Cardiovascular magnetic resonance feature tracking for characterization of patients with heart failure with preserved ejection fraction: correlation of global longitudinal strain with invasive diastolic functional indices

Heart failure with preserved ejection fraction (HFpEF) is a prevalent and growing public health problem [1]. Although the pathophysiology of HFpEF is multifactorial, left ventricular (LV) diastolic dysfunction, which is characterized by LV stiffness and relaxation, is recognized as the main cause [2‐4]. A previous study demonstrated that LV extracellular volume fraction (ECV) is a noninvasive indicator of LV stiffness in patients with HFpEF [5]. A more recent study employing speckle-tracking echocardiography found that systolic function measures such as LV global longitudinal strain (GLS) are frequently abnormal in HFpEF patients [6]. A recent study by Shah et al. also indicated that abnormal GLS is of value to identify patients with HFpEF at high risk for a cardiovascular event [7]. The cardiovascular magnetic resonance (CMR) feature tracking (FT; CMR-FT) technique enables the reproducible assessment of GLS from routine clinical CMR images with reduced observer dependency as compared to echocardiography [8]. However, the association between GLS determined by CMR-FT and the indices of diastolic function determined by cardiac catheterization has not been fully investigated in HFpEF patients.

Consequently, the purposes of this study were to determine the prevalence and severity of GLS impairment in patients with HFpEF by using CMR-FT and to evaluate the relationship between CMR-FT GLS and diastolic functional indices determined by invasive catheterization.

The baseline characteristics of the 18 HFpEF patients and the 18 healthy control subjects are summarized in Table 1. HFpEF patients were more likely to have hypertension and diabetes compared with the control subjects. A more frequent use of antihypertensive and heart failure medications was noted in HFpEF patients compared with the controls.

The main finding of this study is that among HFpEF patients CMR-FT GLS is independently associated with invasive measures of LV relaxation (Tau). Diastolic dysfunction is the hemodynamic consequence of the pathologies involved in HFpEF [5]. Prolongation of active myocardial relaxation and an increase in load-independent LV stiffness have been reported as the main mechanisms for diastolic dysfunction [5]. Therefore, the present finding suggests that systolic longitudinal dysfunction is closely associated with diastolic dysfunction in HFpEF patients. Previous studies have reported an inverse correlation between LV relaxation and LV contractility [26, 27]. According to those studies, the mechanism of the close association between systolic LV longitudinal dysfunction and impaired LV relaxation in HFpEF patients can be attributed to the elastic recoil of the LV myocardium [26]. During systole the myocardial wall stores energy in the form of elastic recoil, and this energy is released when the myocardium relaxes [26]. Thus, strain measurements by CMR-FT may enable the detection of diastolic dysfunction in the absence of an overt reduction in LV EF in HFpEF patients. It is recognized that endocardial dysfunction leads to depressed GLS and the preserved GCS and LV torsion usually compensates for the depressed GLS in HFpEF [28]. In our study, Tau had stronger association with GLS than GCS or GRS, suggesting that LV relaxation are closely associated with endocardial dysfunction. Furthermore, GLS measurement by CMR-FT is reproducible, easy to perform and less time consuming. GLS can be a meaningful tool in the routine clinical practice for patients with HFpEF.

LV stiffness is thought to be a consequence of an increase in extracellular matrix, reflecting abnormal diffuse myocardial fibrosis. In a recent study by Rommel et al., multivariate analysis revealed ECV as the only independent predictor of the myocardial stiffness constant (β), suggesting that in HFpEF patients with elevated ECV, the dominant pathomechanism is an increase in LV stiffness [5]. Rommel et al. mentioned that in addition to myocardial stiffness, impairment of active relaxation may be the important pathomechanism in HFpEF patients too. The findings of our study support their hypothesis. In our study, ECV value in controls was relatively high as compared to the previous studies. This might be because the mean age of controls was relatively high (61 ± 14 years) and 61% were female in which the ECV tends to demonstrate higher value.

In the present study, 39% of HFpEF patients showed impaired GLS with a value of less than − 13.9% (mean + 2SD of controls in our study). A recent study by Shah et al. using echocardiography reported impaired GLS as an independent imaging biomarker for identifying patients with HFpEF at high risk for cardiovascular morbidity and mortality [7], with a cut-off of − 15.8%. CMR-FT may also be able to identify HFpEF patients at high risk for a cardiovascular event; however, further study is required to establish the ideal cut-off value for CMR-FT. Another recent CMR study has shown that high ECV is associated with higher rates of morbidity and mortality in patients with HFpEF [29]. Furthermore, the study conducted by Mordi et al. demonstrated that echo-derived GLS and CMR-derived ECV are able to independently discriminate between hypertensive heart disease and HFpEF and identify patients with prognostically significant functional limitations [30]. In a similar manner, it is considered that measurement of LV strain and ECV by the CMR-only approach may provide two independent parameters of LV relaxation and stiffness that reflect the degree of diastolic dysfunction and may also have the prognostic implications in HFpEF patients. Further studies are needed to confirm the value of the measurement of LV strain and ECV by the CMR-only approach.

