We designed this study as a retrospective analysis. Therefore, we reviewed medical records of 442 patients with heart failure who had been initially diagnosed, evaluated, and followed at Sanggye Paik Hospital, Inje University College of Medicine between March 2013 and February 2018. The diagnosis of DCM was established on the basis of the following criteria: (i) presence of LV dilation (LV end-diastolic diameter ≥ 55 mm); (ii) presence of reduced LV ejection fraction (EF) (all ≤ 45%); (iii) coronary angiographic evidence of absence of coronary artery disease defined as > 50% stenosis of a major epicardial vessel or a history of myocardial infarction; and (iv) absence of cardiac muscle disease secondary to any known systemic disease.7) As shown in Figure 1, patients who received cardiac resynchronization therapy (1), pacemaker (17), implantable cardioverter defibrillator (3), or obtained an LV assist device were excluded. We also excluded patients with AF (118), other major cardiac arrhythmias, pulmonary thromboembolism, idiopathic pulmonary hypertension, hyperthyroidism, more than moderate organic valvular disease except functional mitral regurgitation (MR), and stage 4 or more advanced renal failure or end-stage renal disease requiring renal replacement therapy. We excluded 67 patients who had difficulty in analysis due to poor image on echocardiography or follow-up loss. In total, 160 patients were enrolled in this study. The subjects were in clinically stable condition at the time of enrollment in this study and were undergoing optimal and maximally-tolerated pharmacological therapy. This study complied with the Declaration of Helsinki, and the local ethics committee approved the research protocol. The requirement of informed consent was waived by the committee.

LVRR was defined as the combined presence of: 1) an increase in LVEF of at least 10% points or a follow-up LVEF ≥ 50% (in patients with an LVEF of 40% to 45% at enrolment); and 2) a decrease in indexed left ventricular end-diastolic diameter (LVEDDI) of at least 10% or an LVEDDI ≤ 33 mm/m².2)3)

A standardized complete echocardiographic examination was performed at baseline and during follow-up according to the standard method outlined by the American Society of Echocardiography.8) 2D echocardiographic images were obtained using VIVID E9 and E95 (General Electric Medical Systems, Horten, Norway). All conventional echocardiographic parameters were measured and averaged over three cardiac cycles. All echocardiographic data were reanalyzed with commercially available software (EchoPAC PC software, GE Medical Systems, Horten, Norway).

Significant functional MR should fulfill all of the following three criteria: 1) normal mitral valve leaflets and chords, 2) regional or global wall motion abnormalities of LV with tethering of the leaflet, and 3) MR of EROA > 0.1 cm².9)

2D-speckle tracking measurements of LV GLS were performed in the 2-, 3-, and 4-chamber apical views and aortic valve closure was used for timing of end-systole.10) The endocardial border was traced manually at end-systole. Tracking was adjusted to include the entire myocardial wall from the endocardium to the myoepicardial border. All LV GLS measurements were performed by a single investigator (J. H. Park), with all the images reviewed and validated by a second reader. Reproducibility was performed on a single set of recordings. Variability in the measurement of LV GLS was evaluated in 20 randomly selected patients. To determine intra-observer and inter-observer variability, the investigator measured GLS for each of the selected patients again 2 weeks later. The coefficients of variation of intra- and inter-observer variability for GLS measurements were less than 5%.

Conventional echocardiographic parameters indicating right ventricular (RV) systolic function were measured in a standard manner according to the guidelines of Rudski et al.11) The RV fractional area change was calculated using the percent area change of the end-diastolic and end-systolic areas of the RV in the apical four-chamber views. Tricuspid annular plane systolic excursion was measured as the contraction distance of the RV tricuspid annular plane along its longitudinal plane during systole.

There were two different temporal perspectives in this study. One was to see whether LVRR occurred and the other was to see whether an adverse clinical event occurred. First, echocardiographic evaluation was scheduled to be performed annually after diagnosis of DCM, and the mean interval of echocardiography in these patients was 10.7 ± 7.3 months. The interval of echocardiography in patients without LVRR was shorter than in patients with LVRR (9.5 ± 6.4 vs. 13.7 ± 8.8 months, p = 0.027). We analyzed the echocardiographic data at the time DCM was initially diagnosed. Additionally, follow-up echocardiographic data were collected. For patients with LVRR, the results of echocardiography at the time of recovery were analyzed, and for patients who were not recovered, the results of the last performed tests were analyzed. Second, we collected clinical follow-up data for occurrence of adverse clinical events. We recorded the date of last follow-up in patients without adverse events and recorded the date of occurrence of events in patients with events.

