The present study retrospectively analyzed patients who underwent CMR in our center between February 2011 and March 2017 with a final diagnosis of NIDCM, defined according to current recommendations.11 The patients’ baseline characteristics and electrocardiogram and treatment data were collected from their medical records. Changes over time in the main morphofunctional variables were studied in the patients who underwent follow-up echocardiography. Similarly, the incidences were analyzed of other variables, such as heart failure, implantable cardioverter-defibrillator (ICD) implantation, and death. In addition, a composite endpoint was examined: hospitalization due to heart failure, ICD implantation in secondary prevention, and overall death.

CMR studies were performed using a 1.5-T Signa HDxt system (GE Healthcare, United States); standard SSFP cine sequences were obtained (20 images/cycle) in longitudinal axes and in 10 to 15 contiguous short-axis slices covering both ventricles from the base to the apex. LGE images were acquired about 8 to 10minutes after intravenous infusion of gadobutrol at 0.2 mmol/kg (gadovist 1 mmol/mL). The volumes and ejection fractions of both ventricles were calculated from the cine sequences using the disk summation method with specific software (ReportCARD, GE Healthcare, United States). Hyperenhancement presence and distribution were qualitatively evaluated in the LGE sequences.

The FT analysis was performed using QStrain RE v2.0 software (Medis, The Netherlands). Using short- and long-axis views, end-diastolic and endsystolic volumes were obtained, as well as ejection fraction and the LV myocardial strain parameters: GLS, GCS, and GRS (figure 1). To determine intraobserver variability, the same researcher later reanalyzed 20 randomly chosen patients. The interobserver reproducibility was assessed through the analysis of 20 randomly chosen patients by an experienced CMR reader.

Figure 1.

Determination of myocardial strain using feature tracking. In these longitudinal projections (A), the global longitudinal and circumferential strain can be determined and the radial strain inferred, as well volumes and ejection fractions, whereas, in the short axis (B), the global circumferential and radial strain can be measured and data obtained on myocardial torsion.

(0.52MB).

Qualitative variables are presented as absolute number and percentage and quantitative variables as mean ± standard deviation or median [interquartile range] as appropriate. The intraclass correlation coefficient (ICC) was used to evaluate the intraobserver and interobserver variabilities in the FT determination of strain. Similarly, the Pearson r coefficient and ICC were used to evaluate the concordance between the measurements obtained using the traditional disk summation method and the FT method. The chi-square test was used to compare categorical variables. Differences in continuous variables with normal and nonnormal distributions were evaluated using the unpaired Student t test and the Mann-Whitney U test, respectively. The association was analyzed of the FT-derived LV strain parameters with event occurrence. To identify independent prognostic factors, logistic regression models were built using backward selection with other morphofunctional CMR variables. The results are expressed as odds ratios (ORs) with 95% confidence intervals (95%CIs). ROC curves were used to identify the strain values with best predictive ability. These were used in the Kaplan-Meier survival curve analysis and, together with other morphofunctional CMR variables, using Cox regression with backward selection. Differences were considered statistically significant at a 2-tailed P value < .05. The data analysis was performed with SPSS version 18.0 (SPSS Inc, United States).

In total, 98 patients diagnosed with NIDCM were finally included and FT analysis could be performed in all patients. Significant coronary heart disease was ruled out using invasive coronary angiography (89%) or cardiac computed tomography (4.1%). No patients had LGE with an ischemic pattern.

Regarding the baseline characteristics (table 1), 71.4% were men and the mean age was 69 ± 13 years. The cardiomyopathy was considered idiopathic in 75.5% and alcoholic in 10.2%. Relevant findings in the pre-CMR electrocardiogram included atrial fibrillation in 22% of patients and complete left bundle branch block in 42.9%. A high level of optimal medical therapy was seen (including beta-blockers, 92.8%; angiotensin-converting enzyme inhibitors [ACEIs] or angiotensin II receptor blockers [ARBs], 78.5%; and aldosterone antagonists, 61%).

The morphological and functional characteristics of CMR are collected in table 2. The study patients had severe LV dilatation (end-diastolic volume index [EDVI], 133.6 ± 33.4mL/m2) with severe systolic dysfunction (LVEF, 29.5% ± 9.6%). However, most patients had a right ventricle of normal size and function. In 38.8%, LGE revealed the presence of fibrosis; the most frequent pattern was intramyocardial septal (16.3%). In line with the above LV function values, the myocardial strain parameters were significantly altered: GLS, –9.3% ± 4.2%; GCS, –7.7% ± 3.5%; and GRS, 16% [10.2%-24.6%]. Evaluation of the associations among the variables showed a good correlation of GLS with GCS (r = 0.739; P < .001) and a moderate correlation of GLS with GRS (r = –0.539; P < .001).

In terms of the intraobserver variability, there was an excellent ICC for GLS (0.95; P < .001) and GCS (0.87; P < .001) and a good ICC for GRS (0.62; P = .021). Similar, although slightly worse, concordance values were found for the interobserver evaluation: GLS (0.98; P < .001), GCS (0.56; P = .042), and GRS (0.45; P = .103). The time required for myocardial strain analysis using FT was 7.8 ± 2.3minutes.

In addition, a comparison of the LV volume and function values determined by the traditional disk summation method and by the FT method revealed high concordance between the 2 techniques: EDVI (r = 0.896, P < .001; ICC = 0.838, P < .001), endsystolic volume index (r = 0.944, P < .001; ICC = 0.928, P < .001), and LVEF (r = 0.882, P < .001; ICC = 0.923, P < .001) (figure 2).

Figure 2.

