The interplay between metabolic alterations, diastolic strain rate and exercise capacity in mild heart failure with preserved ejection fraction: a cardiovascular magnetic resonance study

A group of 27 patients with HFpEF defined by the presence of symptoms or signs of HF, a non-dilated LV with LVEF ≥50% and evidence of abnormal diastolic function on Doppler echocardiography were prospectively enrolled between 2011 and 2018 [18, 19]. Additionally, we performed cardiopulmonary exercise testing (CPET) to confirm their exercise limitation by measuring VO₂ max (VO₂ at peak exercise). Patients were enrolled if their VO₂ max was < 80% predicted for age and gender, with a pattern of gas exchange that would indicate a cardiac cause of exercise limitation. Patients who had diabetes mellitus, uncontrolled hypertension, significant valvular disease, previous myocardial infarction, coronary revascularization, previous cardiac surgery, contraindications to CMR or estimated glomerular filtration rate < 30 ml/min were excluded. Fourteen age, gender and body mass index (BMI) matched healthy controls without a history of heart disease, diabetes, hypertension or dyslipidaemia were also recruited. They were identified from the local population by word of mouth and poster advertisements around hospital and university. Controls were included if they were 65 years or older, asymptomatic, not on any medications and had no cardiac abnormalities detected on electrocardiogram (ECG) or CMR. All study participants fasted for at least 6 h prior to CMR and CMRS and were scanned at around the same time of the day.

All study participants underwent resting transthoracic echocardiography, ³¹P-CMRS (n = 20 for HFpEF and n = 10 for controls) and ¹H-CMRS in addition to the standard CMR imaging. All HFpEF patients and a subgroup of healthy controls (n = 9) underwent CPET.

Myocardial ¹H spectra were obtained from the mid interventricular septum as previously described [11]. In short, spectroscopic acquisitions were performed using ECG trigger at end-expiration to minimize motion artefacts. Water-suppressed spectra were acquired to measure cardiac lipid, and spectra without water suppression were acquired and used as an internal standard. Spectroscopic stimulated echo (STEAM) sequence with an echo time (TE) of 10 ms was used. Five to six water-suppressed scans (5 averages each; repetition time (TR) of at least 2 s) were acquired in mid-diastole in a series of single breath-holds of about 10 s each. Next, a water spectrum (3 averages; TR of at least 4 s) was acquired in a single breath-hold to use as internal reference. Spectra were analysed using a custom Matlab (Mathworks, Natick, Massachusetts, USA) implementation of AMARES (Advanced Method of Accurate, Robust and Efficient Spectroscopic) and the AMARES algorithm in jMRUI (Java-based Magnetic Resonance User Interface) as previously described [11]. Myocardial triglyceride content was calculated as a percentage relative to water: (signal amplitude of lipid/signal amplitude of water) × 100%.

³¹P-CMRS was performed to obtain PCr/ATP ratio from a voxel placed in the mid-ventricular septum, with the subjects lying prone with their heart over the centre of the ³¹P heart/liver coil in the magnet isocentre as previously described [20]. ³¹P-CMRS post processing analysis was performed using a custom Matlab analysis tool, as previously described [20]. Coil positioning was confirmed or adjusted with the use of proton localization images as previously described. Spectra were acquired using acquisition-weighted 3D chemical shift imaging (CSI) [21], with 10 averages in the centre of K-space. The TR was 720 ms and the acquisition duration was 8 min [22].

Cine CMR images were acquired for cardiac volume analysis using a 3 T CMR system (Trio, Siemens Healthineers, Erlangen, Germany) using bSSFP cine imaging as previously described [23]. Analysis of cardiac volumes, function and mass was performed using Argus post-processing software (Siemens Healthineers). The systolic circumferential strain and diastolic strain rate parameters were calculated using feature tracking software Circle, Cardiovascular (cvi42®, Circle Cardiovascular Imaging Calgary, Canada) from the short-axis bSSFP cine images. The epicardial and endocardial borders were traced manually at diastole and the software then tracked the deformation throughout the cardiac cycle generating values for global circumferential strain and circumferential diastolic strain rate. In case of insufficient tracking of the endocardial and epicardial borders, contours were manually corrected and the tracking repeated.

Transthoracic echocardiography (echo) was performed using a commercial echocardiographic system (iE33, Philips Healthcare, Best, The Netherlands). Measurements of LV diastolic function were performed according to the guidelines [24]. The following diastolic indices were obtained; transmitral early (E) and late (A) diastolic velocities, mitral annular early (e’) diastolic velocity, with calculation of E/A and E/e’ ratios.

