Intra-study and inter-technique validation of cardiovascular magnetic resonance imaging derived left atrial ejection fraction as a prognostic biomarker in heart failure with preserved ejection fraction

The aim of this study was to assess the agreements of both biplane and short-axis Simpson’s (SAX) methods for left atrial ejection fraction (LAEF) calculation utilising cardiovascular magnetic resonance imaging (CMR) in heart failure with preserved ejection fraction (HFpEF) and evaluate their relation to clinical outcomes. One hundred and thirty six subjects (HFpEF n = 97, controls n = 39) underwent CMR, six-minute walk tests and blood sampling in our prospective, observational, single-centre study. Overall, LAEF (%) was lower in HFpEF patients compared to controls (SAX 34 ± 13 vs 47 ± 8, biplane 34 ± 16 vs 51 ± 11; p < 0.0001 for both). Atrial fibrillation (AF) was present in 24% of HFpEF and was associated with higher LA volumes and lower LAEF compared to sinus rhythm (p < 0.0001) with both methods. Biplane LAEF correlated strongly with SAX measurements (overall Pearson’s r = 0.851, sinus rhythm r = 0.651, AF r = 0.882; p < 0.0001). Biplane LAEF did not differ significantly compared to SAX LAEF (overall 34 ± 16 vs 34 ± 13%; p = 0.307) except in AF subjects in whom biplane LAEF was lower (mean difference 2 ± 4%, p = 0.013). There were 44 composite events (25 deaths, 19 HF hospitalizations) in HFpEF during median follow-up of 1429 days. LAEF below the median was associated with increased risk of composite endpoints (Log-Rank biplane p < 0.0001; SAX p = 0.009). In multivariable Cox proportional hazards regression analysis, both biplane LAEF (hazard ratio [HR] 0.604; 95% confidence interval [CI] (0.406–0.900); p = 0.013) and SAX LAEF (HR 0.636; CI 0.441–0.918; p = 0.016) remained independent predictors along with indexed extracellular volume. CMR LAEF, derived from either the short-axis or biplane method is lower in HFpEF compared to healthy controls and remains a strong marker of prognosis.


Introduction
In our recently published article [1], we reported that cardiovascular magnetic resonance (CMR) derived left atrial ejection fraction (LAEF) measured from the biplane method was lower in a well characterized cohort of heart failure with preserved ejection fraction (HFpEF) compared to healthy controls and was also independently associated with adverse outcomes. Furthermore, CMR biplane LAEF had excellent reproducibility. However, volumetric assessment of left atrial volumes (and hence function) derived from the shortaxis (Simpson's short-axis [SAX]) method is widely recognised as the imaging gold standard [2,3]. In a recent study of HFpEF patients, the biplane method was shown to have good correlation and agreements for LA volumetric analysis compared to the SAX method [4]. However, no prior CMR studies in HFpEF have compared LAEF between both methods. Furthermore, the prognostic role of SAX LAEF in HFpEF has not been reported. We aimed to assess the agreements of both methods for LAEF (and LA volumes) utilising CMR in HFpEF. We also evaluated whether SAX LAEF is also related to clinical outcomes, in order to validate and strengthen our previous findings implicating CMR LAEF as a prognostic biomarker in HFpEF.

Methods
From our original study cohorts [1] of HFpEF (n = 140) and healthy controls (n = 48), paired data for image analysis of both CMR biplane and SAX LAEF derivation was available in 136 subjects (HFpEF n = 97, controls n = 39). HFpEF was defined as clinical or radiographic evidence of heart failure and left ventricular ejection fraction > 50%. Exclusion criteria for HFpEF included: myocardial infarction in the preceding 6 months, suspected or confirmed cardiomyopathy or constrictive pericarditis, non-cardiovascular life expectancy < 6 months, severe chronic obstructive pulmonary disease (or forced expiratory volume [FEV 1 ] < 30% predicted or forced vital capacity [FVC] < 50% predicted), severe native valve disease and significant renal impairment (estimated glomerular filtration rate [eGFR] < 30 ml/min. The control population were age-and sex-matched compared to HFpEF and asymptomatic. As previously reported, hypertensive controls were also included since hypertension is highly prevalent in this age group of patients. Furthermore, hypertension is intimately linked with HFpEF development and is also reportedly associated with LA dysfunction [5] and we wanted to account for this potential confounder. Study recruitment, blood sampling, six-minute walk testing, imaging protocols and analytical methods for CMR have been detailed previously. For the SAX method (see Fig. 1), a contiguous stack of short-axis steady-state-free-precession images were acquired with retrospective ECG gating (or prospective gating in subjects with AF). Both end-diastolic and end-systolic frames were contoured to derive maximal left atrial volume [LAV max] and minimal left atrial volume (LAV min) and LAEF was calculated. Unlike the biplane method (see Fig. 2), SAX volumes were inclusive of the LA appendage but both techniques excluded pulmonary veins. The primary endpoint remained the composite of all-cause mortality or first HF hospitalization. Parameters with univariable association with the composite endpoint at p < 0.1 was entered into Cox proportional hazards regression analysis. In cases of collinearity, variables with historically stronger prognostic importance from published literature were chosen for inclusion in Cox regression analysis and   be compared according to one standard deviation increase, continuous predictor variables were Z-standardized. Event rates were calculated from Kaplan-Meier analysis. The Log-Rank test was used to detect differences in survival curves. In order to further assess the strength of both biplane and SAX LAEF in predicting outcomes, receiver operator characteristics (ROC) analyses were also performed.
Inter-technique comparison of both LAEF derivation methods included paired t-testing, regression and the Bland-Altman method. Ten randomly selected subjects were chosen to undergo intra-observer evaluation for both techniques, a minimum of 4 weeks apart. The two-way mixed-effect intraclass correlation co-efficient (ICC), coefficient of variation (COV) and Bland-Altman analysis were used for test intra-observer agreements.