Adverse clinical events were defined as all-cause death, death from, cardiac transplantation or hospitalization for deteriorating heart failure, and major ventricular arrhythmias or aborted sudden cardiac death. Major ventricular arrhythmias were considered as ventricular fibrillation/flutter or a sustained ventricular tachycardia (hemodynamically unstable or lasting more than 30 seconds) or appropriate ICD intervention. Long-term follow-up was conducted for 37.3 ± 21.7 months (patients with LVRR vs. without LVRR, 38.1 ± 27.8 vs. 37.0 ± 19.4 months, p = 0.855).

Continuous variables with a normal distribution were expressed as mean ± standard deviation or percentages. The parameters of the two subgroups, patients with and without LVRR, were compared using unpaired Student's t-test or Mann-Whitney U test for continuous variables. The χ² tests (or Fisher's exact tests when any expected count was < 5 for a 2 × 2 table) was used for categorical variables. To evaluate change from baseline, a paired Student's t-test was used. Pearson's correlation was used to calculate the association between baseline LV GLS and follow-up LVEF. We performed a time-to-event analysis using a univariate Cox proportional-hazards model for all patients to determine the independent factors influencing LVRR. Variables with a univariate value of p < 0.05 were incorporated into the stepwise selection, while age, sex, medications, and LV dimensions expected to be highly associated with LVRR were forced into the multivariate analysis regardless of its association on univariate analysis. The receiver-operating characteristic (ROC) curve was constructed to determine the optimal cut-off values of LV GLS for LVRR. We then divided the subjects into two groups based on the cut-off value. Kaplan-Meier curve analysis was conducted to assess cardiovascular event-free survival during the follow-up period and compared it by using log-rank test. A probability value of p < 0.05 was considered significant. All statistical analyses were performed using SPSS software package (version 25, IBM SPSS Statistics, Chicago, IL, USA).

A total of 160 patients who fulfilled all inclusion and exclusion criteria were enrolled. The mean age was 64.5 years, and 67.5% were male. Forty-five patients (28%) experienced LVRR within 14.7 ± 10.0 months after medical therapy. Clinical parameters are summarized in Table 1. As compared with patients who did not develop LVRR, patents with LVRR had fewer left bundle branch blocks (LBBB) and more dyslipidemia, and more patients used statins. However, no significant differences between the subgroups were noted in terms of age, gender distribution, body mass index, or blood pressures. A comparison of echocardiographic data between patients with LVRR and those without LVRR is shown in Table 2. The initial LVEF of patients who recovered LV function was 26.1 ± 7.9%, which was not significantly different from the value of 27.1 ± 7.4% (p = 0.49) for those who did not recover LV function. However, patients with LVRR had a larger LV, smaller left atrium (LA), and more frequent significant functional MR than those who did not have LVRR. The E/e′ ratio was higher and the values of LV GLS were lower in patients without LVRR.

Follow-up LV functional echocardiographic parameters of all patients were improved when compared with baseline data except for the LVEF (27.8 ± 7.4%, ΔLVEF of 0.7%, p = 0.126) of patients without LVRR (Table 3). Patients with LVRR had a final LVEF of 49.4 ± 9.5% (ΔLVEF of 23.9%, p < 0.001) and had a significant decrease in LVEDDI (29.8 ± 5.2 mm/m²; ΔLVEDDI of 7.5 mm/m², p < 0.001) as well as an increase in the absolute value of LV GLS (16.9 ± 2.7%; ΔLV GLS of 5.1%, p < 0.001). The mean differences of the LV GLS between baseline and follow-up in patients were 0.5 ± 1.3% for patients without LVRR (Figure 2A) and 5.1 ± 2.9% for patients with LVRR (Figure 2B).

Baseline LV GLS had a significant negative correlation between follow-up LV dimensions and positive correlation with follow-up LVEF and Δ EF (i.e., difference between baseline and follow-up LVEF) (Table 4). Although baseline LV GLS was widely distributed at the follow-up LVEF levels, there was a moderate and highly significant correlation between baseline LV GLS and follow-up LVEF (r = 0.717; p < 0.001) (Figure 3).

