Background

Heart failure (HF) with preserved ejection fraction (HFpEF) has been established as a major cause of cardiovascular morbidity and mortality, especially among the elderly [1, 2]. Prevalence is increasing, affecting half of the patients with clinical signs of heart failure [3]. However, compared to HF with reduced ejection fraction (HFrEF), survival in HFpEF has not improved over time, and to date, no treatment effectively improves outcomes, probably because of the phenotypic heterogeneity of this syndrome [4].

Several mechanisms have been implicated in HFpEF, including advanced age and cardiovascular, metabolic, and pro-inflammatory comorbidities such as hypertension, diabetes, obesity, chronic obstructive pulmonary disease), coronary disease and renal failure [1, 4,5,6]. The exact pathophysiology of HFpEF remains unclear as a result of the absence of a proper animal model and the presence of numerous confounding effects. Recently, several studies using autopsies or myocardial biopsies have highlighted the key role of myocardial extracellular matrix abnormalities (fibrosis [7,8,9]) and myocardial structural changes such as altered cardiomyocyte function (hypertrophy [7]), systemic and coronary microvascular inflammation, and endothelial dysfunction (oxidative stress) as all being involved in myocardial stiffness and diastolic dysfunction [10, 11].

Quantification of extracellular volume fraction (ECV) by cardiovascular magnetic resonance (CMR) has recently emerged as a novel non-invasive diagnostic tool to assess myocardial fibrosis [12,13,14]. Recently, some studies have demonstrated the importance of diffuse or focal fibrosis estimated by biopsies or/and by CMR in patients with HFpEF [15, 16].

The aim of our study was to evaluate the presence, associations, and prognostic significance of ECV, likely reflecting diffuse myocardial fibrosis, in HFpEF patients.

Methods

Study population

Between December 2014 and October 2016, consecutive patients with suspected HFpEF were prospectively evaluated for inclusion in the study. The local ethics committee approved the study, and all patients gave written informed consent before study enrollment (Clinical trial NCT03197350).

The following criteria had to be fulfilled for study inclusion: New York Heart Association functional (NYHA) class ≥II, typical signs of HF, NT-proBNP > 350 pg/ml and/or a hospitalization for HF in the previous 12 months, left ventricular (LV) ejection fraction (EF) ≥50%, and relevant structural heart disease (LV hypertrophy/left atrial (LA) enlargement) and/or diastolic dysfunction by echocardiography [17]. Ischemic cardiomyopathy was defined as history of myocardial infarction or revascularization by either coronary artery bypass graft surgery (CABG) or coronary artery angioplasty. Diabetes was defined as an abnormal fasting glycaemia (> 126 mg/dl) or the use of antidiabetic drugs.

The exclusion criteria were any contraindications to CMR (pacemaker, estimated glomerular filtration rate (eGFR) < 30 ml/min/m2, claustrophobia), severe valvular disease, infiltrative (ie amyloidosis, sarcoidosis or hypertrophic cardiomyopathy, acute coronary syndrome in the last 30 days, chronic obstructive pulmonary disease GOLD 3 or 4, congenital heart disease, pericardial disease, atrial fibrillation with a ventricular response > 140 bpm, and severe anemia (hemoglobin < 7 g/dl). Patients with severe cognitive disorder were also excluded. One patient screened for the study was excluded due to presence of unknown cardiac amyloidosis.

Patients were compared to 26 age- and sex-matched controls without history of cardiovascular disease. Controls were recruited by advertisement in the community. All subjects underwent a full clinical exam, electrocardiogram (ECG), transthoracic echocardiography and exercise stress test, which all had to be normal prior to inclusion.

Controls and patients underwent blood sampling, complete transthoracic echocardiography and a CMR.

Echocardiography

All subjects underwent a two-dimensional (2D) transthoracic echocardiography at inclusion (iE33 system Philips Healthcare, Best, The Netherlands) with parasternal long- and short-axis views and apical views to assess LV and right ventricular (RV) systolic and diastolic functions and measurements of LA and right atrial (RA) volumes, as well as a valvular evaluation. Mitral valve inflow pattern (E and A velocity) and septal and lateral mitral valve annular velocities (e’) were recorded.

RV function was assessed by systolic annular tissue velocity of the lateral tricuspid annulus, tricuspid annular plane systolic excursion (TAPSE), and tracing the RV endocardium in the apical four-chamber view in systole and diastole to obtain fractional area change (RV FAC%) [18]. All measurements were averaged over three beats in atrial fibrillation.

