The review board of the First Affiliated Hospital of Soochow University approved this study, which adhered to all principles of the revised Declaration of Helsinki. Written informed consent (for the planned procedures and use of medical data) was obtained from all patients as dictated by the guidelines of the Chinese National Ethics Regulation Committee. All participants and their immediate relatives, caregivers, or legal guardians were told that they could withdraw consent at any time.

Dilated cardiomyopathy patients treated between January 2013 and June 2016 were enrolled. DCM was diagnosed when the ejection fraction was < 0.45 and/or a fractional shortening < 25% was apparent, combined with a left ventricular end-diastolic dimension (LVEDD) > 112% of the predicted value after correction for age and body surface area (Elliott, 2000). Exclusion criteria were: systemic hypertension (>160/100 mmHg); coronary artery disease (>50% in one or more major branches); chronic alcohol consumption (>40 g/day in females; >80 g/day in males); any systemic disease known to cause DCM; any pericardial disease; congenital heart disease; and/or a cor pulmonale (Elliott, 2000). A total of 100 DCM patients (65 males and 35 females) were finally enrolled (Figure 1A). Heart failure (HF) classification followed the recommendations of the American College of Cardiology Foundation, and the criteria of the American Heart Association were used to diagnose HF stage (Yancy et al., 2013).

A total of 202 healthy controls who underwent physical examinations in our hospital served as normal controls (males, 86; females, 116, age 54 ± 13.2 years). All controls were confirmed to not have DCM, atrial fibrillation, atrial flutter, pacemakers, sick sinus syndrome, an atrioventricular block, hypertension, coronary heart disease, even limited cardiomyopathy, pulmonary heart disease, myocarditis, valvular heart disease, congenital heart disease, any neuromuscular disease, any liver or kidney dysfunction, and diabetes.

As we reported previously (Figure 1B) (Zou et al., 2016), a PRSA algorithm detects and quantifies quasi-periodic oscillations masked by non-periodic components, artifacts, and ectopic beats (Kantelhardt et al., 2007). We used this algorithm to analyze 24-h ambulatory electrocardiograms. The DC and AC were computed at a timescale (T) of unity and a wavelet scale (s) of 2. Abnormal RR intervals (intervals varying by >5% from previous RR intervals (in terms of atrial fibrillation, premature atrial or ventricular contraction, and/or ventricular tachycardia) were removed to minimize errors caused by artifacts and variations in ectopic rhythm. Briefly, in step 1 (definition of decelerating and accelerating anchors), heartbeat intervals longer than those of preceding intervals were considered to reflect deceleration and intervals shorter than the preceding intervals were defined as accelerating anchors. In step 2 (definition of cardiac electrical segments), segmental interval data surrounding the decelerating and accelerating anchors were evaluated. In step 3 (phase rectification), the electrical cardiac segments described above were aligned to the decelerating and accelerating anchors. In step 4 (signal averaging), phase-rectified signal averages [X(i) values] were obtained by averaging signals within aligned cardiac electrical segments. In step 5, deceleration and acceleration capacities were quantified using the following formula: DC (AC) = [X(0)+X(1)–X(–1)–X(–2)]/4 (Bauer et al., 2006; Hu et al., 2016; Zou et al., 2016).

As previously described in detail (Zou et al., 2016), transthoracic echocardiographic examinations were performed using a Sonos 5500 Ultrasound machine (Philips, Best, Netherlands) fitted with a 2.5-Hz transducer. Parameters measured using the M-mode technique included the left ventricular end-diastolic diameter (LVEDD, normal range 35–56 mm), the left ventricular end-systolic diameter (LVESD, normal range: 20–40 mm), and the left atrial diameter (LAD, normal range 27–40 mm). The left ventricular ejection fraction (LVEF, normal value: >0.45) was measured using the Simpson biplane method. All echocardiographic examinations were performed and data were analyzed by the same experienced echocardiographer who was blinded to clinical data and patient group. Any uncertainties in echocardiographic data were resolved by discussion among senior technicians in the department of echocardiography.

The distributions of all variables were first assessed by drawing histograms; correlations between variables were then calculated using scatter plots and Pearson’s correlation coefficient analysis. All analyses were performed separately for either gender.

