Consecutive symptomatic HF patients (New York Heart Association [NYHA] functional class ≥ 2) who had an LVEF > 40% on echocardiography between November 2012 and December 2019 were studied. Patients were eligible if N-terminal prohormone of brain natriuretic peptide (NT-proBNP) was > 125 pg/ml and echocardiography showed evidence of LV diastolic dysfunction, left atrial dilatation and/or LV hypertrophy, according to current European Society of Cardiology criteria [5]. Patients were excluded from the present analysis if they had LVEF ≤ 40% on either echocardiography or CMR imaging, (corrected) congenital heart disease, genetically proven hypertrophic-, or dilated cardiomyopathy, or if they had more than moderate left-sided valvular disease. Other exclusion criteria included myocardial infarction < 3 months prior to CMR imaging contra-indications for CMR imaging, including claustrophobia and presence of pacemaker or internal cardiac defibrillator (ICD). A study flowchart is shown in Fig. 1. Previous myocardial infarction was assessed by reviewing the medical records. All patients were part of a standard diagnostic protocol for HF patients with an LVEF > 40%. This protocol consisted of a thorough examination including laboratory testing, echocardiography and routine cardiac magnetic resonance (CMR) imaging for the aetiology of HF. Serum NT-proBNP and estimated glomerular filtration rate (eGFR) were determined using an immunoassay based on electrochemiluminescence (Elecsys, Roche Diagnostics, Mannheim, Germany). The Institutional Review Board of the University Medical Centre Groningen approved the study and because of the retrospective nature of the study, the need for individual informed consent was waived. The present study was in concordance with the principles outlined in the Declaration of Helsinki.

Echocardiographic parameters were assessed according to the current recommendations for cardiac chamber quantification and included: LV systolic function, LV diastolic function, left atrial volume index, LV mass index, valvular stenosis and/or regurgitation and the peak pressure gradient across the tricuspid valve [12].

The endpoint used in this study was all-cause mortality. Follow-up time was defined as the time from CMR to death. Medical records were reviewed to assess whether the endpoint was met.

Data are presented as numbers and percentages, mean ± standard deviation or median with interquartile ranges. Differences in categorical variables were analysed using the Chi-squared test or Fisher’s exact test, where appropriate. Differences in normally distributed continuous variables were compared using the independent-sample t test or Wilcoxon signed rank test if variables were not normally distributed. Univariable logistic regression was used to determine which covariates were associated with the presence of LGE lesions. To minimize the risk of type II errors, we considered p < 0.1 as a significant univariable association and these variables were entered into a multivariable backward selection model. First-line interaction between all covariates with p < 0.1 for the association with the presence of LGE was assessed.

A Kaplan–Meier plot with log rank test was used to display relation between LGE and outcome, which was defined as all-cause mortality. The association between LGE and all-cause mortality was assessed using a univariable Cox proportional hazard regression model. The multivariable Cox regression model was adjusted for age, sex, NYHA functional class, NT-proBNP, LGE mass and LVEF as these variables are strong determinants of outcome in patients with HFpEF [16‐18]. The Cox proportional hazards assumption was assessed by visually inspecting the log minus log survival plots over time, which showed proportionality. Statistical analyses were performed using SPSS (Version 23, Chicago, Illinois). Statistical significance was considered achieved at a p value < 0.05.

The study population consisted of 110 HF patients (Fig. 1). In 15 patients, LGE imaging was either not performed or the scan was of insufficient quality to assess LGE and these patients were excluded. Mean age was 71 ± 10 years, 49% were female and the median NT-proBNP was 1259 pg/ml (Table 1). The clinical characteristics stratified by HFpEF or HFmrEF are depicted in Supplementary Table 1. HFpEF patients were more often female (59 vs. 30%, p = 0.004), and more often had a history of hypertension (82 vs. 60%, p = 0.01) and less often renal dysfunction (30 vs. 51%, p = 0.03).

LGE lesions were observed in 37 (34%) patients, of which 22 (20%) were ischemic-, and 15 (14%) were considered to be non-ischemic LGE lesions (Fig. 2). Of the 22 patients with an ischemic LGE lesion, 8 (36%) did not have a history of myocardial infarction.

Patients with an LGE lesion were more often male (76 vs. 38%, p < 0.001) and had a previous myocardial infarction more frequently (49 vs. 11%, p < 0.001). In addition, patients with an LGE lesion had increased levels of Troponin T compared to those without LGE (27 vs. 17 ng/L, p = 0.007). The median time interval between troponin T assessment and CMR imaging was 37 [16–77] days. CMR measurements are depicted in Table 2. LV mass index was almost 20% higher in patients with LGE lesions compared to patients without LGE (p = 0.02). LVEF on echocardiography and CMR were moderately associated (r = 0.38, p < 0.001). CMR characteristics for HFpEF and HFmrEF are displayed in Supplementary Table 2.

T1 mapping was performed in 75 (68%) patients. In patients with LGE lesions, native T1 times and ECV were significantly increased compared to patients without LGE lesions (1043 vs. 1013 ms, p = 0.01 and 28.6 vs. 26.6%, p = 0.04, respectively (Table 2).

Patient- and CMR characteristics with a univariable association (p < 0.1) with LGE lesions are depicted in Table 3. Previous myocardial infarction and LV mass index were strong predictors for the presence of LGE lesions and remained independently associated after adjustment (Odds ratio 6.32 with 95% confidence interval 2.07–19.31, p = 0.001 and 1.68 (1.03–2.73), p = 0.04, respectively). Forced entry of age and sex into the model did not improve the predictive accuracy of the model.

During a median follow-up of 34 months (interquartile range 19–53), 19 (17%) patients died. Of these 19 patients, 11 died due to cardiovascular causes and 8 due to a variety of non-cardiovascular causes such as cancer, and chronic obstructive pulmonary disease. In Fig. 3, a typical CMR example is shown of a patient with a non-ischemic LGE lesion, who died due to cardiovascular causes. Mortality was significantly higher in patients with LGE lesions (Log Rank p = 0.002, Fig. 4) The presence of LGE was an independent determinant of all-cause mortality after the adjustment for age, sex, NYHA functional class, NT-proBNP, LGE mass and LVEF (adjusted hazards ratio 5.3, 95% confidence interval 1.5–18.1, p = 0.009) (Table 4). LV diastolic dysfunction, LV hypertrophy and left atrial volume index on echocardiography were all not associated with mortality (data not shown).

Our study is exploratory in nature and suffers from general limitations associated with its retrospective nature. In addition, a few specific limitations should be acknowledged. First, because T1/ECV mapping became available in our center during the course of the study, it is only available in 75 out of 110 patients. Second, although our population was meticulously characterised, our sample size is relatively small and our mortality rate was lower than in most HFmrEF/HFpEF trials. The incidence and prognostic significance of LGE may, therefore, be different in the general HFmrEF/HFpEF population. Fourth, we did not include a control group, and we cannot rule out that the association between LGE and mortality also applies to similar subjects without HF.

Our study has shown that a considerable amount of HF patients with LVEF > 40% have myocardial scars and these patients are at increased risk. The presence of an LGE lesion may not only aid in follow-up, but the pattern of LGE may also assist in the diagnosis of the underlying cause of HFpEF/HFmrEF. For instance, of the 22 patients in our study with an ischemic LGE scar, 8 (36%) did not have a history of myocardial infarction. These findings suggest that HF was caused by a silent myocardial infarction, and these patients may benefit from anti-ischemic therapy. On the other hand, clinical trials may also profit from selecting HF patients with LGE lesions, as this would result in a study population at increased risk.