Cardiovascular magnetic resonance

CMR was performed using a 3 Tesla system (Ingenia, Philips Hearlthcare). Briefly, after localization of the heart, to assess LV and RV myocardial function and mass, 10–12 consecutive short-axis (SAX) images and 2-, 3-, and 4-chamber long-axis images of the LV were acquired using a cine balanced steady-state free precession sequence (bSSFP). Then, mid-ventricular short-axis modified Look-Locker inversion recovery (MOLLI) images were acquired for T1 determination using an 11-image, 18-heartbeat 3-(3)-3-(3)-5 bSSFP sequence. A total dose of 0.2 mmol/kg gadobutrol (Gadavist, Bayer Healthcare Leverkusen, Germany) was injected, and 10–15 min after contrast injection, short- and long-axis 2D inversion recovery late gadolinium enhancement (LGE) images were acquired to evaluate focal myocardial fibrosis. Finally, 15-min post-contrast, MOLLI T1 mapping was repeated in a protocol identical to that used for pre-contrast T1 mapping. The presence of LGE was visually assessed.

Pre- and post-contrast MOLLI images were processed using the open-source software MRmap v1.4 [19] under IDL. Images were corrected for respiratory motion when needed. T1 maps were exported to Osirix 5.7 (Pixmeo, Geneva, Switzerland) and pre- and post-myocardial T1 times were measured in six regions-of-interest (ROI) in the myocardium (anterior, anterolateral, inferolateral, inferior, inferoseptal, anteroseptal). We calculated the average T1 time of the six different ROIs. Areas of ischemic focal fibrosis identified by late gadolinium enhancement were excluded from the analysis. The partition coefficient lambda (λ) and ECV were then computed according to the formula [20].

End-diastolic and end-systolic LV and RV volumes as well as LV mass and presence of LGE were analyzed using the freely available software Segment 2.0 (Medviso, Lund, Sweden), as previously described. LGE was considered present if myocardial enhancement was observed on both short-axis and long-axis views. The total LGE volume was calculated by summing the LGE volume of all slices, and the ratio (%) of LGE was then calculated using the same software.

Follow-up

Patients were prospectively followed by ambulatory visits and telephone calls at 6-month intervals. Clinical and survival status was obtained by follow up visits and by phone contact with the patients, their relatives, and their physician if necessarily. The primary outcome was a composite of all-cause mortality or a first hospitalization for HF. Vital status was ascertained by medical record review. First HF hospitalization was defined as patients treated in the emergency room or admitted to a hospital and requiring intravenous diuretics. Patients had at least one symptom and 2 signs of HF (peripheral edema, pulmonary crackles, high NT-proBNP level, radiological signs of pulmonary congestion or hemodynamic evidence).

Statistical analysis

Statistical analyses were performed using SPSS version 22 (International Business Machines, Inc., Armonk, New York, USA) and STATA version 11 (Stata Corporation, College Station, Texas, USA) software. All tests were two-sided, and a P < 0.05 was considered statistically significant. Continuous variables were expressed as mean ± 1 SD if normally distributed or as medians (25th and 75th percentiles) if not normally distributed. Categorical variables were expressed as counts and percentages. Comparison between groups was performed using ANOVA, chi-square test, or unpaired t-tests when appropriate. We determined an ECV age and sex-adjusted cutoff value corresponding to the upper 95% confidence interval of 26 age- and sex matched volunteers (ECV ≥ 33%).

Logistic regression was performed to determine predictors of abnormal diffuse fibrosis (ECV above or below 95% confidence intervals in controls). For this purpose, after univariate comparison of the two groups, parameters with a P < 0.10 were proposed for inclusion in the multiple logistic regression analysis with a backward selection procedure.

Event-free survival was estimated using Kaplan–Meier methods and Cox regression analysis. All baseline and imaging variables were initially proposed for inclusion in a univariate Cox proportional hazard model. To avoid colinearity in the Cox regression model, the correlation coefficients between covariates were examined. In cases of colinearity (r > 0.50), only the strongest of the two covariates was proposed for inclusion into the multivariate model. After univariate Cox regression analysis, all significant variables (P < 0.10) were entered into a stepwise forward multivariate Cox regression model. Two different models were evaluated using ECV as either a continuous or a categorical variable (ECV ≥ 33%).

A prognostic score was established in patients followed up for at least 6 months. The accuracy of ECV, risk score or LGE to predict composite outcome were evaluated by area under the receiver operator curve (ROC) curves. This score was validated in 53 consecutive HFpEF patients recruited between October 2016 and July 2017.

