Prognosis and diastolic dysfunction predictors in patients with heart failure and recovered ejection fraction

There is limited data on whether diastolic dysfunction in patients with heart failure (HF) and recovered ejection fraction (HFrecEF) is associated with worse prognosis. We retrospectively assessed 96 patients diagnosed with HFrecEF and created ROC curve of their diastolic function at the 1-year follow-up for the composite endpoint of cardiovascular death and HF readmission after the follow-up. Eligible patients were divided into two groups according to the cutoff value of E/e′ ratio (12.1) with the highest AUC (0.70). Kaplan–Meier analysis showed that HFrecEF with high E/e′ group had a significantly poorer prognosis than the low E/e′ group (log-rank, p = 0.01). Multivariate Cox regression analysis revealed that the high E/e′ group was significantly related to the composite endpoint (hazard ratio 5.45, 95% confidence interval [CI] 1.23–24.1). The independent predictors at discharge for high E/e′ ratio at the 1-year follow-up were older age and female sex after adjustment for covariates (odds ratio [OR] 1.07, 95% CI 1.01–1.13 and OR 4.70, 95% CI 1.08–20.5). In conclusion, HFrecEF with high E/e′ ratio might be associated with a poor prognosis. Older age and female sex were independent predictors for a sustained high E/e′ ratio in patients with HFrecEF.

www.nature.com/scientificreports/ Although, there is limited data regarding appropriate risk stratification and management in patients with HFrecEF, we hypothesized that diastolic function would be the prognostic indicator among HFrecEF patients as it is based on findings regarding the functional phenotypes of bioengineered cardiac tissue in hypoxia and reoxygenation. Both the systolic and relaxation functions of the cardiac tissue had deteriorated in the hypoxic condition, while in reoxygenation, the relaxation dysfunction remained, even after systolic function had fully recovered 9 . These results suggest that, in hypoxia, the relaxation dysfunction might be prolonged and the recovery or remission difficult to achieve when compared with systolic dysfunction. Relaxation dysfunction is one of the component of diastolic dysfunction 10 . Peak velocity of the early wave (E) to early diastole (e′) (E/e′) ratio and left atrial volume index (LAVI) assessed by echocardiography were correlated with LV filling pressure as well as the indexes for diagnosis of LV diastolic dysfunction 10,11 .
The aims of our study were to elucidate the predictors for sustained diastolic dysfunction at, as well as the long-term prognoses after, the 1-year follow-up.

Methods
Study population and endpoints. We retrospectively assessed consecutive patients who were hospitalized for HFrEF and discharge alive at the Tokyo Women's Medical University Hospital between July 2013 and October 2018. We followed-up and reassessed them via echocardiography at the 1-year follow-up and we confirmed that the patients were either diagnosed with HFrecEF or not. Inclusion criteria were follows: a diagnosis of HFrecEF at the 1-year follow-up; more than 1 year of follow-up; and obtaining the data describing the LVEF and E/e′ (septal e′) ratio (Fig. 1A,B).
Patients were diagnosed with HF using the Framingham HF diagnostic criteria 12 : HFrEF was defined as HF and an LVEF < 40% at discharge 13,14 ; HFrecEF was defined as an LVEF < 40% at discharge that improved to ≥ 40% at the 1-year follow-up, and this was based on the results of previous studies 3,7,14 . Exclusion criteria were as follows: a diagnosis of heart failure with unchanged ejection fraction (HFuncEF: LVEF < 40% at discharge and Changes in LVEF and E/e′ ratio in HFrecEF with low or high E/e′ ratio group. E/e′ peak velocity of the early wave (E) to early diastole (e′), HF heart failure, HFrecEF heart failure with recovered ejection fraction, HFrEF heart failure with reduced ejection fraction, HFuncEF heart failure with unchanged ejection fraction, ROC receiver operating characteristic.  18 . Anemia was defined as hemoglobin levels < 12.0 g/dL in women and < 13.0 g/dL in men 13 . The definition of ischemic cardiomyopathy was that left ventricular (LV) dysfunction due to severe ischemic heart disease including the myocardial infarction and angina pectoris according to Japanese Circulation Society 2018 guideline on diagnosis and treatment of cardiomyopathies.
Statistical analyses. The data are expressed as means and standard deviations, median values and interquartile range (IQR), or percentages, as appropriate. The Student's t-test was used to compare normally distributed continuous variables between the groups, while the Mann-Whitney U test was used for skewed continuous variables; Fischer's exact test was used to evaluate the categorical variables. A paired t-test was used to compare the LVEF or E/e′ ratio at discharge and the 1-year follow-up. Pearson product-moment correlation coefficient was used to identify the correlation between E/e′ ratio at the 1-year follow-up and LVEF changes from the discharge to the 1-year follow-up. We plotted ROC curves for the composite of CV death and readmission for HF using the E/e′, LAVI, TR Vmax, and e′ at the 1-year follow-up; additionally, we estimated the optimal cutoff value based on the Youden index. BNP levels were log-transformed. Logistic regression analysis was performed to identify independent factors at discharge related to high E/e′ ratio at the 1-year follow-up. Variables were considered clinically significant if they reached a significance level of p < 0.05 and were subsequently included in the multivariable logistic regression model. The Kaplan-Meier method and log-rank tests were used to compare the event-free survival ratios between the two groups during the follow-up period after the 1-year follow-up. Multivariable Cox regression analysis was performed to assess whether high E/e′ ratio at the 1-year follow-up was associated with the primary endpoint after adjusting for covariates. Due to small number of patients who experienced the composite endpoint, the multivariate analysis was performed by adjusting for age and sex only. Two-sided significance was set at p < 0.05. Furthermore, sensitivity analysis was performed to remove the influence of different definition of HFrecEF on the results. We applied another definition of HFrecEF for the sensitivity analysis. The definition was (1) decreased LVEF < 40% at baseline; (2) ≥ 10% absolute improvement in LVEF; and (3) a second measurement of LVEF ≥ 40%, according to the previous reports 19 . All statistical analyses were performed using R software (version 1.41.1; R Foundation for Statistical Computing, Vienna, Austria) 20 .

