The principal finding in our study is that stress CMR unmasks potentially clinically relevant undiagnosed cardiac pathology in a significant proportion of patients (27%) labelled as HFpEF after echocardiography. A clinically relevant proportion of our patients was identified as having hitherto unknown coronary artery disease or microvascular dysfunction. Moreover, despite being part of the TTE-based exclusion criteria at study entry, new cases of HCM and constrictive pericarditis were identified during subsequent CMR evaluation. Our observations suggest that previous intervention trials in HFpEF are likely to have included patients meeting one or more exclusion criteria, thereby possibly influencing treatment response. These additional pathologies, when grouped together in our cohort, were associated with adverse outcomes.
‘New CMR diagnoses’
The reasons for the higher pick-up rate of new clinical diagnoses with CMR are multiple. Firstly, the overall image quality for TTE in our study was poor, reflecting the clinical profile of our challenging population, with a high prevalence of obesity, lung disease and atrial fibrillation [
21]. These comorbidities are typical of HFpEF as reported in the literature [
1]. Furthermore, the low feasibility (inadequate endocardial border definition in nearly one-third) and diagnostic utility of TTE in HF has previously been reported and is subject to wider limits of agreement compared with CMR [
9,
22]. The ability of CMR to interrogate any imaging plane and perform in vivo tissue characterization (e.g. by LGE) makes this the reference standard for detection of new diagnoses in our cohort [
9‐
11].
Previous reports quote a wide range for the prevalence of CAD in HFpEF, comprising primarily data from epidemiological studies and registries. Furthermore, the presence of CAD was variably based on patient reporting, use of insensitive and non-specific investigations (e.g. ECG, exercise treadmill tests), inconsistent diagnostic cut-offs for angiographic disease severity, and did not incorporate CMR [
23]. In this study, CMR increased the overall proportion of significant CAD (silent MI and/or ischaemia) from 21% to 34%, equivalent to a relative increase of 63%. These findings (and microvascular dysfunction) might be expected, given the proportion of elderly, hypertensive and diabetic patients in our cohort [
24]. Furthermore, these greater number of ‘new’ CAD diagnoses is perhaps unsurprising given that CAD was not part of our exclusion criteria. We used a practical definition of HFpEF and current clinical guidelines [
25] for HF do not mandate routine investigation for CAD unless accompanied by anginal symptoms recalcitrant to medical therapy. Additionally, the higher numbers of ‘silent’ CAD could also be explained by the inability of some patients to provoke clinical symptoms due to limited exercise capacity owing to co-morbidities. Conversely, exertional breathlessness may represent angina equivalent. The typical patterns of infarction (small number of segments and ≤50% transmurality) in our study are in keeping with overall preservation of LVEF. In such cases, the diagnostic accuracies of both ECG (Q wave) and TTE (RWMAs) are low in concordance with published literature [
26].
Diagnosing HCM represents an imaging challenge in this cohort of patients. The latest HCM diagnostic guidelines [
12] advocate a morphological description of imaging in suspected subjects. These guidelines are also more inclusive of considering HCM as a diagnosis in any patients whereby increased LV wall thickness cannot solely be explained by abnormal loading conditions. CMR features supportive of HCM in hypertensive patients include a more asymmetric pattern of LVH and LGE at the insertion points and in segments of maximal LV wall thickening [
27,
28]. Furthermore, LGE is reportedly present in 65% with HCM, similar to our cohort [
12].
HCM is characterized by non-specific diverse patterns of hypertrophy with or without left ventricular outflow tract obstruction or systolic anterior motion of the mitral valve [
12,
14,
29]. In HFpEF, LVH is a common finding [
1] and co-existing conditions such as ageing, obesity and hypertension are additional confounders [
30]. Furthermore, hypertensive heart disease classically presents with concentric hypertrophy and wall thickness rarely exceeds 15–16 mm [
28]. Deciphering the pattern of LVH according to mass and relative wall thickness calculations traditionally used in TTE is fraught with intrinsic methodological limitations [
31]. These factors along with sub-optimal image quality [
29] and the very high prevalence of hypertension (90%) may explain the underreporting of HCM by TTE in our cohort. In our study, patients who met wall thickness criteria for HCM on TTE were not reported as likely HCM most probably due to a predominant concentric pattern of LVH. Whilst TTE traditionally risks overestimating wall thickness (e.g. oblique cuts) [
12], underestimation has been noted in a small (12%) proportion, especially if confined to the inferolateral, anterolateral or apical segments. In contrast, the superior endocardial definition afforded by CMR allows a more precise measurement of LV wall thickness and hypertrophy [
29].