Results

Baseline characteristics

Between December 2014 and October 2016, a total of 118 consecutive HFpEF patients (78 ± 8 years, 63% women) and 26 age- and sex-matched controls (76 ± 5 years, 62% women) were included in the study. The demographic, clinical, and laboratory characteristics of HFpEF patients and controls are summarized in Table 1. We observed a high prevalence of established cardiovascular risk factors in our HFpEF population; including arterial hypertension (93%), diabetes (39%), hypercholesterolemia (67%), and higher body mass index (BMI). HFpEF patients had worse renal function and lower hemoglobin and hematocrit levels than healthy controls, and 62% of patients had a history of atrial fibrillation.

Table 1 Baseline characteristics of HFpEF patients and age- and sex-matched controls
Full size table

Table 2 summarizes echocardiographic and CMR measurements. As expected, compared to age- and sex-matched healthy controls, HFpEF patients had higher E/e’ ratio, higher indexed LA and RA volumes, higher RV/RA gradient, and worse RV function as evaluated by TAPSE and FAC. Prevalence of pulmonary hypertension (RV–RA gradient > 35 mmHg) in our population was 39% (n = 46). By CMR, HFpEF patients had a slightly lower LVEF and higher indexed LV mass than age- and sex-matched healthy controls. Twenty-six HFpEF patients had LGE (8 = focal spots, 18 = ischemic pattern). When present, the average percentage of LGE was 5.5 ± 2.9%.

Table 2 Echocardiographic and CMR parameters of HFpEF patients and age- and sex-matched controls
Full size table

ECV was significantly higher in HFpEF patients than in the healthy control group (Fig. 1). Forty-nine (42%) HFpEF patients had significant diffuse fibrosis based on the ECV cutoff (defined as ≥33%). Sixteen (14%) HFpEF patients had impaired RV systolic function defined as RVEF≤45% by CMR. HFpEF patients with RV dysfunction had higher ECV (36 ± 6% vs. 32 ± 4%, P < 0.001), higher indexed LA volume (81 ± 26 ml/m2 vs. 64 ± 28 ml/m2, P = 0.03), lower LVEF (59 ± 7% vs. 64 ± 8%, P = 0.010), lower TAPSE (14 ± 5 mm vs. 19 ± 5 mm, P < 0.001), lower FAC (32 ± 6% vs. 43 ± 8%, P < 0.001), and higher septal E/e’ ratio (24 ± 9 vs. 17 ± 7, P = 0.001).

Fig. 1
figure 1

Extracellular volume fraction (ECV) values between age- and sex-matched healthy controls and heart failure with preserved ejection fraction (HFpEF) patients

Full size image

Diabetic HFpEF patients were younger (75 ± 9 vs. 80 ± 7 years, P = 0.002), had a higher body mass index (BMI) (31.2 ± 6.9 vs. 27.3 ± 6.1, P = 0.002), a lower atrial fibrillation history rate (48% vs. 71%, P = 0.012), and more known ischemic cardiomyopathy (43% vs. 26%, P = 0.055) than non-diabetic HFpEF patients. Echocardiographic and CMR parameters were quite similar except for a trend to higher indexed LV mass in diabetics (71.0 ± 12.6 vs. 65.6 ± 16.1 g/m2, P = 0.056). Diabetic HFpEF patients presented more frequently ECV ≥33% than non-diabetic patients (52% vs. 35%, P = 0.061).

Predictors of high ECV

Table 3 compares HFpEF patients with high and low ECV values. ECV and the lambda coefficient were different between the two groups, demonstrating that the difference in ECV was not solely the result of a significant difference in hematocrit level..

Table 3 Baseline characteristics of HFpEF patients with ECV < or ≥ 33%
Full size table

HFpEF patients with ECV ≥ 33% had lower BMI, lower hematocrit, higher prevalence of diabetes, higher transmitral peak E wave velocity and higher E/e’ ratio. A higher proportion of HFpEF patients with ECV ≥ 33% had an impaired RV systolic function (CMR RVEF ≤45% (20% vs. 9%, P = 0.073). There were no differences in NT-proBNP and LGE in HFpEF patients between the two groups.

In multivariate logistic regression, BMI (OR = 0.92 [0.86–0.98], P = 0.011), presence of diabetes (OR = 2.62 [1.11–6.18], P = 0.028), and higher transmitral peak E wave velocity (OR = 1.02 [1.00–1.03], P = 0.022) were significantly associated with high ECV value.

Outcomes

During a mean follow-up of 11 ± 6 months, we observed 43 events (11 deaths and 32 hospitalizations for HF) (Fig. 2a). The primary outcome (all-cause mortality or first hospitalization for HF) occurred in 38 patients (32%). Only one patient was lost to follow up. The percentage of combined event at 18 months in our population was 50%.

