The participants underwent echocardiography using a commercially available ultrasound system (iE33, Philips Medical Systems, Andover, MA, USA) equipped with a 2.5–4 MHz transducer. Images were taken by an investigator and reviewed offline by two experienced echocardiographic observers. Three cines of cardiac cycles were digitally recorded and stored for subsequent analysis. Chamber dimensions and thicknesses were obtained according to American Society of Echocardiography (ASE) recommendations.¹⁷

Two-dimensional images were obtained at a cycle frequency of 50–70 cycles per second using tissue harmonic imaging and a high-quality electrocardiogram signal. The images were evaluated on a workstation equipped with QLab software (Philips Medical Systems). STE was used to assess myocardial deformation. All images were analyzed using the model proposed by the American Heart Association/American College of Cardiology with 16 segments.¹⁸ Apical four-, three-, and two-chamber views were used to measure the global longitudinal strain (GLS) and the mechanical dispersion index (MDI), also obtained by longitudinal strain, which refers to the standard deviation of time for maximum deformation of the 16 segments. Mid-ventricular short-axis images were used to assess circumferential (SCi) and radial (SR) strain.

The normal percentage of peak longitudinal systolic strain (LS) varies among software references. The Phillips QLab reference value is 18.9% ± 2.5%. The lower limit of normal is 14%. The SCi and SR reference values were 22.1% ± 3.4% and 35.1% ± 11.8%, respectively. As suggested, the absolute numerical value without the negative or positive sign was used when referring to LS, SCi, or SR.¹⁰ An MDI greater than 55 ms is considered high.^19,20

CMR was performed on an Avanto 1.5 T full-body scanner (Siemens Medical Solutions, Germany) using an eight-channel cardiac coil array. Cine images were acquired using a proton-free processing sequence synchronized with heart rate during apnea (acquisition rate, 20 frames/cardiac cycle; slice thickness, 8 mm; field of view, 300; matrix, 208; Åx, 80; bandwidth, 925 KHz/pixel). LV function was analyzed using eight and 12 8-mm-thick and 2-mm apart short-axis slices. Ventricular volume, mass, and systolic function, including right ventricle and LV ejection fraction, were calculated using CMR cine images and Argus 4D VF software. Endocardial contours were manually drawn during systole and diastole.

Multi-axis acquisition of cardiac prototypes was performed using the T1 MOLLI sequence for individual short-axis slices. Automated T1 maps were generated for manual calculation using analysis of the region of interest.

Fibrosis was analyzed using delayed myocardial enhancement (DME) after the intravenous infusion of 0.2 mmol/kg of gadolinium-tetraazacyclododecane tetraacetic acid (Dotarem^® [gadoteric acid]; Gerbet Produtos Médicos, Rio de Janeiro, Brazil). DME images were taken in short-axis view with 8-mm thickness from the LV base to the apex at 5–15 minutes after the gadolinium administration using the following parameters: repetition time, two RR intervals; excitation time, 4–4.8 ms; 256 × 192 matrix image; field of view, 300–400 mm; spacing, 2.0 mm; deflection angle, 30°; and inversion time, adjusted on a case-by-case basis to nullify the normal myocardial signal.

Continuous variables are expressed as mean ± standard deviation or median and interquartile range (IQR; 25%–75%) depending on normality on the Shapiro-Wilk test. Categorical variables are expressed as total number (percentage). Student’s t-test was used to examine normally distributed numerical variables, while the Mann-Whitney nonparametric test was used to examine non-normally distributed variables. Pearson’s correlation coefficient was used to assess the association between GLS and TTE and T1 mapping parameters. Binary logistic regression was used to test the independent association between strain and CMR variables. Receiver operating characteristic (ROC) curves were constructed to identify the best strain and T1 mapping cutoff points to predict fibrosis. Considering the study’s exploratory nature, sample size was not measured. Based on a published article,¹⁵ we included 25 participants. The analyses were performed using SPSS software version 21.0 (IBM), and p values <0.05 were considered statistically significant ( test).

