The association of reduced left ventricular strains with increased extracellular volume and their collective impact on clinical outcomes

This was a prospective study approved by the institutional review board. Informed consent was obtained from all participants who were recruited from patients referred for clinical CMR due to a referring diagnosis of cardiomyopathy at our institution between January 2012 and December 2018. All participants completed a questionnaire for demographic information and medical history. The clinical charts were reviewed to confirm cardiovascular history. Exclusion criteria included implanted pacemaker or defibrillator, renal insufficiency with estimated glomerular filtration rate < 45 ml/min/1.73 m², significant arrhythmia, claustrophobia and metallic hazards. Normal controls were recruited among subjects without cardiovascular history or risk factors and had a normal electrocardiogram (ECG), transthoracic echocardiogram and CMR.

All subjects underwent scans in a 1.5 T CMR scanner (Avanto, Siemens Healthineers, Erlangen, Germany) with an 8-element phased array surface coil. No adenosine or regadenoson stress was administered. Cardiac volumes and systolic function were assessed using balanced steady-state free precession cine images with retrospective ECG gating during a breath-hold. Average temporal resolution was 50 ms. A stack of short axis planes (8 mm thickness, no gap) and 3 long axis planes (2-, 3-, and 4-chamber) were obtained using the following imaging parameter: echo time (TE) 1.3 ms, repetition time (TR) 3.1 ms, flip angle 70°, and average in-plane resolution 1.3 × 1.3 mm². T1 mapping was performed with a modified Look Locker inversion recovery sequence with a 3(3)5-scheme before and 15 min after contrast administration [6]. Native and post-contrast T1 were acquired from 3 short-axis slices (basal, mid and apical). The apical slice was chosen from the most proximal slice of the apical segment to avoid partial volume. The same T1 mapping sequence was used for the duration of the study. For late gadolinium enhancement (LGE) imaging a phase sensitive inversion recovery (PSIR) gradient echo sequence was performed 8–10 min after the administration of 0.15 mmol/kg of gadopentetate dimeglumine (Magnevist, Bayer Healthcare, Berlin, Germany) following TI scout imaging to determine the optimal inversion time.

Biventricular volumes and ejection fraction (EF) were analyzed based on a consecutive short axis stack of cine images using commercially available QMASS software (Medis Medical Imaging, Leiden, The Netherlands). LV volume and mass were normalized to body surface area.

T1 values of the myocardium and blood pool were assessed in the pre- and post-contrast T1 mapping images. We averaged T1 values of the basal, mid and apical slices. Hematocrit was drawn on the day of CMR. ECV was calculated using the formula:

where R1 = 1/T1 time [7].

LV strains, including GLS, global circumferential strain (GCS) and global radial strain (GRS) were analyzed using feature tracking on cine images. The feature tracking contours of the LV endocardium and epicardium were drawn manually at end diastole which were propagated automatically to all phases [8]. Contours were manually corrected when automatic tracking failed. The peak strain values for GLS, GCS and GRS were determined on the strain curves where GLS and GCS were of negative values and GRS was of positive value. Both T1 mapping and feature tracking analyses were performed using commercial software (cvi42, version 5.6.5, Circle Cardiovascular Imaging Inc., Calgary, Alberta, Canada). The examples of ECV and strain profiles from a normal control and a patient are shown in Fig. 1. To test the reproducibility of T1 mapping we randomly selected 15 cases and evaluated by 2 readers with consensus reading. The reproducibility of strain was reported by our group previously using the same program [9].

In the reproducibility analysis, the intra-observer reliability measured by ICC was 0.992 (0.981 to 0.996) for ECV, 0.990 (0.977 to 0.996) for native T1 and 0.993 (0.985 to 0.997) for post-contrast T1. The inter-observer reliability measured by ICC was 0.961 (0.759 to 0.988) for ECV, 0.976 (0.940 to 0.990) for native T1, and 0.968 (0.854 to 0.990) for post T1. All data are presented as the averaged ICC of each component variable.

There are several limitations to our study. First, optimal sampling for ECV calculation is debatable. We chose basal, mid and apical representative slices to aim for better coverage while some published reports favor not including the apical slice due to concerns over the partial volume effect in apical region [23]. Our approach to analyze the apical slice that is closest to the mid ventricle allowed us to avoid the thin apical wall and ensure good reproducibility. While ECV assessment based on 3 slices generated better coverage for tissue characterization than that of 2 slices, it may still underestimate the disease burden particularly for disease with heterogeneous distribution of fibrosis. The future development of 3D T1 mapping with free breathing will improve coverage but will not burden the patient. Second, our cohort was from a single center with a relatively modest population. However, we were able to demonstrate the intricate link between tissue property alteration and mechanical impairment, as well as their collective impact on adverse clinical outcomes. Third, this was a clinically referred cohort with mixed diagnosis of ischemic and non-ischemic diseases. While a particular referring pattern may be subjective to institutional practice, the collective cases in our cohort have typical representations of common CMR referrals, thereby making our findings generalizable. Fourth, the absolute value of GLS appears to be low, which is subject to the selection of feature tracking software as we demonstrated in our previous publication [9]. We used feature tracking for strain analysis in our study because it has the advantage of using cine images that are routinely captured nearly in all clinical cases. But we acknowledge other important modalities for strain evaluation such as tagged cine images, SENC and displacement encoding with stimulated echoes (DENSE) which allow for myocardial deformation to be analyzed using features encoded directly onto the myocardium [24‐26] although additional dedicated breath-hold imaging is required that adds scan time and burden to the patient. Finally, this is a single center study largely limited to patients of Caucasian background. The incidence of cardiac death was low. In addition, the outcome event rate was low in the subgroup of dilated cardiomyopathy (12 out of 81) rendering insufficient statistical power for outcome analysis. Future studies with multi-center designs and large clinical cohorts are warranted to compare ECV and strain with other well-established biomarkers, such as B-type natriuretic peptide and diastolic function, to improve risk stratification.