Current TTE diagnostic criteria for constrictive pericarditis have lower sensitivities compared to CMR (pericardial thickening: 36% vs 88%, septal bounce: 62% vs 81%) [
5,
20]. In our cohort, the majority of these TTE parameters were not detected, which again is a likely reflection of poor image quality.
Implications
Our CMR findings reinforce the marked clinical heterogeneity in HFpEF [
1] and provide alternative explanations for symptoms in a significant minority of patients. Survival following silent MI is comparable to known MI [
32]. Importantly, diagnosis by CMR enables initiation of effective secondary prevention treatment and guides revascularization, given that most affected myocardial segments identified in our cohort were viable [
13]. Our data suggest that screening for significant CAD should be undertaken in patients with suspected HFpEF. A diagnosis of HCM has implications for both patients and relatives. CMR improves risk stratification and may enable earlier initiation of therapies such as implantable defibrillator devices [
12]. Constrictive pericarditis is potentially curable and pericardial enhancement on LGE may predict treatment response [
5].
Implications for current HFpEF clinical trials
Our study has important implications and ramifications for HFpEF clinical trials and current treatment strategies. Variable definitions of HFpEF and phenotypic heterogeneity displayed in prior studies have previously been proposed as a key reasons for treatment failure [
1,
3]. This has led to a paradigm shift in focus to study ‘purer’ subsets of HFpEF resulting in more detailed mechanistic studies. Our CMR study findings provide additional explanations for such poor outcomes whereby TTE remains the primary entry tool for trial enrolment. Our data suggests that TTE alone is incapable of rigorously excluding imaging phenocopies of HFpEF prior to study entry. Such conditions have alternate pathophysiological mechanisms, respond differently to existing therapies and contribute to adverse outcomes. While TTE is comparatively more extensively available, and therefore attractive for clinical trial design, access to CMR is rapidly increasing. Furthermore, CMR refines the diagnosis and sub-categorises HFpEF into ‘purer forms’ and alternative pathologies, enabling disease-specific tailored therapies, and provides prognostic data.
The routine use of stress CMR in HFpEF patients should refine diagnosis and treatment strategies as we move towards an era of precision medicine. However, further randomised trials are needed to assess the wider impact of CMR in terms of clinical outcome, resource utilization and cost-effectiveness.
Limitations
The definition of HFpEF used in our study was not in accordance with current European Society of Cardiology (ESC) guidelines [
8]. However, we took a pragmatic approach to reflect a real world setting. In particular, the presence of diastolic dysfunction was not a pre-requisite for study entry since recent contemporary clinical trials have highlighted normal diastolic function at rest in approximately a third of such patients [
7]. Although all patients meeting inclusion criteria were invited, 26 out of 180 (14%) did not undergo CMR, which might raise concerns about its applicability to the wider HFpEF population. Whilst chronic obstructive pulmonary disease is quite prevalent in the clinical scenario of HFpEF, we only excluded patients with severe disease (and likewise severe valvular disease) to minimise the contribution from alternate causes of HF symptoms. Besides our cohort still comprised chronic obstructive pulmonary disease subjects in nearly one-fifth who underwent CMR. Six patients with pacemakers did not undergo CMR: at the time the study was conducted, our centre was not implanting CMR conditional devices. Although all CMR scans were performed solely at 3 T, we expect the study findings to be similar with a 1.5 T system.
Discriminating microvascular dysfunction from global coronary ischaemia can be challenging with CMR and raises the possibility of under-reporting of CAD. Furthermore, patients did not have stress echocardiography which may have identified more patients with ischaemia. In this cohort of patients with multiple risk factors for LVH, ultimately the imaging diagnosis of HCM is one of exclusion. However, the most recent ESC guidelines recommend defining HCM in patients with LVH ≥ 15 mm not solely explained by loading conditions [
12]. Our CMR reports were generated using a clinical protocol exclusive of T1 and T2 mapping which were not routinely used at the time of study conduct. T1 mapping may have unmasked further hypertrophic phenotypes [
12] such as cardiac amyloid and Anderson-Fabry’s disease, and T2 mapping may have been helpful in cases of constrictive pericarditis [
5].
While the CMR reports were generated by GPM and ASHC, clinical endpoints were collated by PK who was not blind to CMR results. However, the HF hospitalization events were clearly objectively defined (see methods section) and assessment of vital status is robust. Some patients may have had hospitalizations exclusive of our hospital. However, there should be no systematic bias for those with or without ‘new’ diagnoses.