Left atrial (LA) remodeling and dysfunction have been implicated in the pathophysiology of heart failure (HF) and are associated with poorer outcomes across a range of pathologies [1]. To date, the evidence base for such observations has largely been derived from echocardiography which is reliant upon adequate LA endocardial border definition for both volumetric and strain assessments [2]. Cardiovascular magnetic resonance imaging (CMR) affords superior spatial resolution, excellent contrast between blood pool and myocardium, has excellent reproducibility and is the current gold standard for LA volumetric [3] and functional assessment in sinus rhythm [4] or atrial fibrillation (AF) [5].

Heart failure with preserved ejection fraction (HFpEF) currently accounts for a significant proportion of all HF patients. Unlike heart failure with reduced ejection fraction (HFrEF), HFpEF still remains poorly understood and lacks proven, effective therapies [6]. Currently, prospective CMR studies assessing LA dysfunction in HFpEF are lacking. In this prospective, observational study of a well-characterized cohort with HFpEF, we assessed whether CMR-derived left atrial ejection fraction (LAEF) is different compared to controls and is of prognostic value.

This study prospectively evaluated the prognostic relevance of CMR-derived LAEF in a well-phenotyped cohort of HFpEF and healthy subjects. The principal findings are that (a) LAEF is lower in HFpEF compared to age- and sex-matched healthy controls (b) LAEF is associated with LA volumes and plasma biomarkers of atrial stress/stretch (c) inclusive of AF or sinus rhythm alone, lower LAEF is associated with adverse outcomes in HFpEF and (d) CMR-LAEF is also an independent prognostic marker in HFpEF.

Our work adds to a growing body of evidence implicating LA remodeling and dysfunction in HF [1]. Impaired LA function has previously been noted in conditions associated with HFpEF (e.g. diabetes, hypertension) even in the presence of normal LA size [13]. Furthermore, LAEF is reportedly lower in HFpEF compared to hypertensive subjects with LVH, corroborating our findings [14]. Diminished LA contractile reserve as a predictor of exercise incapacity has also been shown in subjects with preserved LV ejection fraction with [15] and without heart failure [16].

Current ESC guidelines advocate measurement of LA volumes and LV mass in all subjects with suspected HFpEF [6]. However, these measures are reliant on image quality and adequate endocardial border definition, unfortunately lacking in a third of HF cases when assessed with TTE [17]. Excellent spatial resolution and contrast, as well as the ability to scan in any image plane make CMR the current imaging gold standard [17]. Current imaging diagnostic criteria provide cut-offs for LAVImax and LV mass that are echocardiography-based and do not routinely incorporate CMR [6]. In our study, ROC analyses confirmed the strong signal from LAEF in discriminating HFpEF from controls. The reasons for this are likely multiple. Firstly, LAEF reduction might be a more precise reflection of elevated filling pressures than the other traditional surrogate imaging markers of chronic diastolic dysfunction [18]. Similar to our study, published literature has demonstrated normal-sized LA in approximately one-third of HFpEF subjects [19]. Our findings of reduced LAEF even in the presence of normal-sized atria reaffirms prior observations that LA dysfunction likely precedes overt LA remodeling in HFpEF [20]. Towards the other end of the spectrum, with worsening LA dilatation (and likely chronic LV&LA pressure overload), we have also demonstrated a close relationship between LA systolic function and volumes akin to the Frank-Starling mechanism i.e. LAEF reduces significantly more at higher volumes as contractile reserve becomes exhausted [12, 21]. In our subjects, more specific derangements in both reservoir (increased LAVmax and reservoir volume) and booster pump (increased LAVmin) function were also noted. LA reservoir function may be compromised by reduced LA compliance and LV longitudinal dysfunction typical of HFpEF [22]. In addition, LV diastolic dysfunction and concomitant elevated filling pressures further contribute to ineffective LA active emptying through increasing LA afterload and wall tension [23]. Compensatory improvements in conduit function may in part explain the lack of difference in conduit volume between HFpEF and controls in our study [24].