Results
The median follow-up duration was 537 (IQR 309-923) days after the 1-year follow-up. During the study period, 14 patients (15%) were readmitted for HF or died as a consequence of CV events; only one patient died from CV disease. The ROC curve for a composite of CV death and readmission of HF using the E/e′ ratio, LAVI, TR Vmax, and e′ at 1-year follow-up revealed cutoff values of 12 Fig. 2). There were no statistically significant differences regarding the AUC between the E/e′ ratio and LAVI at the 1-year follow-up (p = 0.76).
Patient characteristics at discharge. Table 1 shows the patient characteristics of the study population and comorbidities at discharge. Significant differences were evident between the two groups concerning age, sex, principal cause of HF due to ischemic cardiomyopathy, and sodium, hemoglobin, and BNP levels. Regarding echocardiography during the hospitalization, LVEF and E/e′ were higher in the HFrecEF with high E/e′ group than the HFrecEF with low E/e′ group. The estimated RVSP, E-wave deceleration time, e′, LAVI, LVMI, IVC diameter, and presence of atrial arrhythmias at the echocardiography were comparable between the two groups. Angiotensin-converting enzyme inhibitors or angiotensin receptor blocker inhibitors (ACEi/ARB) and betablockers were prescribed in approximately 90% of patients at discharge; statin was more often prescribed in the high E/e′ than low E/e′ group.