Fig. 2
figure 2

(a) Mortality and HF hospitalization Kaplan Meier Curves (b) Event-free survival in HFpEF patients according to ECV cutoff

Full size image

In univariate Cox regression analysis, lower eGFR, lower hemoglobin, presence of diabetes, higher transmitral E wave velocity, and ECV ≥ 33% were significant predictors of the composite outcome (Table 4). In multivariate Cox regression analysis, only the presence of diabetes (HR = 1.98 [1.04; 3.76], P = 0.038) and hemoglobin level (HR = 0.81 [0.67; 0.98], P = 0.028) were significantly associated with the composite outcome (Table 5). The ability of either ECV as a continuous variable or ECV as a categorical variable (ECV > 33%) to improve this model was then tested and added significant prognostic information (χ2 4.33, P = 0.037 and χ2 4.46, P = 0.035, respectively), as opposed to transmitral peak E wave velocity and eGFR (χ2 3.00, p = 0.083 and χ2 2.86, p = 0.091 respectively). The ability of the model to discriminate outcome was good with an area under the curve (AUC) of 0.76. Kaplan–Meier event-free survival curves showed that HFpEF patients with ECV ≥ 33% had poorer one-year prognosis than those with ECV < 33% (56 ± 8% vs 82 ± 5%, P = 0.001, Fig. 2b).

Table 4 Univariate Cox analysis for composite event (death and first HF)
Full size table
Table 5 Multivariate Cox analysis for event-free survival in HFpEF
Full size table

A prognostic score was established based on the significant predictors of composite outcome in patients with at least 6 months of follow up (n = 97). Diabetes, hemoglobin, and ECV > 33% were thus used to build the risk score. ROC curves showed a better discrimination with the prognostic score and ECV (c-statistic of 0,76 and 0,67 respectively) compared to LGE (c statistic 0,51) (Additional file 1). In the validation cohort of 53 patients, 2 deaths and 9 HF hospitalizations were observed during a mean follow up of 11 ± 5 months.

Our risk score based on the initial cohort and applied in the validation cohort had an AUC of 0.71 to predict primary outcome.

Discussion

We sought to evaluate the presence, associations, and prognostic significance of quantification of ECV using 3 T CMR in a prospective and well-characterized HFpEF cohort. The salient findings of our study are as follows. Mean ECV by T1-mapping was significantly higher in HFpEF patients than in age-matched healthy controls; ECV in HFpEF patients was related to the presence of anemia and diabetes and associated with altered diastolic function by echocardiography and lower CMR RV systolic function; and finally, increased ECV and the presence of anemia and diabetes were independent risk markers of short-term poor prognosis with increased rehospitalization or all-cause mortality in HFpEF.

HFpEF

HFpEF is a heterogeneous disease observed mainly in the aging population. We and others demonstrated that aging is associated with structural changes in the heart, such as alterations of diastolic function, increased stiffening, LV and atrial remodeling characterized by decreased LV and RV volumes and mass, increased atrial volumes, and an increase in ECV [21]. It is crucial to separate age-related changes from disease-related processes associated with HFpEF. Therefore, we compared our population to carefully age- and sex-matched healthy subjects, allowing for a better understanding of these two processes. As compared to previous work [22, 23], our HFpEF population was older and had more comorbidities, probably reflecting a less selected population but a more advanced stage because patients were highly symptomatic with high NT-proBNP levels (45% NYHA III/IV). Overall, and in accordance with prior reports, we observed relatively poor outcomes with a high rehospitalization rate and high mortality despite optimal medical treatment.

Role of extracellular matrix abnormalities in HFpEF

Extracellular matrix abnormalities causing LV stiffening and secondary LV diastolic dysfunction are probably among the main potential pathophysiologic mechanisms involved in HFpEF. Previous studies using endomyocardial biopsies or autopsies have already shown that the extent of fibrosis is higher in HFpEF patients than in controls [7].Studies have demonstrated that ECV closely correlates with histologically determined diffuse interstitial fibrosis, providing a non-invasive estimation for its quantification; however, only a few studies have evaluated ECV in HFpEF patients [15, 22, 23]. Our findings confirm that high ECV likely reflecting diffuse extracellular matrix abnormalities, a potential surrogate for myocardial fibrosis, may play a key role in the pathophysiology of HFpEF. Indeed, ECV was significantly increased relative to the age-matched population, yet only 42% of our HFpEF patients had high ECV, suggesting that other mechanisms are involved in HFpEF pathophysiology. Assessment of ECV allows for a direct evaluation of the extracellular space, reflecting interstitial disease. However, it lacks the ability to provide information about the relative contribution of edema, fibrosis, inflammation, or deposition of other extracellular proteins such as amyloid [13] and therefore may not allow for a full understanding of the pathophysiological mechanisms underlying HFpEF. In our study, low BMI, diabetes, and high transmitral peak E wave velocity were significant determinants of high ECV. High transmitral peak E wave velocity is a good surrogate for high filling pressures, and it could be easily anticipated as an important determinant of fibrosis.