Cardiopulmonary exercise testing was undertaken using an upright treadmill or bicycle ergometer protocol with simultaneous respiratory gas analysis after performing a spirometry, as described [16]. An incremental ramp protocol was utilised whereby speed and inclination (for treadmill) or resistance and speed (for bicycle exercise) were gradually increased with continual heart rate, blood pressure and ECG recording. All subjects were exercised to volitional fatigue, with a corresponding adequate respiratory exchange ratio (RER) achieved as a requirement for satisfactory effort defined as RER of > 1. Maximal oxygen consumption (VO₂ max) was determined by averaging VO₂ measures over 30 s of peak exercise.

A priori sample size calculation was performed which was based on a change in PCr/ATP ratio. With a power of 80% and p < 0.05, a sample size of 18 would be required to identify a difference of 0.39 in PCr/ATP. As the true effect size for MTG was unknown in this population, a post-hoc analysis using MTG was performed which showed that 30 subjects (20 HFpEF and 10 controls) would have a power of 86%. Data are expressed as mean ± SD for description of study cohorts and mean ± SE for CMR and CMRS variables. Categorical data are presented as numbers and percentages. Comparisons between the two groups were performed by non-parametric method due to small sample size. The Chi-squared test or Fisher’s exact test were used to compare categorical data as appropriate. Bivariate correlations were performed in all subjects using Pearson’s or Spearman’s method as appropriate. Non-normally distributed data were log transformed to construct normal data. Variables with significant correlations with diastolic strain rate and VO₂ max were entered into a stepwise multivariate model to determine predictors of diastolic strain rate and VO₂ max. A P-value < 0.05 was considered significant. All statistical analyses were performed with SPSS (version 21, International Business Machines, Armonk, New York, USA).

Sixty-five patients with chronic breathlessness under follow up in the cardiology clinics were identified and screened in addition to 7 patients who were recruited by posters. Of these, 37 were eligible to enter the study, of which the first 31 consecutive patients took part in the study. Four patients were excluded due to incomplete study protocol. None of the patients had prior HF admissions, prior insults such as chemotherapy or radiation. None of the patients had been diagnosed or suspected to have amyloidosis, and CMR did not show findings suggestive of amyloidosis. Patients with history of angina, previous myocardial infarction or coronary revascularization were excluded. Although coronary angiography was not performed prior to entering the study, none of the patients had a history of angina, and they did not have angina or develop any ischaemic changes during CPET.

Table 1 shows demographic, clinical, echocardiographic, biochemical and VO₂ max results for both HFpEF and healthy control groups. Both groups were matched with regards to age, gender, BMI and glucose and blood cholesterols. Frequency matching instead of individual (one-to-one) matching was performed which was based upon proportion of the age and gender for both cohorts. It was challenging to achieve equal number of controls to the HFpEF group, as a significant number of healthy elderly subjects had to be excluded due to concomitant medical problems. Although the age (72 vs 69) and gender (67 vs 57%) were not perfectly matched, the differences are numerically small (3 years for age and 10% for gender), and statistically they were not significant (p > 0.05). All HFpEF patients had (1) signs or symptoms of HF, (2) normal LVEF and LV cavity size, (3) evidence of mild diastolic dysfunction on Doppler echocardiography, and, additionally (4), objective evidence of cardiac cause of exercise limitation on CPET. We decided to use the latter as an additional inclusion criterion, as this provided evidence of a cardiac cause of exercise intolerance in our mild HFpEF population, rather than being attributed to physical deconditioning, commonly seen in healthy elderly [16, 25, 26], and the predicted values are entirely consistent with a meta-analysis of studies from which reference values were developed [27].

Table 2 summarises CMR results for both groups. When compared to controls, HFpEF showed concentric remodeling as indicated by increased LV mass to LV end diastolic volume (EDV) ratio but normal LV mass index. As expected diastolic strain rate in HFpEF was significantly impaired. Despite normal LVEF, peak systolic circumferential strain was significantly impaired in HFpEF, indicating additional subtle contractile dysfunction. None of the HFpEF or controls had a late gadolinium enhancement (LGE) pattern indicating previous myocardial infarction. To assess gender difference in LV functional parameters, analyses were performed comparing LVEDV, LV end systolic volume (ESV) and LVEF between women and men separately in both groups. There were no significant gender differences in LV volumes and function in both groups (results not shown).