Patient characteristics and echocardiography data at the 1-year follow-up.
The median followup period between discharge and the 1-year follow-up was 365 (IQR 332-398) days. Although LVEF in both HFrecEF with low and high E/e′ ratio groups significantly improved from 30 ± 5.5 and 33 ± 4.1% at discharge to 49 ± 6.1% and 47 ± 5.8% at the 1-year (paired-t test: p < 0.001, respectively, Fig. 1D), E/e′ ratio in HFrecEF with high E/e′ group were unchanged from 18 ± 6.3 at discharge to 18 ± 5.4 at the 1-year follow-up (paired-t test: p = 0.68), while E/e′ ratio in HFrecEF with low E/e′ group improved from 15 ± 5.4 at discharge to 8.8 ± 1.9 at the 1-year follow-up (paired-t test: p < 0.001, Fig. 1D). Interestingly, there was inversely mild correlation between the E/e′ ratio at the 1-year follow-up and LVEF changes (%) from the discharge to the 1-year follow-up (r = − 0.25, p = 0.02, Supplemental Fig. 1). There were no significant differences in systolic blood pressure, diastolic blood pressure, heart rate, atrial arrhythmias at the 1-year follow-up, LVEF, LVDd, LVDs, LVMI, E-wave deceleration time, IVC diameter, the presence of atrial arrhythmias at the echocardiography, and the prescription rates of GDMT for HF between the two groups at the 1-year follow-up. Changes in heart rate from discharge to the 1-year follow-up did not differ between the two groups (low E/e′ group: − 2.4 ± 14 bpm vs. high E/e′ group: 0.8 ± 18 bpm; p = 0.33). Patients in the high E/e′ group had larger LADs and LAVIs, as well as a tendency to exhibit a higher RVSP and TR Vmax, compared with those in the low E/e′ group. Then, e′ was lower and BNP level was higher in HFrecEF with high E/e′ group compared with HFrecEF with low E/e′ group. There was statistically significant difference in the furosemide dose between the two groups. No patient had moderate to severe mitral valve regurgitation. The prescription rates for ACEi/ARB and β-blockers were compatible between discharge and the 1-year follow-up; however, that of aldosterone antagonists decreased from 72% at discharge to 58% at the 1-year follow-up (Table 2).
Prognosis. The Kaplan-Meier curve revealed that the rate of the composite endpoint was significantly higher in the high E/e′ group than the low E/e′ group (log-rank test, p = 0.01; Fig. 3). Furthermore, multivariate Cox regression analysis revealed that an E/e′ ratio ≥ 12.1 at the 1-year follow-up was associated with the composite endpoint after the 1-year follow-up, after adjusting for age and sex (hazard ratio [HR] 5.45, 95% CI 1.23-24.1; Table 3). This result was consistent with sensitivity analysis with other definition of HFrecEF (Supplemental Fig. 2). Meanwhile, when the study population was divided into two groups according to a cutoff of 49.4 mL/m 2 for LAVI, the Kaplan-Meier curve demonstrated that patients in the high LAVI group exhibited higher rate of the composite endpoint, compared with those in the low LAVI group (log-rank test, p = 0.02; Supplemental Fig. 3). However, the age and sex adjusted HR for the composite endpoint did not significantly differ between the two groups (HR 1.86, 95% CI 0.60-5.73).
Predicting high E/e′ ratio at the 1-year follow-up. Discharge parameters including age, female sex, ischemic cardiomyopathy, high BNP level, low sodium level, high LVEF, high E/e′ ratio, and a prescription for statins correlated with a high E/e′ ratio at the 1-year follow-up in the univariate logistic regression analyses. After   Table 4).