Surprisingly, we observed that a higher BMI was associated with lower ECV values, suggesting that diffuse fibrosis is not the sole mechanism involved in HFpEF pathogenesis in the obese population. Indeed, adiposity-induced inflammation has considerable adverse effects, including endothelial dysfunction, capillary rarefaction, and mitochondrial dysfunction in both the cardiac and systemic beds [24].

Diabetic cardiomyopathy could be considered as a form of HFpEF, explaining the increased incidence of HF in the diabetic population [25]. This cardiomyopathy is characterized by insulin resistance and a loss of metabolic flexibility, and subsequently by cardiomyocyte hypertrophy and increased fibrosis. Increased reactive oxygen species production is one of the major pathophysiological mechanisms triggered by hyperglycemia and high free fatty acid level.

Finally, native T1 time was surprisingly significantly lower in HFpEF patients than in healthy controls. We can only speculate on the explanation. Possible explanation could be that HFpEF patients have slightly higher intramyocadial iron concentration or more intramyocardial fat, particularly related to presence of diabetes.

Prognosis in HFpEF

As demonstrated in our study, the prognosis of HFpEF patients is still quite poor. Many studies have compared HFpEF and HFrEF prognosis and demonstrated the same or even a worse mortality and morbidity rate [26, 27] for HFpEF than for HFrEF. Our study identified high ECV, hemoglobin and diabetes as independent prognostic predictors for the short-term composite outcome. This result is in accordance with observations by Redfield et al. [28], who showed that profibrotic pathways may contribute to adverse outcomes in diabetic HFpEF patients. However, in our work, presence of LGE was not a significant predictor of the composite outcome (HR = 1.08 [0.96; 1.21], p = 0.22) in univariate analysis, in contrast to another recent study [16]. A potential explanation was that patients in this pre cited study had significantly higher percentage of LGE quantification (13 ± 8%), as opposed to our population where only 26 patients had small amounts of LGE (5.5 ± 2.9%).

The prognostic role of anemia has already been demonstrated in large HFpEF studies (SENIOR [29], MAGGIC [30], ARIC [26], respectively).

Clinical implications

Because HFpEF syndrome is a heterogeneous disease, better characterization of HFpEF phenotypes based on clinical presentation and biological and/or imaging data is crucial for better designing therapies [31]. In particular, the identification of anemia is relevant because its association with composite outcome could suggest a potential beneficial effect of iron-replacement therapy in HFpEF patients, and this hypothesis could be evaluated in larger randomized trials. Our study contributes to a better understanding of the heterogeneous and complicated nature of HFpEF [32]. Larger studies are needed to confirm our findings and prospectively validate our risk markers in other populations.

Study limitations

Our study is a single-center study of relatively small size, and its power is limited by a modest number of events, which bears a risk of overfitting in multivariable models. Yet our study was prospective, and the echocardiographic and CMR imaging, as well as biomarker sampling, were standardized. Because we excluded patients with contraindications to CMR, in particular renal failure, our conclusions cannot be generalized to all patients with HFpEF. Although we validated ECV assessment by T1 mapping in other populations against histopathology [15] with reasonably good correlation, in this fragile elderly population, we did not sample cardiac biopsies and thus could not ascertain the pathophysiological correlates of increased ECV in our patients. Another limitation is the lack of others SAX slices for the MOLLI acquisition to have a better idea of ECV value in the rest of the myocardium.

Focal fibrosis is probably also very important in HFpEF patients but only 26 patients had LGE.

Moreover, we interpreted CMR data from patients in atrial fibrillation but the impact of atrial fibrillation on MOLLI sequences has not been fully studied yet.

In addition, our study was performed mainly in a white population, and findings might differ in other groups, particularly in African American or other populations. Finally, the presence of different comorbidities is an important confounding factor in HFpEF. Thus, findings might be affected by selection criteria and presence of comorbidities in such HFpEF populations.

Conclusions

Among HFpEF patients, high ECV likely reflecting increased myocardial fibrosis was associated with BMI, diabetes, and transmitral peak E wave velocity. Among HFpEF patients, abnormal diffuse myocardial fibrosis estimated by ECV was associated with a higher rate of all-cause death and first HF hospitalization in the short term follow up.