Table 2 also shows the comparison of ¹H and ³¹P CMRS results for all study groups. Compared to controls, HFpEF patients had pronounced steatosis (a 2.3-fold increase) and impaired energetics despite similar age, gender and BMI (Fig. 1).

Elevated MTG (Fig. 2) and reduced PCr/ATP ratio correlated significantly with impaired diastolic strain rate but no significant correlations with other diastolic indices such as transmitral E/A or E/e’. There were no significant correlations observed between MTG and PCr/ATP with echo diastolic indices, as the echo variables were not powered to detect these differences. On stepwise multivariate regression analysis, MTG but not PCr/ATP independently correlated with diastolic strain rate (adjusted R² = 0.48) (Table 3). Increased MTG, reduced PCr/ATP, and impaired diastolic strain rate (Fig. 2) correlated significantly with reduced VO₂ max but on stepwise multivariate analysis only diastolic strain rate was an independent correlate of VO₂ max (adjusted R² = 0.30). As expected both MTG (r = 0.35, p = 0.027) and diastolic strain rate (r = − 0.48, p = 0.001) correlated significantly with New York Heart Association (NYHA) class suggesting there may be a relationship between cardiac lipid and exercise tolerance.

By using the cut off values of abnormal lipid/water > 0.64 and abnormal PCr/ATP of < 2.0 from the current study, of the 27 HFpEF patients, 21/27 (78%) had abnormal lipid/water and 17/20 (85%) had abnormal PCr/ATP. Of the 14 controls, 3/14 (21%) had abnormal lipid and 4/10 (40%) had abnormal PCr/ATP and the rates of these abnormalities between HFpEF and controls were statistically significant (p < 0.05). Furthermore, the patients with abnormal lipid/water and PCr/ATP ratio had significantly lower LV diastolic strain rate and VO₂ max than those with normal lipid/water and PCr/ATP ratio (results not shown). By using the cut-off values of abnormal lipid/water > 0.77 and abnormal PCr/ATP < 1.6, of the 27 HFpEF patients, 19/27 (70%) had abnormal lipid/water and 15/20 (75%) had abnormal PCr/ATP.

In line with our previous work [13], we found that MTG had a significant correlation with peak systolic circumferential strain (r = 0.61, p < 0.001). In addition we found that reduced PCr/ATP correlated with impaired peak systolic circumferential strain (r = − 0.55, p = 0.002) but no significant correlations with age, BMI, systolic blood pressure (SBP) or BNP.

As this study is limited by a small sample size, further corroboration of these findings in larger-scale multi-centre studies is required. Although the current study is powered to detect a difference, it may be underpowered to confirm the null hypothesis. Therefore, the non-statistical age difference between control and HFpEF cohorts would imply that age could still be a confounding variable for the difference in MTG observed. However, it would not be physiologically possible for a small age difference to result in > two-fold increase in MTG. In the present study, we excluded patients with a history of coronary artery disease and there was no evidence of myocardial infarction on CMR LGE imaging. Myocardial stress perfusion to assess microvascular dysfunction was not performed, as we did not want to burden these frail patients with a longer scan protocol. Thus, reduced coronary microvascular reserve during stress cannot be excluded as a potential mechanism contributing to the metabolic alteration in HFpEF. Due to the relatively long scan protocol, it was not feasible to perform reproducibility tests in the current study, thus reproducibility of lipid/water and PCr/ATP techniques in HFpEF is unknown. However, the reproducibility of these techniques have been shown to be excellent in healthy subjects [11, 20]. The current study is not representative of the HFpEF population as patients with diabetes and uncontrolled hypertension were excluded. In the future, the analysis should be repeated in a larger sample size including patients with these comorbidities. As there is lack of accepted standard MTG normal threshold, the interpretation of the results may be limited as the findings are based on the normal MTG value at our institution. The relatively low MTG (0.64) in controls could reflect the healthiness of the cohort control that is not representative of normal aging. Future studies will be needed to investigate the true MTG in old sedentary individuals. Similarly, the participation of very healthy subjects cannot be excluded as a potential factor contributing to the supra normal PCr/ATP. Although results vary slightly between centres, the mean PCr/ATP value for healthy controls in the current study is consistent with the values reported previously in this age group [16, 41]. Finally, the observational nature of our study precludes inferences of attribution. Further interventional research is needed to examine the change in MTG or PCr/ATP to confirm causal relationship between diastolic dysfunction/VO₂ max and metabolic abnormalities.