Discussion
This study investigated the predictors at discharge for high E/e′ ratio at the 1-year follow-up in patients with HFrecEF, as well as their prognosis. The principal findings were as follows: (1) patients with HFrecEF and high E/e′ ratio at the 1-year follow-up had a poor prognosis after the 1-year follow-up, and (2) older age and female sex at the hospitalization were associated with a high E/e′ ratio at the 1-year follow-up. Several studies have been conducted regarding the recovery of LVEF during the follow-up period 2-6 . The current study reveals that the proportion of HFrecEF at the 1-year follow-up in patients with HFrEF is consistent with previous reports 3,4 . Additionally, Bermejo et al. reported that the percentage of the implantable cardioverter defibrillator (ICD) implantation, ischemic etiology of HF, ACEi/ARB, and beta blocker in patients with HFrecEF as the predictors of LVEF improvement were 3%, 20%, 90%, and 78%, respectively 21 . Those percentages were similar to those of the current study; ICD implantation: 2%, ischemic etiology of HF: 19%, ACEi/ARB: 89%, and beta blocker: 96% (Table 1). However, despite the LVEF improving in patients with HFrEF, the predictors of remaining diastolic dysfunction and the relationship between prognosis and diastolic function at follow-up remain unknown. We thus focused on patients with HFrecEF and their diastolic function at follow-up. We selected the E/e′ ratio at the 1-year follow-up to classify the study population as the diastolic dysfunction parameter because this parameter demonstrated the highest AUC for the composite endpoint when compared with LAVI, e′, or TR Vmax, which were also related to the diastolic dysfunction. The E/e′ ratio is one of the parameters assessing left ventricular filling pressures, which can be assessed by combining the effects of the transmitral driving pressure and myocardial relaxation 22 . Patients in HFrecEF with high E/e′ ratio group had lower e′ (< 7 cm/s) compared with patients in HFrecEF with low E/e′ group, suggesting that patients in HFrecEF with high E/e′ ratio group had the relaxation dysfunction at the 1-year follow-up 10,11 . Furthermore, patients in the high E/e′ group had a larger LAVI (> 34 mL/mm 2 ) and higher TR Vmax than those in the low E/e′ group at the 1-year follow-up. It suggested that patients in the high E/e′ group exhibited diastolic dysfunction regardless of the improvement in LVEF, in reference to previous reports 10, 23 . Table 1 showed that HFrecEF with high E/e′ group had higher LVEF and smaller LV size than HFrecEF with low E/e′ group. HFrecEF with high E/e′ group had higher proportion of female patients compared with low E/e′ group, which might influence the LV size. Further, sex-specific differences in the distribution of LVEF was reported in nationwide register (N = 499,153), which demonstrated that mean LVEF was higher in women than men 24 . We speculated that sex differences between the two groups might affect the LVEF and LV size at the discharge. In addition, because Zhang et al. reported that the degree of LVEF improvement was associated with the prognosis 25 , we created ROC curve and compared the AUC of E/e′ ratio at the 1-year follow-up and LVEF changes from the discharge to the 1-year follow-up for the composite of CV death and rehospitalization for HF. Although there was no significant difference between them, AUC was higher in E/e′ ratio than LVEF change (E/e′ ratio: 0.70 vs. LVEF change: 0.61, p = 0.48, Supplemental Fig. 4).
In previous studies, patients with HFrecEF were shown to have a lower risk of mortality or hospitalization for HF [3][4][5] . One of the predictors of increased EF was female sex 3,6,7 . Interestingly, elderly women in the current study tended to exhibit diastolic dysfunction at the 1-year follow-up, regardless of recovered LVEF. This can Table 1. Patients' characteristics at discharge according to HFrecEF with low or high E/e′ ratio at the 1-year follow-up. ACEi angiotensin converting enzyme inhibitor, ARB angiotensin receptor blocker, OAC oral anticoagulants, BMI body mass index, BNP brain natriuretic peptide, BP blood pressure, bpm beats per minute, CABG coronary artery bypass grafting, CRP C-reactive protein, CRT cardiac resynchronization therapy, DPP4i dipeptidyl peptidase-4 inhibitor, E/e′ peak velocity of the early wave (E) to early diastole (e′) ratio, eGFR estimated glomerular filtration rate, HDL high density lipoprotein, HFrecEF heart failure with recovered ejection fraction, HT hypertension, ICD implantable cardioverter defibrillator, IDDM insulin dependent diabetes mellitus, IVC inferior vena cava, LAD left atrial dimension, LAVI left atrial volume index, LDL low density lipoprotein, LVEF left ventricular ejection fraction, LVDd left ventricular end-diastolic diameter, LVDs left ventricular end-systolic diameter, LVMI left ventricular mass index, OAC oral anticoagulants, PM pacemaker, RVSP right ventricular systolic pressure, SGLT2i sodium-glucose cotransporter 2 inhibitor, T-Chol total cholesterol, TR Vmax maximal tricuspid regurgitation velocity, WBC white blood cell.