Three (13%) research subjects were excluded from the analysis for poor echocardiographic image quality, while another was excluded due to cardiac arrhythmia. Thus, the final study population was 21 participants. Women accounted for 62% of the patients, and the mean patient age was 54 ± 5 years.

Six (29%) participants had a significant segmental change, including four (19%) with apical LV aneurysms, one (5%) with a basal septal aneurysm, and one (5%) with basal hypokinesia in the inferior lateral wall. The patients’ clinical characteristics and echocardiographic data are shown in Table 1.

The GLS was reduced in 81% of the samples (median, 14.1%; IQR, 12.1–16.3). Of them, 10 (48%) participants already had a frankly abnormal GLS decrease (< 14%). Table 2 shows the LS values.

The apical and middle portions of the LV were the most involved (62% and 57%, respectively), while only 38% of the basal portion had a GLS with a frankly abnormal decrease.

The myocardial performance index presented a mean of 88.6 ms and a median of 83.4 ms (IQR, 58.6–108.6 ms).

The GLS was significantly correlated with LVEF (r = -0.512; p = 0.018) and the S’ value (r = - 0.511; p = 0.018). GLS differed significantly between groups with and those without segmental changes (12% ± 1.2% versus 14.5% ± 2.8%, p = 0.049).

The patients’ SCi and SR myocardial deformations at the mid-ventricular level are shown in Table 3. The mean SR was significantly reduced in 13 patients (65%), while the mean SCi was significantly reduced in 11 patients (55%).

Sixteen participants underwent CMR. The prevalence of MF by DME in the sample was 50%. The patients were divided into two groups according to the presence or absence of MF. The clinical characteristics did not differ significantly between the groups. The patients’ clinical and echocardiographic characteristics (Table 4) did not differ significantly between groups.

The prevalence of segmental contractility changes detected on TTE was higher in the MF group (37.5% versus 12.5%). The two cases of grade II diastolic dysfunction were also in the MF group.

Strain values in the different orthogonal, longitudinal, circumferential, and radial axes did not differ by the presence of MF by DME. The median 16-segment MDI obtained by strain was higher in the MF group (90.7 m; IQR, 72–121 ms) than in the group without MF (67 ms; IQR, 53.5–109 ms), but the difference was not statistically significant (p = 0.23).

ROC curves were constructed to analyze the discrimination capacity of GLS for the presence of MF using delayed gadolinium enhancement. The GLS presented a C-statistic (AUC) of 0.516 (95% confidence interval [CI], 0.215–0.816). The best GLS cutoff point was 13.97%, with a sensitivity of 62.5% and specificity of 50%. MDI had the highest AUC (0.696; 95% CI, 0.412–0.981) (Figure 1). The best MDI cutoff point was 85 ms, with a sensitivity of 75% and specificity of 71%. Six of the eight patients (75%) with MF had an MDI > 85 ms, while only two (25%) without MF had an MDI > 85 ms, and the difference was of borderline statistical significance (p = 0.067) (Table 7).

T1 mapping with native T1-weighted early global enhancement sequence was also used to assess MF. T1 mapping values were high in patients with CD (mean, 993 ± 163 ms). The median T1 mapping values were 1,033 for the group with MF (IQR, 998–1.081), and 1,010 for the group without MF (IQR, 1,002–1,047). The times were even higher in the MF group, corroborating the finding in cases of other dilated cardiomyopathies. According to the ROC curve analysis, the best T1 mapping cutoff point for identifying MF was > 1,013.6, with a sensitivity of 71%, specificity of 57%, and AUC of 0.592 (Figure 1).

In the simple linear regression analysis, T1 mapping value was significantly correlated with GLS (r = -0.634; p = 0.015). A GLS < 13% had 83% sensitivity and 50% specificity for identifying a T1 mapping value > 1,013.6.