We used three progressive analysis strategies to evaluate associations between AC, DC, and DCM. First, we compared the AC, DC, and echocardiographic indices between DCM patients and healthy controls. The paired Student’s t-test was used for comparisons within groups and the non-paired Student’s t-test for comparisons between groups; all tests were two-sided. Next, the associations between the AC/DC and echocardiographic indices and age were subjected to univariate linear logistic regression, and factors associated with P-values < 0.1 underwent further multivariable linear logistic regression. Pearson’s correlation revealed a close link between AC and DC, and we thus performed linear logistic regression using either AC or DC as a dependent variable. Third, we performed Kaplan–Meier analyses with log-rank testing to calculate survival rates over the 60 months of follow-up and to compare survival rates between groups. The LAD, DC, and AC cutoffs were selected by reference to the areas under the receiver operating characteristic (ROC) curves (AUCs). P-values < 0.05 were deemed to reflect statistical significance. All statistical analyses were performed with the aid of SPSS software version 17.0 (SPSS Inc., Chicago, IL, United States).

In males, the average AC, LAD, LVEDD, and LVESD were significantly higher in DCM patients than controls; conversely, the average DC and LVEF were significantly lower than in controls (Table 1). In females, DCM patients were significant older than controls and, as with males, the average AC, LAD, LVEDD, and LVESD were significantly higher in DCM patients than controls; conversely, the average DC and LVEF were significantly lower in DCM patients (Table 1). In summary, the echocardiographic data confirmed the pathophysiological changes associated with DCM; both the AC and DC were significantly affected.

To explore correlations between AC, DC, age, and echocardiographic parameters, we drew scatter plots and calculated Pearson’s correlation coefficients. AC was positively and DC was negatively associated with age in both male and female controls (Figures 2A,B; age column). DCM greatly affected this pattern; in males, DCM abolished the association between AC, DC, and age (Figure 2A; age column); the LVESD was now significantly (negatively) associated with age (Figure 2A; age column). In females, DCM abolished the association between AC, DC, and age, but no new association emerged (Figure 2B; age column).

The AC displayed a very precise negative correlation with the DC in both males and females (Figures 2A,B). Healthy control males exhibited a significant negative association between the AC and LVEF (Figure 2A; AC column). Conversely, no significant association between AC and any echocardiographic parameter was observed in female controls (Figure 2B; AC column). DCM dramatically changed the above associations in males; significant positive associations between AC, LVESD, and LAD were now observed; and the negative slope of the LVEF/AC relationship fell from -0.30 to -0.44 (Figure 2A; AC column). Interestingly, no similar changes were observed in females (Figure 2B; AC column). In conclusion, Pearson’s correlation coefficient analysis revealed extensive associations between the AC and echocardiographic changes in males with DCM.

In male controls, DC was significantly (positively) associated with the LVEF (Figure 2A; DC column). Conversely, no significant association was observed in female controls (Figure 2B; DC column). Again, DCM changed the associations between DC and echocardiographic parameters. DC now became negatively associated with the LAD and LVESD; and the slope of the (positive) association between LVEF and DC increased from 0.27 to 0.41 (Figure 2A; DC column). Interestingly, no similar changes were observed in females (Figure 2B; DC column).

We used a univariate linear logistic regression to further assess the association between AC and echocardiographic parameters. Similar to what was noted on Pearson’s correlation coefficient analysis, in male controls, AC increased with age with a B value of 0.074 (95% CI, 0.036 ∼ 0.113; P < 0.001) and LVEF associated with AC increased with a B value of -8.632 (95% CI, -14.666 ∼ -2.598; P = 0.006) (Table 2). In male DCM patients, LVESD was positively associated with the AC with a B value of 0.064 (95% CI, 0.008 ∼ 0.121; P = 0.026); the association of LVEF with AC increased with a B value of -14.791 (95% CI, -22.453 ∼-7.129; P < 0.001); and the LAD was also positively associated with the AC with a B value of 0.107 (95% CI, 0.053 ∼ 0.161; P < 0.001) (Table 2). To adjust for possible interactions among variables, a multivariable linear logistic regression was next performed using variables associated with P-values < 0.1 on both univariate regression and Pearson’s correlation coefficient analysis. As shown in Table 2, age was independently (positively) associated with AC in male controls; and the LAD and LVEF were independently (positively and negatively, respectively) associated with AC in male DCM patients.

