Left atrial stiffness, exercise intolerance and injury biomarkers predict adverse outcomes in patients with heart failure with preserved ejection fraction Authors: Navazh Jalaludeen, Faye Forsyth, Jamie O’Driscoll, Jamie Edwards, Holly Pavey, Annette Hubsch, Joanna Helmy, Carmel M McEniery, Ian B Wilkinson, M Justin Zaman, Christi Deaton, Joseph Cheriyan Published: 2026 ABSTRACT AbstractHeart failure with preserved ejection fraction (HFpEF) is a complex condition associated with high morbidity and mortality. This exploratory study aimed to assess the prognostic value of left atrial (LA) stiffness, exercise tolerance, and selected biomarkers in patients with HFpEF. We evaluated LA mechanics using 2D speckle tracking in 43 patients (mean age: 77.2 years) from the OPTIMISE HFpEF study (NCT03617848). Biomarkers such as N-terminal pro-brain-type natriuretic peptide, growth differentiation factor 15, and galectin-3 were measured. Patients also completed a six-minute walk test. Over a mean follow-up of 24.5 months, there were five deaths (11.4%) and 21 all-cause hospitalisations (48.8%). LA stiffness (HR: 1.81; p = 0.028) and N-terminal pro-brain-type natriuretic peptide (HR: 1.01; p = 0.013) were independent mortality predictors. For hospitalisations, six-minute walk distance (HR: 0.99; p = 0.011) and growth differentiation factor 15 (HR: 1.01; p = 0.007) were significant independent predictors. This exploratory analysis indicates that LA stiffness, N-terminal pro-brain-type natriuretic peptide, exercise intolerance, and growth differentiation factor 15 independently predict adverse events in HFpEF. Implementing a comprehensive therapeutic approach, including lifestyle modifications, medical therapy, and, if necessary, more advanced interventions to improve LA mechanics, functional capacity, and clinical biomarkers, may improve outcomes for HFpEF patients.Subject terms: Circulation, Cardiovascular diseases Introduction Heart failure with preserved ejection fraction (HFpEF) is a complex, heterogeneous clinical syndrome accounting for approximately half of all heart failure patients1. Commonly presenting with symptoms such as exertional dyspnoea, fatigue and fluid retention, HFpEF carries significant implications for quality of life and outcomes in affected patients2. Indeed, community-based studies commonly evidence poor health-related quality of life scores3 and 5-year mortality rates upwards of 50%4,5. Despite advances in the understanding of its pathophysiological mechanisms, HFpEF remains largely under-recognised and under-diagnosed1,2,6. Left ventricular diastolic dysfunction and left atrial (LA) pressures are key in the development of HFpEF2,7. As a result of elevated left ventricular (LV) filling pressures, the LA is subject to increases in pressure and wall tension, resulting in well-documented structural and functional maladaptation. Consequently, LA stiffness is commonly increased in HFpEF as shown in speckle tracking echocardiographic studies of LA reservoir mechanics8. This has resulted in the inclusion of the LA reservoir strain in the recent European Association of Cardiovascular Imaging recommendations for multimodality imaging in HFpEF9. Given the imperative role of the LA in the regulation of LV filling and holistic cardiac performance, several studies have also demonstrated a positive correlation between LA stiffness and a reduction in exercise capacity, which remains an independent hallmark of HFpEF1. Despite the prevalence and pathophysiological importance of LA stiffness and