LA dysfunction is common in HFpEF because it is linked with LV dysfunction [20, 31]. Significant impairment of LA total strain in HFpEF patients was found in the present study, and LA total strain has recently been identified as a powerful prognostic factor in HFpEF patients [32]. Schuster et al. has shown that in the survivors of acute myocardial infarction LA total strain has incremental prognostic value in addition to any CMR measurements [33]. LA total strain might have incremental value for stratifying HFpEF patient prognosis, in addition to LV GLS and ECV. However, LA strain did not show the significant correlation with invasive measures of diastolic dysfunction in the current study and previous study [20], suggesting that LA strain parameters might offer exclusive clinical information which is only possible non-invasively. LA conduit function is closely related to LV stiffness [34]. von Roder et al. demonstrated that LA conduit strain was significantly impaired in HFpEF patients than controls and was the strongest predictor of exercise capacity [20]. However, in the present study, significant difference was not observed in LA conduit strain between HFpEF patients and control subjects. It might be speculated that LV stiffness in our patients was milder than the patients in the study by von Roder et al. Interestingly, Kowallick et al. demonstrated that LA conduit functions evaluated by CMR-FT can make discrimination among a hypertrophied phenotype, HFpEF and volunteers, as demonstrated by Mordi et al. where the discrimination between HFpEF and hypertensive heart disease was achieved based on GLS and ECV [35]. In the present study, there was no difference in RV volume or function between HFpEF patients and controls, whereas RA EF was significantly impaired in HFpEF patients compared with the controls. Recent study demonstrated that RV systolic function was preserved while RV early filling was impaired and compensated by increased RA booster pump function in compensated HFpEF patients [36]. Although it is difficult to interpret clinical implications of our finding at the present time, RV systolic function can be preserved with the impairment of RA function parameters in a certain condition. Future study focusing on RV and RA function in HFpEF patients is warranted.

The values of Tau and β in HFpEF patients were substantially different between our study and Rommel’s study. The reason for the difference may be due to the difference in the fitting equations to determine those values. In our study the best fit method assuming that pressure decayed to a non-zero asymptote was used for calculating Tau (P = P₀e^-t/Tau + P_B) [24], while Rommel et al. employed Weiss’s method (P = e^At + B) (asymptote = 0) [37]. To determine β, the formula of “P = A e^βV” was used in Rommel’s study [38], whereas the formula of “P = A (e^βV-1)” was used in the present study [25], where P is the LV pressure, V is the LV volume, A is a curve fitting constant. Consequently, our method provide larger Tau and smaller β compared to Rommel’s method. LV pressure volume analysis has remained a more research-based reference standard for confirming definite evidence of HFpEF due to its invasive nature. The pulmonary wedge pressure (PCWP) during physiological exercise emerged as the clinical reference standard to define HFpEF [39], which is clinically beneficial as it can avoid the risk of conventional LV pressure volume analysis. However, the study investigating the association between exercise PCWP and CMR functional parameters is still lacking. Further study will be needed.

In our study, cine images consisted of 20 phases per cardiac cycle. Cine images with lower temporal resolution are more prone to miss the short-lived events during the isovolumic period. However, longer breath-hold duration is required to obtain higher temporal resolution in cine CMR imaging. In this respect, typically cine CMR with 20–30 phases per cardiac cycle was used in the most of previous studies [15, 18, 20, 40, 41]. CMR-FT using cine images consisting of 20–30 phases per cardiac cycle substantially underestimated true GLS [42].However, the previous studies and our study successfully demonstrated that cine images with 20–30 phases per cardiac cycle can provide useful information in a clinical setting [15, 18, 20, 40, 41].

Echocardiography enables the noninvasive identification of diastolic dysfunction based on transmitral inflow (E and A values) or on myocardial compliance sampled at regional myocardial locations (e′) [43]. The global LV filling curves of cine CMR provide an alternative means of assessing diastolic physiology based on the timing and pattern of dynamic changes in LV chamber volume [13, 14]. The LV filling curve is usually transformed to the first derivative to obtain the early filling profile (i.e., PFR), which corresponds to the E value measured by echocardiography. Because it is influenced by filling pressure as well as inversely altered by changes in relaxation [26], E is usually corrected for the influence of relaxation (e′) in echocardiography (E/e′). In agreement with previous studies, in the present study PFR normalized by LV end-diastolic volume (nPFR) was significantly impaired in HFpEF patients compared with controls [13, 14]. However, the present study found no association of PFR or nPFR with Tau, which can be attributed to the load dependence of global LV filling curves. Abnormalities observed in the global LV filling curve may be less specific to the pathomechanism of diastolic dysfunction in the individuals with HFpEF when compared to GLS and Tau.

CMR-FT is a noninvasive approach that enables identification of the subgroup of HFpEF patients who have impaired LV GLS, and LV GLS assessed by this technique independently predicts abnormal LV relaxation measured by invasive conductance catheter in HFpEF patients. Diastolic dysfunction can be evaluated noninvasively in patients with HFpEF via the CMR approach using LV GLS in conjunction with ECV, which was previously shown to be an imaging biomarker for predicting LV stiffness.