Our prospective study shows CMR-derived LAEF is an independent prognostic marker in HFpEF, both with and without AF. Previously, TTE-based observational studies [25] and HFpEF clinical trials [20, 23] have highlighted perturbed LA function as a marker of adverse outcomes. Using indices of LA strain measured by speckle tracking, LA dysfunction was independently associated with either prior [20] or subsequent [23] HF hospitalizations and death [25]. In a further retrospective TTE study involving both HFpEF and HFrEF, LAEF was independently associated with death only in HFpEF [21]. However, in the latter study, the groups were not evenly matched and controls comprised subjects referred for cardiac catheterization and were perhaps not truly representative of a healthy comparator group. In the one published CMR study to date evaluating the role of LA function in HF (heterogeneous population primarily comprising HFrEF), LAEF independently predicted mortality and incident AF. However, this retrospective study was again limited by referral bias, lacking a control group and excluding subjects who were in AF (nearly one-third) [12].

The potential value of LAEF as a prognostic biomarker may not be confined to HF alone. In a previous study of 312 subjects free of HF, who were in sinus rhythm and of similar age to our cohort, LAEF and LA strain were independent predictors of outcomes including future development of AF, HF and cardiovascular death [26]. All of the aforementioned studies however share intrinsic limitations of TTE [4].

Beyond HF, CMR data also further support LA dysfunction as a marker of outcomes. Similar to our findings, the incremental prognostic value of LA function beyond LAVImax has previously been shown in a prospective study of asymptomatic subjects from the general population [27] and in chronic hypertensive subjects without prevalent cardiovascular disease [28]. These findings suggest that LAEF also reflects a more advanced state of LA remodeling than LA dilation alone [1]. In another population study, LA strain using CMR feature tracking was independently associated with future development of incident heart failure [29].

The association between LA dilation and AF and their attendant cardiovascular risk is well recognized [1]. In HF, AF risk is also known to increase with diminishing LAEF [12]. Interestingly, in our study, AF was not associated with adverse outcomes even though event rates were higher in this sub-group and LAEF was significantly lower compared to those in sinus rhythm. This suggests that LAEF exerts its influence on outcomes through alternate mechanisms, either directly or indirectly [24, 27]. LA dysfunction as a mediator of pulmonary vascular damage, RV dysfunction and progressive biventricular failure has also been proposed [21]. Additional reports have also highlighted that LA dysfunction in the presence of AF has incremental thromboembolic and mortality risk, beyond the CHADS2 (congestive heart failure = 1, hypertension = 1, age ≥ 75 = 1, stroke/transient ischaemic attack = 2) score. Furthermore, LA dysfunction (using echo strain measures) predicts the success of restoring and maintaining sinus rhythm following either direct-current cardioversion or AF ablation [1].

Our study reaffirms the pathophysiological role of LA dysfunction in HFpEF. CMR-measured biplane LAEF is simple, reproducible and provides prognostic information which are strengths for consideration as a potential biomarker. CMR is becoming increasingly accessible and may more reliably discriminate breathless individuals with equivocal BNP levels and suboptimal echocardiographic imaging windows (especially HFpEF) [6, 17]. Recent data have also suggested that LA dysfunction may be a potential therapeutic target [24]. While our study suggests that iECV and LA function are not related, ongoing clinical trial data may shed further insight into whether myocardial fibrosis regression in HFpEF (with the anti-fibrotic agent Pirfenidone) may alter LA function as a secondary outcome measure [30].

This is a single centre study and the results should be confirmed in additional HFpEF cohorts. We used a pragmatic approach to define our HFpEF population to reflect a real world setting as opposed to latest ESC guidelines [6]. The presence of diastolic dysfunction was not a pre-requisite for study inclusion since recent contemporary clinical trial data have highlighted normal diastolic function in approximately a third of such patients [31]. Our data does however also provide compelling evidence (natriuretic peptides) that our patient cohort truly had HF and the event rates (48%) are similar to that of previous outcome studies in HFpEF. Our control group included hypertensive subjects and was therefore not totally free of cardiovascular disease. However, LAEF was also lower in HFpEF compared to both hypertensive and non-hypertensive controls.

CMR-derived LAEF is lower in HFpEF compared to age- and sex-matched controls and independently predicts outcomes.