exercise intolerance in HFpEF, the prognostic utility of these parameters is not entirely clear. Similarly, while some biomarkers, such as BNP, have well-established roles in heart failure, the prognostic value of different clinical biomarkers in HFpEF is not entirely understood. Growth differentiation factor 15 (GDF-15) has emerged as a potentially important parameter in HFpEF due to its up-regulated expression with stressors such as inflammation and oxidative stress, which are distinct pathophysiological characteristics of this population10. Galectin-3, which is secreted by activated macrophages, is strongly implicated in the regulation of pro-fibrotic and pro-inflammatory pathways and has also been shown to be elevated in HFpEF patients11. Similar to LA stiffness, Galectin-3 is also associated with exercise intolerance in heart failure patients12. We hypothesise that LA stiffness represents an integrative haemodynamic phenotype in HFpEF that reflects chronically elevated LV filling pressures and impaired atrial compliance, which in turn contributes to reduced exercise capacity and adverse clinical outcomes. Furthermore, we postulated that circulating biomarkers of myocardial stress, inflammation and fibrosis would track with this atrial remodelling process, providing complementary biological insight into the mechanisms linking LA dysfunction to prognosis. Despite previous work examining the independent impact of LA stiffness, exercise intolerance and different biomarkers on adverse events13,14, to our knowledge, no work to date has analysed these parameters together in the same cohort. As such, the present study was designed as an exploratory multimodal, hypothesis-generating investigation to assess the prognostic utility of LA stiffness, exercise tolerance as assessed via a six-minute walk test (6MWT) and biomarkers including GDF-15 and galectin-3 in patients with HFpEF. Results Hospitalisation and all-cause mortality rates were determined at the end of a mean follow-up period of 24.5 ± 4.4 months. Data for all 43 analysed HFpEF patients, separated by hospitalisation and all-cause mortality, can be found in Tables 1 and 2, respectively. Table 1Characteristics of patients according to hospital admissionCharacteristicNo admission (n = 22)Admission (n = 21)p ValueDemographics Age (years)74 ± 7.681 ± 5.5<0.001 Height (cm)1.64 ± 0.11.64 ± 0.10.433 Weight (kg)88.8 ± 19.688.2 ± 19.00.916 BMI (m2)32.6 ± 5.732.8 ± 6.30.924 Sex0.758 Female (%)11 (50%)10 (47.6%) Male (%)11 (50%)11 (52.4%)Haemodynamics Systolic blood pressure (mmHg)143 ± 25147 ± 330.687 Diastolic blood pressure (mmHg)80 ± 1378 ± 120.605 Mean blood pressure (mmHg)97 ± 26101 ± 180.518Traditional echocardiography Septal E’6.6 ± 2.05.8 ± 1.10.110 Lateral E’10.0 ± 3.78.1 ± 1.60.040 E/E’ ratio10.6 ± 4.414.6 ± 5.50.013 LA volume46.8 ± 21.446.0 ± 16.30.910 LVEF (%)59.8 ± 8.658.8 ± 8.70.710LA and LV mechanics LA reservoir (%)15.7 ± 8.715.3 ± 8.90.908 LA conduit (%)8.1 ± 3.58.3 ± 4.40.826 LA contractile (%)11.7 ± 5.79.3 ± 5.40.271 LA stiffness index1.0 ± 0.91.4 ± 1.30.269 LV GLS (%)16.6 ± 4.016.7 ± 3.70.902 Basal and mid LV GLS (%)15.4 ± 3.614.7 ± 3.60.542 Apical LV GLS (%)18.9 ± 6.620.1 ± 5.20.540 LV strain ratio1.2 ± 0.31.4 ± 0.30.106Functional status NYHA classification0.058 I10 (45.5)3 (14.3) II11 (50)14 (66.7) III1 (4.5)4 (19.0) 6-minute walk distance (m)397.5 ± 120.2211.8 ± 126.20.02Adverse events All-cause mortality0 (0%)5 (23.8%)0.015LA left atrial, LVEF left ventricular ejection fraction, GLS global longitudinal strain, NYHA New York