In females, only age was significantly associated with AC on the univariate linear logistic regression analysis, with a B-value of 0.050 (95% CI, 0.028 ∼ 0.072; P < 0.001); a multivariable linear logistic regression analysis was thus not performed (Table 2).

In summary, age was an independent risk factor associated with AC in both male and female controls, and the LAD and LVEF were independent risk factors associated with AC in male DCM patients.

Univariate linear logistic regression analyses revealed that in male controls, age was negatively and LVEF was positively associated with DC with B values of -0.05 (95% CI, -0.090 ∼-0.018; P = 0.003) and 7.063 (95% CI, 1.587 ∼ -12.539; P = 0.012) respectively (Table 3); in DCM patients, the LAD and LVESD were negatively associated with DC with B values of -0.102 (95% CI, -0.154 ∼-0.051; P < 0.001) and -0.058 (95% CI, -0.112 ∼-0.004; P = 0.036) respectively; and LVEF was associated with DC with a B value of 13.342 (95% CI, 5.965 ∼ 20.720; P = 0.001) (Table 3). Multivariable linear logistic regression revealed that only age was independently associated with DC in healthy controls, with a small change in the B value (Table 3); the LAD and LVEF were both independently associated with DC in DCM patients with B values of -0.075 (95% CI, -0.130 ∼-0.020; P = 0.008) and 8.428 (95% CI, 0.194 ∼ 16.662; P = 0.045), respectively (Table 4).

In females, only age was significantly associated with DC on the univariate linear logistic regression analysis with a B value of -0.04 (95% CI, -0.055 ∼-0.015; P = 0.001); a multivariable linear logistic regression analysis was thus not performed (Table 3).

In summary, age was an independent risk factor for DC in both male and female controls, and the LAD and LVEF were independent risk factors for DC in male DCM patients.

To further explore whether the AC and DC differed between DCM patients who survived and died, the AC, DC, echocardiographic data, age, and stages of HF were compared between subgroups. As showed in Table 4, 65 males and 35 females with DCM were followed up; 20 males and 5 females died during the 60 months follow up. Of males, the average follow up time of survived and died groups were 29.2 ± 15.9 and 29.5 ± 14.6 months, respectively; Of females, the follow up time of survived and died groups were 33.2 ± 15.0 and 28.2 ± 15.4 months, respectively (Table 4). In males, the LAD and AC were significantly higher and the DC was significantly lower in those who died. In females, only the LVEDD differed significantly between those who died and survived, being significantly higher in the former cases (Table 4). In summary, in males, the absolute values of both AC and DC decreased and that of LAD increased significantly in those who died; in females, only the LVEDD was significantly higher in those who died.

We found that AC, DC, and the LAD could predict the prognoses and survival of male DCM patients; we next calculated cumulative survival rates using appropriate cutoffs of these parameters. We first plotted ROC curves using the AC, DC, and LAD values of individual patients; the AUCs of AC, DC, and LAD were 0.647 (95%CI, 0.514 ∼ 0.779), 0.359 (95%CI, 0.225 ∼ 0.493), and 0.754 (0.637 ∼ 0.871), respectively (Figure 3A). The cutoffs for AC, DC, and the LAD were -4.75, 4.72, and 46.5 mm, respectively (Figure 3A). We next calculated cumulative survival rates using these cutoffs. During 60 months of follow-up, we counted deaths in subgroups formed using the cutoffs LAD ≥6.5 mm, AC ≥ –4.75 ms, and DC ≥ 4.72 ms; the numbers were 17, 16, and 15, respectively. Kaplan–Meier analysis revealed that DCM patients with greater LADs (log-rank Chi-square = 11.1, P = 0.001), a higher AC (log-rank Chi-square = 6.8, P = 0.009), and/or a lower DC (log-rank Chi-square = 9.1, P = 0.003) were at higher risk of cardiac death (Figure 3).

Our number of female patients was small and we could not identify any parameter predicting survival; we thus did not perform a Kaplan–Meier analysis of female DCM patients.