*2.3. LVRR Definition and Study Outcome Measures*

LVRR was defined as an increase in the LVEF ≥ 10% (or LVEF > 50%) associated with a decrease <sup>≥</sup> 10% in indexed left ventricular end-diastolic diameter (LVEDDI) or (LVEDDI <sup>≤</sup> 33 mm/m2) at 24-month follow-up after enrolment, as previously described [5].

The main outcome measure was considered a composite of all-cause mortality, heart transplantation (HTx), and ventricular assist device (VAD) as destination therapy. A composite of sudden cardiac death (SCD) or major ventricular arrhythmias (MVA) was considered as the secondary outcome measure.

Specifically, MVA was defined as sustained ventricular tachycardia, ventricular fibrillation/flutter, or appropriate intervention of an ICD. SCD was defined as a death occurred within 1 h from the symptom's onset, or as a death occurred during sleep in clinically stable patients with New York Heart Association (NYHA) class I–III.

To evaluate the association with the study outcome measures, the population was stratified into four groups, based on sex and the occurrence of LVRR.

Outcomes were investigated directly from the patient during the follow-up visit, medical records from the referral hospital or by telephone interview with the patient, relatives, or the general practitioner.

#### *2.4. Statistical Analysis*

Variables were expressed as mean and standard deviation, median and interquartile range (IQR), or counts and percentage, as appropriate. Comparisons between groups were made by the analysis of variance (ANOVA) test on continuous variables using the Brown-Forsythe statistic when the assumption of equal variances did not hold, or the nonparametric Mann-Whitney test when necessary; the chi-square test or the Fisher's exact test were calculated for discrete variables.

Survival curves for the composite outcome measure of all-cause mortality/HTx/VAD were estimated and compared between groups by means of the Log-rank test. Cumulative incidence curves for the composite outcome measure of SCD/MVA were estimated and compared taking into account competing risks of death from other causes, and the appropriate statistical test suitable for competing risks was performed [14]. To investigate the impact of sex and LVRR on the outcomes, cause-specific multivariable Cox models were estimated from a list of candidate prognostic variables obtained from the univariable analyses (i.e., those with a *p*-value ≤ 0.1). For this analysis, the follow-up started after 24 months from enrolment, when the LVRR is considered to be completed [5]. Moreover, to further evaluate the relationship between sex and LVRR, a Markov illness-death model with all-cause mortality/HTx/VAD as absorbing state and the risk of LVRR as an intermediate state was estimated. The model consists of three discrete health states (i.e., alive without LVRR; alive with LVRR; dead or HTx or VAD) and a transition probability matrix (P) is calculated between states (see Supplementary Figure S1 for schematic representation). Specifically, a multi-state model fitting a Cox-type regression for each transition was used to estimate transition-specific hazard ratio (HR) for Sex. In this case, the follow-up started at the time of enrolment and this model was adjusted for a list of candidate variables significantly different at the univariable analysis of the multi-state model. The IBM-SPSS (New York, NY, USA) statistical software version 19 was used for descriptive analyses; the software R (R Foundation for Statistical Computing, Vienna, Austria. https://www.r-project.org/) was used for the cumulative incidence curves estimation (library "cmprsk"), to test the proportional hazards assumption for the Cox model and for the multi-state model (packages "ggplot2", "survival" and "mstate") [15].

#### **3. Results**

A total cohort of 605 consecutive DCM patients with available data at a median follow-up of 24 (IQR 20–26) months was analysed (Figure 1). The main characteristics of the population at 24-month follow-up evaluation are summarized in Table 1. Patients were predominantly males (73% *n* = 440), and males were slightly younger than females (47 ± 15 vs. 51 ± 14 years respectively, *p* = 0.007). Females

had a higher incidence of left bundle branch block (LBBB) compared to their male counterparts (34% vs. 25%, respectively, *p* = 0.02). All patients received optimal medical treatment, without differences between sexes.

**Figure 1.** Diagram of study population. Legend. F: Females; LVRR: Left Ventricular Reverse Remodelling; M: Males.

subsectionLeft Ventricular Reverse Remodeling

Overall, 30% of patients experienced a LVRR (*n* = 181), without significant differences between sexes: the cumulative incidence of LVRR at the 24 months evaluation was 33% in women vs. 29% in men (*p* = 0.26) (Figure 1). Indeed, the probability of undergoing LVRR was similar between men and women (Hazard Ratio for male sex [HR] 0.81, 95% Confidence Intervals [CI] 0.53–1.22, *p* = 0.31). Interestingly, at the 24-months evaluation, despite a comparable LVEF (40 ± 12% in women vs. 41 ± 11% in men, *p* = 0.32), women had a higher incidence of moderate to severe sex-specific LV dilation compared to men (59% vs. 28% respectively, *p* ≤ 0.001).
