We retrospectively screened 217 patients (185 at Osaka University Hospital and 32 at the University of Tsukuba Hospital) with DCM and LVEF ≤ 35%, without a history of open-heart surgery, including mitral valve repair and LV reduction surgery, or ischemic heart disease. These patients were hospitalized for HF at either hospital from January 2009 to December 2017. DCM was diagnosed according to the World Health Organization/International Society and Federation of Cardiology criteria [14] as cardiomyopathy with LV dilatation and a reduced EF without ischemic heart disease, valvular heart disease, or secondary cardiomyopathy by several examinations, including coronary angiography and endomyocardial biopsy. Cardiac amyloidosis was ruled out in all patients by endocardial biopsy or clinical examinations including scintigraphy, wall thickness assessments, and laboratory data. We excluded 44 patients because of poor images for acquiring the speckle tracking data. Furthermore, 34 patients were excluded due to a scheduled hospitalization to implant a left ventricular assist device (LVAD). Therefore, 139 patients were included in this study. This study complies with the principles of the Declaration of Helsinki. The study protocol was approved by the ethics committees of each hospital (17371-4 and H30-053).

Transthoracic echocardiographic data including LS were collected in stable conditions after admission. Echocardiography was performed by experienced sonographers and doctors using a commercially available ultrasound machine (EPIQ and iE33, Philips Healthcare, Andover, MA, USA; Vivid7 and Vivid E9, GE Healthcare, Milwaukee, WI, USA; Aplio Artida, Canon Medical Systems, Otawara, Japan). We obtained the measurements and recordings based on the American Society of Echocardiography recommendations [15]. The LVEF and left atrial volume index were calculated using the biplane disk summation method with two-dimensional images and were indexed to the body surface area. Based on a previous report, we calculated the wall stress on the left ventricle at diastole and systole [16]. The height and body weight were also obtained on echocardiography. In addition, we calculated the body mass index.

Peak systolic LS measurements were obtained from the gray-scale images, recorded in the apical four- and two-chambers and long-axis views. The frame rate was maintained at > 50 frames/s. After gathering all DICOM data at Osaka University Hospital, we analyzed the LV strain offline using the speckle tracking software (TOMTEC, TomTec Imaging Systems GmbH, Munich, Germany). Good image quality was defined as a clear detection of the endocardial border throughout the cardiac cycle, and the regions of interest at the apex and annulus were ensured. Following the manual definition of the LV endocardial border, the endocardium was automatically tracked throughout the cardiac cycle. The software algorithm automatically divided the LV apical view into six segments for the speckle tracking throughout the cardiac cycle. We obtained the GLS by averaging all the segmental LS values from the aforementioned chambers and long-axis views. LS values for the basal, mid, and apical LV segments (six each) were averaged to obtain the regional LS values (basal, mid, and apical, respectively). The relative apical LS index (RapLSI) is one of the indices used for evaluating the regional abnormality of the myocardium. It is calculated by dividing the apical LS by the sum of the basal and mid-LS values [12]. In this study, we divided the patients into three groups according to RapLSI. When apical LS was more than 50% higher (relatively-impaired) or lower (relatively preserved) than the average of LS in mid and basal segments, RapLSI becomes < 0.25 (Apical relatively-impaired group) or > 0.75 (Apical relatively preserved group). The patients with RapLSI from 0.25 to 0.75 were grouped as a Scattered/Homogenously impaired group. The inter- and intra-observer intraclass correlation coefficients for obtaining the RapLSI were determined for a subset of 20 patients. The estimated inter- and intra-observer intraclass correlation coefficients were 0.931 [0.811–0.974] and 0.975 [0.938–0.990], respectively.

We obtained data on the patient characteristics and laboratory tests, including the brain natriuretic peptide (BNP) levels on admission. HF duration was defined as the duration from the initial event that requires hospitalization for decompensated HF to indexed hospitalization by asking the patient and/or searching hospital records. Medication data were collected at the discharge of indexed hospitalization, and optimal medical therapy (OMT) scores were calculated according to the previous report [17]. Details on all cardiac deaths, LVAD implantation, and HF-associated re-hospitalization after the discharge of indexed hospitalization were collected from the medical records, and the first event was used for survival analyses. One year after discharge, data on LVEF and beta-blocker use (carvedilol equivalent dose) were collected to assess LVRR. LVRR was defined as achieving ≥ 10% unit increase in the LVEF and ≥ 10% decrease in the LV end-diastolic dimension [18]. At the follow-up, the patients on LVAD support were classified as having non-reverse remodeling.

Continuous data are expressed as mean and standard deviation for normally distributed variables and as median and interquartile range for non‐normally. We compared the categorical variables using the Chi-square (χ2) test. The continuous variables were compared using a one-way analysis of variance with Bonferroni correction or the Steel–Dwass test for multiple unadjusted comparisons, and when appropriate, following an assessment of the normal distribution. Cochran–Mantel–Haensze test or two-way ANOVA was used to adjust for differences between institutions, as appropriate. The correlations among echocardiographic and clinical parameters were evaluated by the Pearson correlation coefficient. We conducted the Kaplan–Meier method to test for differences in the event-free rate between groups using the log-rank test. We used the Cox proportional-hazards model to adjust for the effects of differences in the baseline characteristics or pertinent covariates on the outcomes. Statistical comparisons were performed using the JMP, version 13 (SAS Institute Inc., North Carolina). All analyses were two-sided, and the significance was set at p < 0.05.