant myocardial fibrosis and reduced myocardial viability. Furthermore, the blinded cross‐over design facilitated comparison of participants irrespective of their co‐morbidity and health status, consequently reducing variability and enhancing study power to detect any differences. Reducing the impact of physiological variability is of importance given the HFpEF phenotype heterogeneity.

In summary, a well‐defined, relatively early‐stage, symptomatic HFpEF population with a compromised myocardial energy homeostasis was investigated in this trial, suggesting that the specific patient characteristics for our trial were not precluding the detection of any beneficial effects of trimetazidine.

Dosage, duration, and cross‐over design

Both dosage and duration of treatment in this study were similar to several studies that demonstrated positive effects of trimetazidine (60–70 mg/day for >4 weeks) in HFrEF and/or coronary artery disease. ¹¹ Unfortunately, these studies did not report trimetazidine plasma levels. We confirmed excellent treatment compliance through pill count and the detection of adequate trimetazidine plasma levels, which were similar to other pharmacological studies with trimetazidine. ²⁹ Moreover, (neuromuscular) side effects were abundantly present, indicating that trimetazidine was biologically available and that higher dosages will likely not be tolerated.

Trimetazidine is a drug with a short elimination half‐time (6 hours). ³⁰ Implementation of the two‐week wash‐out period and the 2:1 either placebo‐first or trimetazidine‐first randomization mitigated a potential carry‐over effect. Absence of detectable trimetazidine plasma levels after cross‐over, supported by the results of additional statistical tests that demonstrated no influence of randomization allocation, strengthens our confidence that a carry‐over effect did not obscure any effects of trimetazidine.

Outcome measures

Elevated filling pressures during exercise play a central role in the pathophysiology of HFpEF, and are highly associated with symptoms of dyspnoea. ⁶ Therefore, PCWP measured at multiple levels of exercise is a strong surrogate endpoint for phase II studies in HFpEF, particularly because indirect estimates of elevated filling pressures from circulating NT‐proBNP levels or from echocardiography are inaccurate in HFpEF. ³¹ Others have suggested using PCWP/cardiac output slope as the primary readout for exercise haemodynamics, ³² but this parameter did not reveal any differences between placebo and trimetazidine either. Notably, Omar et al. found a change in PCWP during exercise in HFrEF with empagliflozin, whereas the PCWP/cardiac output slope remained unchanged. ³³ This underscores the validity of exercise PCWP as a surrogate endpoint for phase II HFpEF studies, with empagliflozin now shown to be beneficial in both HFrEF and HFpEF. ¹ , ²⁸

Although each methodology has its own merits and limitations, taken together, the lack of effect was not likely to be related to methodological limitations in study design or execution. Therefore, based on the plethora of different but relevant test results that we collected in this study, we conclude that there is a lack of improvement by trimetazidine in HFpEF.

Limitations

An important limitation to our single‐centre study is the small sample size. Using a cross‐over design, the study was sufficiently powered to detect differences in the primary and secondary endpoints. Of note, a cross‐over trial of 25 participants is powered equal to a parallel‐group study of 100 participants (with an assumed moderate correlation (r = 0.5) between repeated measurements). However, the study was not powered for other, potentially less sensitive, readouts such as echocardiography, 6‐MWD and QoL, nor was the study powered for sub‐analyses of different phenotypes.

Furthermore, metabolic assessments were performed in the fasting state and at rest, which is required to reduce measurement variability. Potential metabolic effects of trimetazidine during a non‐fasting state and during exercise (a situation with higher energy demand and potential hypoxaemia) may have been missed. Still, all clinical and cardiac performance endpoints at rest and during exercise remained unchanged, indicating that there was indeed no clinically relevant improvement in myocardial energy homeostasis.

Conclusion

Trimetazidine did not improve myocardial energy homeostasis, exercise haemodynamics, or cardiac performance in HFpEF, as assessed by an extensive combination of state‐of‐the‐art imaging and functional tests. Future studies should investigate whether other metabolic modulators can improve cardiac performance in HFpEF through the improvement of myocardial energy metabolism.

Funding

M.L.H. is supported by the Dutch Heart Foundation (NHS; 2020T058) and M.L.H. and A.J.B. by an ‘Out‐of‐the‐Box’ grant from the Amsterdam Cardiovascular Sciences (ACS) Institute, Amsterdam, The Netherlands. F.S.d.M is supported by the Netherlands Organisation for Scientific Research (NWO; 917.18.338, CVON DOLPHIN, PHAEDRA) and the Dutch Heart Foundation (NHS; 2018T059). There was no relationship with industry for this study.

Conflict of interest

M.L.H. received an educational grant and/or speaker/consultancy fees from Novartis, Boehringer Ingelheim, Daiichi Sankyo, Vifor Pharma, AstraZeneca, Bayer, MSD, and Quin; all not related to this work. All other authors report no (potential) conflict of interest.

Supporting information

Table S1. Inclusion and Exclusion Criteria.

Table S2. Reasons for Exclusion.

Table S3. Baseline Characteristics of All Included Patients.

Table S4. Exercise Right Heart Catheterisation Parameters.

Table S5. Exploratory Parameters.

Table S6. EQ‐5D‐5L Questionnaire.

Figure S1. Pulmonary Capillary Wedge Pressures at Multiple Levels of Exercise: Individual Data.

Figure S2. Phosphorus‐31 MR Spectra of the Heart during Placebo and Trimetazidine Treatment: Individual Data.

Figure S3. Mean Myocardial Proton MR Spectra during Placebo and Trimetazidine Treatment.

Figure S4. Kansas City Cardiomyopathy Questionnaires.

Figure S5. Plasma Acylcarnitine Profiles.

van de Bovenkamp, A. A. , Geurkink, K. T. J. , Oosterveer, F. T. P. , de Man, F. S. , Kok, W. E. M. , Bronzwaer, P. N. A. , Allaart, C. P. , Nederveen, A. J. , van Rossum, A. C. , Bakermans, A. J. , and Handoko, M. L. (2023) Trimetazidine in heart failure with preserved ejection fraction: a randomized controlled cross‐over trial. ESC Heart Failure, 10: 2998–3010. 10.1002/ehf2.14418.