Results
Baseline clinical and imaging characteristics are shown in Tables 1 and 2  1 3 min r = 0.884, LAEF r = 0.851), in sinus rhythm (LAV max 0.926, LAV min r = 0.920, LAEF r = 0.651) and moderate to good in AF (LAV max r = 0.776, LAV min r = 0.788, LAEF r = 0.882); p < 0.0001 for all. LA volumes calculated from the biplane method (excluding the LA appendage) were significantly lower compared to the SAX method overall and irrespective of whether subjects were in sinus rhythm or AF ( Table 2). While overall agreements for LA volumes were good (Table 3), the limits of agreement were wider in AF. Biplane LAEF did not differ significantly compared to SAX LAEF except in AF subjects in whom biplane LAEF was comparably lower (mean difference 2 ± 4%, p = 0.013). Intraobserver agreements for both methods were excellent, albeit biplane measures fared slightly worse (Table 4). Forty four composite events (45%, 25 deaths, 19 HF hospitalizations) were observed in the HFpEF group during median follow-up of 1429 days (1157-1657). Sixteen parameters were associated with outcomes during Cox regression analysis, including both biplane and SAX LAEF (Table 5). Urea and extracellular volume (ECV) were excluded from regression analysis due to collinearity. On Cox regression analysis (Table 6), both biplane and SAX LAEF remained significantly associated with outcome in 3 separate models incorporating clinical factors, biochemical markers and imaging parameters. In a final model comprising the strongest predictors overall, biplane LAEF (HR 0.604; 95% confidence interval [CI] (0.406-0.900); p = 0.013) and SAX LAEF (HR 0.636; CI 0.441-0.918; p = 0.016) remained independent predictors along with indexed extracellular volume (iECV). Irrespective of methodology, a lower LAEF group (below median) was associated with increased risk of the composite endpoint (biplane Log-Rank p < 0.0001; SAX Log-Rank p = 0.009); see Fig. 3. The area under curve for predicting outcomes was higher for SAX LAEF (0.709, p < 0.0001), albeit not significantly different compared to biplane LAEF (0.690, p = 0.001).

Discussion
Our study is the first to compare the agreements of both CMR biplane and SAX LAEF in HFpEF and assess their prognostic capabilities in the same setting. Our results confirm LAEF as an independent marker of prognosis in HFpEF, irrespective of CMR technique and validate our earlier findings of biplane LAEF as a strong prognostic biomarker [1]. The ability of CMR LAEF to be measured via 2 separate methods with a high degree of precision and reproducibility in addition to its prognostic capabilities are important strengths for consideration when being proposed as an imaging biomarker [6]. While CMR SAX LAEF is considered the more accurate of both measures [2,3], the CMR biplane method represents a potentially useful alternative for both research and clinical settings in HFpEF and has been recently proposed as the preferred method of LA volumetric analysis in an expert consensus document by the European Association of Cardiovascular Imaging [7]. Biplane LAEF derivation does not necessitate additional, multiple short axis image acquisitions which can prolong scan times in predominantly elderly subjects typical of HFpEF, who may struggle with multiple breath-holds. Furthermore, image analysis of the biplane method is also significantly shorter. On the other hand, SAX LAEF is an alternative if long axis images are degraded by artefact. Our findings are also important since LAEF represents a potential therapeutic target in HFpEF [8] where there are currently no effective therapies [9] and may also act as a trial endpoint [10].
While LA volumes were underestimated by the biplane method compared to SAX evaluation, LAEF did not differ significantly except in AF. Since LAEF is a fraction derived from volumetric analysis, systematic bias of similar magnitudes in the same direction i.e. underestimation of both LAV max and LAV min likely explains this lack of LAEF difference between both the techniques. The difference in LA volumes between both methods however is unsurprising given that the LA appendage is typically excluded from the biplane method but included in the SAX method [2,3].
These volumetric differences are likely further exaggerated in HFpEF where LA dilation is typical [2,4] and associated with a higher prevalence of AF. Furthermore, in AF we  employed prospective gating which can lower the margins for erroneous measurements. As addressed in our recent publication, we also recognise that having some subjects with hypertension in our control group is a limitation given that these subjects were not totally free from cardiovascular disease. However, LAEF was again lower in HFpEF with either methodology, irrespective of controls' hypertensive status, reinforcing altered LA contractile function in the pathophysiology of HFpEF.

Conclusions
CMR LAEF calculated from either the short-axis or biplane method has excellent reproducibility and remains a strong marker of prognosis in HFpEF.