All patients
HFrecEF with low E/e′ (< 12.  www.nature.com/scientificreports/ be explained by the effect of sex hormones as postmenopausal women with low levels of estrogen are prone to cardiac diastolic dysfunction through the suppression of sarcoplasmic reticulum Ca 2+ -ATPase (SERCA) activity 26,27 . Furthermore, menopause is associated with a reduction in cGMP-protein kinase G signaling by decreasing plasma natriuretic peptide levels [28][29][30] . These mechanisms may lead to CV events. Although Ghimire et al. reported that female patients exhibited lower mortality risk than men among HFrecEF patients 31 , the Table 3. Cox regression analysis for the composite of CV death and heart failure readmission after the 1-year follow-up. CI confidence interval, CV cardiovascular, E/e′ peak velocity of the early wave (E) to early diastole (e′) ratio, HFrecEF heart failure with recovered ejection fraction. www.nature.com/scientificreports/ impact of sex differences on the prognosis was not evaluated in the current study because of small sample size. HFrecEF with high E/e′ ratio could indicate a more severe comorbidity, potentially causing the poor prognosis. Regarding the possible reason for persistent diastolic dysfunction, the fabricated human cardiac tissue revealed systolic and relaxation dysfunction in a hypoxic environment 9 . After reoxidation, systolic function had completely improved; however, the relaxation dysfunction remained, with the mRNA expression of phospholamban upregulated. This suggests that compared with systolic dysfunction, it might be difficult to achieve normalization of the relaxation dysfunction in hypoxia. The myocardium in heart failure is constantly exposed to the hypoxic and normoxemic environment. In cases of acute decompensated HF, the myocardium may be exposed to the hypoxic environment as a result of decreasing oxygen supply due to pulmonary congestion-as well as increasing oxygen demand due to volume or pressure overload-regardless of the etiology of HF 32,33 . The microvascular density decreased in patients with chronic HF compared with healthy subjects 32,34,35 . Arnold et al. reported that the microvascular dysfunction in HF with preserved ejection fraction (HFpEF) was correlated with E/e′ ratio, which was related to the long-term prognosis 36 . On the other hand, the compensatory mechanisms in HF help to supply the oxygen and to suppress the oxygen demand as well as GDMT can improve the hypoxic condition in the myocardium by supporting the decrease of oxygen demand and suppression of sympathetic activity, thus possibly contributing to the improvement of systolic dysfunction 33,37 . Meanwhile, diastolic function was affected by the relaxation function, left atrial size, and fibrosis in the cardiac tissue, in addition to oxygen demand 38 . Therefore, as shown in a previous report 9 , recovery of diastolic dysfunction may be difficult in patients with HFrecEF who exhibit cardiac fibrosis due to aging, despite the condition being improved by GDMT.
As the clinical implication, patients with HFrecEF and high E/e′ ratio exhibited unfavorable clinical outcomes in this study. It was reported that the changes in heart rate during the follow-up periods were associated with relapse in patients with dilated cardiomyopathy and recovered LVEF 39 . However, heart rate at discharge and at the 1-year follow-up, as well as changes in heart rate, were compatible between the two groups in the current study; therefore, monitoring of the E/e′ ratio may be a useful risk stratification tool for future CV events in patients with HFrecEF. Meanwhile, Pritchett et al. reported that LAVI was also correlated with the diastolic dysfunction as well as long-term prognosis in a cross-sectional sample with > 45 years of age 40 . In the current study, AUC of E/e′ ratio at the 1-year follow-up for the composite endpoint was higher than that of LAVI (0.70 vs. 0.67), but not statistically significance between them (p = 0.76). It is still unclear which parameters are better for assessing the prognosis in patients with HFrecEF. The combination of the parameters might predict the prognosis accurately.
Although appropriate strategies for improving a patient's diastolic dysfunction remain unestablished, and careful observation and management are needed for these patients, it was reported that the angiotensin receptor neprilysin inhibitor (ARNI) might be effective in female patients with HFpEF, those who are older (> 65 years; postmenopausal women), or those with relatively lower LVEF (< 57%) 31,41 . Furthermore, ARNI altered the biomarker of abnormal extracellular matrix (ECM) homeostasis and improved clinical outcomes in patients with HFpEF, likely through antifibrotic effects 42 . This may indicate that ARNI had favorable effects on diastolic dysfunction. In this study population, ARNI might be effective in patients with HFrecEF and high E/e′ ratio because of the mean age of 68 years, high rate of females, and mean LVEF of 47% at the 1-year follow-up. Therefore, we considered that clinicians should assess the implications of ARNI for patients with HFrecEF and diastolic dysfunction, which may lead to better clinical outcomes.
There were several limitations in this study. This was a retrospective study performed at a single center with a small sample size. There may be a selection bias as some patients were excluded from this study due to missing echocardiographic data. Additionally, patients who underwent mitral valve plasty or replacement were excluded as their E/e′ ratio could not be accurately measured. Although we grouped patients according to E/e′ ratio using a cutoff value of 12.1 (Youden index), other cutoff values-such as an E/e′ ratio of 14 or 15-were not assessed, since the number of patients with an E/e′ ratio greater than 14 was very small. Although we assessed the septal e′, the diagnostic accuracy of tissue Doppler for evaluating LV filling pressure and diastolic dysfunction have been still discussed 43 . We did not show the NYHA classification at the discharge and the 1-year follow-up because there was not enough amount of data on it. No patients were prescribed ARNI and ivabradine at both discharge and the 1-year follow-up as these drugs were not approved in Japan at the time, and sodium-glucose cotransporter 2 inhibitors were only approved for patients with diabetes during the study period; these differences may have influenced the results. The multivariate analysis might be overfitting. Further large-scale, prospective investigations are needed to determine the best management for patients with HFrecEF and high E/e′ ratio.

Conclusions
Elderly and female patients hospitalized for HFrEF may exhibit diastolic dysfunction at the 1-year follow-up even if their LVEF had improved; additionally, patients with HFrecEF and high E/e′ ratio at the 1-year follow-up had a poor prognosis. Close observation and novel strategies are, thus, needed for this population.

Data availability
Data are available from the corresponding authors upon reasonable request.