During follow-up, 25 patients (15.6%) had serious adverse events including 3 cardiac deaths and 1 cardiac transplantation. Among patients without LVRR, 13 patients were admitted for worsening heart failure, and 4 patients experienced major ventricular arrhythmia or aborted sudden cardiac death. In the group that recovered LV function, there were 2 patients who were admitted for worsening heart failure and 2 patients who experienced major arrhythmia, but there were no deaths (Table 5). There was no significant difference in the incidence of adverse events between the two groups (p = 0.142).

The hazard ratio (HR) of each variable is shown in Table 6. By univariate analysis using a Cox proportional hazard model, no LBBB, higher value of LV GLS (−%), lower LA volume index, and E/e′ ratio were significantly associated with LVRR. Using multivariate Cox analysis, LV GLS (HR: 1.474, p = 0.001) was an independent predictor of LVRR in DCM patients with sinus rhythm. Among the medications, ACE-I/ARB was not a valid variable for LVRR in univariate analysis (HR: 1.686, p = 0.332), but it was a significant variable in multivariate analysis (HR: 4.031, p = 0.029).

The survival curves of the study population classified according to LVRR are shown in Figure 4A. DCM patients with LVRR had a significantly better long-term prognosis with respect to the others (Log-rank test, p = 0.029).

By analyzing the ROC curve, LV GLS ≤ −10% was selected as the best cut-off value to discriminate between patients with and without LVRR with 98% sensitivity and 63% specificity (area under the curve = 0.854, p < 0.001). However, patients with LV GLS ≤ −10% was not lower in occurrence of adverse event compared to those with LV GLS > −10% (Log-rank test, p = 0.086) when we divided the subjects into two groups based on the cut-off value for LV GLS of −10% (Figure 4B).

The present study has several limitations. First, this was a single-center observational study of subjects who were registered in our institute and was performed using a single vendor-specific software. Therefore, our study group may not represent all patients with DCM, and our results may not apply to other equipment. Second, only 61.9% of patients enrolled in our study used beta-blockers, which was significantly lower than that reported in previous studies. At the time of enrollment in this study, it is possible that many of the patients with DCM had low blood pressure. And this reason forced them to use a combination of low dose beta blocker and ivabradine (25 patients without LVRR vs. 9 with LVRR, 21.7% vs. 20.0%) but often use ivabradine alone for patients who cannot tolerate beta blocker (19 patients without LVRR vs. 8 with LVRR, 16.5% vs. 17.8%). However, there was no significant difference between the two groups in the frequency of ivabradine alone or beta blocker combination (p = 0.785). Furthermore, Cox regression multivariate analysis revealed that ACE-i / ARB, rather than beta blockers or ivabradine, had a significant effect on the occurrence of LVRR in this study population. Third, the number of patients with cardiovascular events was relatively small, although the mean follow-up period of about 40 months was adequate. This is likely because we excluded subjects with AF and enrolled a relatively small number of patients with significant MR; AF and MR are considered important prognostic factors for clinical outcomes in DCM patients. We think this is one of the reasons why we failed to demonstrate the differences of event-free survival rate between groups based on the cut-off value of LV GLS for LVRR. Finally, because LV GLS values are highly correlated with LVEF, the difference in LV GLS values between patients who have already reduced LVEF, such as DCM patients, is not numerically large. We suggest that these results disprove that LV GLS is a more objective indicator than LVEF in patients with DCM; however, it would be inappropriate to predict the prognosis or a recovery of LV function in DCM patients based on only the LV GLS value. Nevertheless, in this study we showed that recovery of LV function occurred more frequently when DCM patients had better LV GLS values, even in patients with similar LVEF, which was significantly reduced at the time of diagnosis.

In our single-center study, we demonstrated that LV GLS was an independent predictor for LVRR, and the optimal cut-off point of LV GLS for LVRR was −10% in DCM patients with sinus rhythm. Patients with LVRR showed a significantly better long-term prognosis than those without LVRR. There was a significant correlation between baseline LV GLS and follow-up LVEF in this population. Therefore, we suggest LV GLS should be measured in DCM patients with sinus rhythm at the time of initial diagnosis. To validate the role of LV GLS in LVRR and prognosis in DCM patients with sinus rhythm, however, future studies must include long-term data compiled from a larger patient population.