Correlation between LVEF measurements obtained via the traditional disk summation method and feature tracking. LVEF, left ventricular ejection fraction.

(0.08MB).

No associations were found between the baseline clinical characteristics and the FT-derived LV myocardial strain values, except for a weak negative correlation between age and GRS (r = –0.28; P = .005). Similarly, there were no differences in the values obtained according to the presence of atrial fibrillation or complete left bundle branch block during the test.

Both LV volumes and functions and left atrial size were significantly correlated with all LV strain values obtained using FT (table 3), particularly GCS. However, no differences were seen in the values according to the presence of fibrosis on LGE.

Follow-up echocardiography was performed in 84 of the patients (85.7%) at a median of 2.4 [1.8-3.4] years after the CMR study. These patients had an LVEF of 43.6% ± 10.8%, with at least moderate mitral regurgitation in 20 patients (20.4%) and an E/e’ value of 10 [7-14]. In addition, 71.4% had a higher LVEF in this study than in the baseline CMR, with a median increase of 13%; a LVEF > 50% was found in 25 patients (25.5%). Both the LVEF during follow-up and the change from baseline were significantly correlated with all FT-derived LV values and with the baseline LVEF on CMR (table 4). In contrast, the presence of LGE was only associated with a lower LVEF during follow-up (40.4% vs 45.5%; P = .04). Those patients with preserved systolic function (LVEF > 50%) during follow-up had been found to have better GCS values (–9% vs –7.1%; P = .019), without differences in the other FT-derived LV values or the baseline LVEF or EDVI. LGE was less frequent in these patients (16.1% vs 37.7%; P = .037). On multivariate analysis, GCS was independently associated with the presence of an LVEF > 50% during follow-up (OR = 1.16; 95%CI, 1-1.34; P = .045), and not LGE.

Event incidence during a median follow-up of 3.2 [2.2-4] years is shown in table 5. An ICD was implanted in 22 patients, mainly in primary prevention. Of the patients with an ICD indication in secondary prevention, 2 had sustained monomorphic ventricular tachycardia, 2 had syncope and nonsustained monomorphic ventricular tachycardia, and 1 had cardiac arrest. In total, 25.5% of patients required admission due to HF. Ten patients died during follow-up, with a cardiovascular cause confirmed in 5 of them. Of the remaining deaths, 4 patients died at home less than 24hours after symptom onset, with an unidentified immediate cause of death, and 1 patient died of respiratory failure of unknown cause.

No significant association was found between admission due to HF and FT-derived LV myocardial strain values, except a tendency for lower GCS values (–7.1% vs –8.5%; P = .10). Mortality during follow-up was significantly associated with GCS (–5.9% vs –7.9%; P = .012) and tended to be associated with GLS (–6.9% vs –9.6%; P = .051). There were no differences in FT-derived strain values according to ICD implantation in secondary prevention. During follow-up, 34 patients (34.7%) had the composite event (admission due to HF, ICD in secondary prevention, or overall mortality); of the FT-derived strain parameters, only GCS was associated with this outcome (table 6). Of the remaining baseline morphological variables on CMR, only LVEF and EDVI were significantly associated with worse prognosis. Accordingly, the multivariate analysis included GCS, LVEF, and EDVI. GCS was independently associated with prognosis (OR = 1.15; 95%CI, 1.01-1.31; P = .038). Through the ROC curve, a GCS value of –8.2% was selected because it showed the best predictive performance (sensitivity, 74%; specificity, 45%). The Kaplan-Meier curve showed that this value was able to identify patients who had the composite event during follow-up (figure 3). This association was confirmed to be independent of other morphofunctional CMR variables with prognostic value (LVEF and EDVI) through Cox regression (figure 4), and a GCS < –8.2% protected against events during follow-up (OR = 0.44; 95%CI, 0.20-0.95; P = .037).

Figure 3.

Kaplan-Meier curve for composite event-free survival. A GCS value of –8.2% was able to discriminate patients who had heart failure, implantable cardioverter-defibrillator implantation in secondary prevention, or death during follow-up in the univariate analysis. GCS, global circumferential strain.

(0.09MB).

Figure 4.

Cox regression curve for composite event-free survival. A GCS value of –8.2% was an independent predictor of the occurrence of heart failure, implantable cardioverter-defibrillator implantation, or death during follow-up, after adjustment of the survival curves for other morphofunctional CMR variables with prognostic relevance (LVEF and EDVI). CMR, cardiac magnetic resonance; EDVI, end-diastolic volume index; GCS, global circumferential strain; LVEF, left ventricular ejection fraction.

(0.07MB).

First, despite the inherent limitation of a retrospective study in terms of the identification of events, their incidence was similar to that of other prospective studies.1,9 In addition, the small number of events only allowed the introduction of imaging variables in the multivariate analysis. Although this could limit the prognostic value of FT-derived strain from a more general context, it did allow demonstration of its superiority to other morphological predictors such as LVEF. Furthermore, our population had severe systolic dysfunction at diagnosis, which is why our findings might not be applicable to patients with less advanced disease. Finally, consideration is required of FT-related limitations. Compared with echocardiographic studies, FT has lower temporal and spatial resolution, although it does not suffer from acoustic window problems. In this regard, a standard SSFP sequence was used, with a temporal resolution of 20 images/cycle, reaching the minimum recommended for clinical studies of 45ms/phase.18 Although a higher temporal resolution could optimize the evaluation of myocardial strain, our work shows that, even with this resolution, common in clinical studies and used previously,10 FT has prognostic utility in NIDCM. In addition, although the GCS values are more reproducible, there is variability among the different software programs used for FT,19 which is why the reported cutoff points are not applicable to other tools.