Heart Association. Characteristics of patients according to hospital admission LA left atrial, LVEF left ventricular ejection fraction, GLS global longitudinal strain, NYHA New York Heart Association. Table 2Characteristics of patients according to alive and all-cause mortalityCharacteristicAlive (n = 38)All-cause mortality (n = 5)p ValueDemographics Age (years)77 ± 7.582 ± 5.80.169 Height (cm)1.64 ± 0.11.69 ± 0.10.414 Weight (kg)88.3 ± 18.890.0 ± 24.50.869 BMI (m2)32.8 ± 5.831.5 ± 8.20.671 Sex0.153 Female (%)20 (52.6%)1 (20%) Male (%)18 (47.4%)4 (80%)Haemodynamics Systolic blood pressure (mmHg)148 ± 29122 ± 180.028 Diastolic blood pressure (mmHg)80 ± 1272 ± 110.165 Mean blood pressure (mmHg)100 ± 2389 ± 130.284Traditional echocardiography Septal E’6.2 ± 1.75.6 ± 1.50.416 Lateral E’8.9 ± 3.19.4 ± 1.10.765 E/E’ ratio12.3±4.715.0 ± 9.20.286 LA volume45.0 ± 17.256.0 ± 28.50.224 LVEF (%)59.9 ± 8.454.8 ± 9.20.129LA and LV mechanics LA reservoir (%)16.3 ± 8.79.6 ± 6.00.05 LA conduit (%)8.7 ± 3.94.8 ± 2.90.038 LA contractile (%)10.6 ± 5.68.0 ± 7.10.536 LA stiffness index1.1 ± 1.02.1 ± 1.80.032 LV GLS (%)17.1 ± 3.713.5 ± 2.50.023 Basal and mid LV GLS (%)15.5 ± 3.511.6 ± 2.70.02 Apical LV GLS (%)19.8 ± 6.116.9 ± 3.90.313 LV strain ratio1.3 ± 0.31.5 ± 0.20.269Functional status NYHA classification0.495 I13 (34.2%)0 (0%) II21 (55.3%)4 (80%) III4 (10.5%)1 (20%) 6-minute walk distance (m)311.2 ± 143.3170.0 ± 89.30.04LA left atrial, LVEF left ventricular ejection fraction, GLS global longitudinal strain, NYHA New York Heart Association. Characteristics of patients according to alive and all-cause mortality LA left atrial, LVEF left ventricular ejection fraction, GLS global longitudinal strain, NYHA New York Heart Association. Hospitalisation Twenty-one patients (48.8%) were hospitalised during follow-up. The hospitalised patients were also significantly older (81 ± 5.5 vs. 74 ± 7.6, P < 0.001), with a higher E/E’ ratio (14.6 ± 5.5 vs. 10.6 ± 4.4, P = 0.013), and lower Lateral E’ (8.1 ± 1.6 vs. 10.0 ± 3.7, P = 0.040) and 6MWD (211.8 ± 126.2 m vs. 397.5 ± 120.2 m, P = 0.02). As seen in Table 3, hospitalised patients also had a significantly higher GDF-15 (3900 ± 3276 vs. 2180 ± 1133 pg/mL, p = 0.02) and IL-6 (4.1 ± 3.1 vs. 1.8 ± 1.4 pg/mL, p = 0.004). There were no other significant differences between the hospitalised and non-hospitalised patients. Importantly, 6MWD (HR: 0.99; 95% CI: 0.98–0.99, p = 0.004) and GDF-15 (HR: 1.01; 95% CI: 1.00–1.01, p = 0.007) were independent predictors for all-cause hospitalisation (Table 4 and Fig. 1). Table 3Biomarker differences according to hospitalisation and all-cause mortalityBiomarkerNo hospital admissionHospital admissionP valueAliveAll-cause mortalityP valueFerritin (μg/L)136 ± 112170 ± 1700.510130 ± 102237 ± 2400.104Fibroblast factor 21 (pg/mL)374 ± 247823 ± 14480.182397 ± 2451243 ± 21830.024Galectin 3 (ng/mL)15.3 ± 7.915.2 ± 3.70.99115.5 ± 6.913.8 ± 5.60.648Growth differentiation factor 15 (pg/mL)2180 ± 11333900 ± 32760.022399 ± 13164825 ± 46030.013High sensitivity C-reactive protein (mg/L)6.8 ± 16.44.4 ± 4.10.6336.6 ± 14.52.8 ± 2.90.566Interferon gamma (pg/mL)8.9 ± 10.414.4 ± 15.40.25510.9 ± 13.410.9 ± 6.10.994Interleukin-6 (pg/mL)1.8 ± 1.44.1 ± 3.10.0042.6 ± 2.52.8 ± 1.60.888Iron (μmol/L)17.7 ± 4.514±5.10.02916.4 ± 4.215.1 ± 5.30.617N-terminal brain natriuretic peptide (pg/mL)576 ± 5981321 ± 18850.065632 ± 5862257 ± 30420.002Total iron binding capacity (μg/dL)348 ± 46368 ± 510.265352 ± 48371 ± 490.464Transferrin (g/L)2.5 ± 0.42.7 ± 0.40.2132.5 ± 0.42.7 ± 0.30.360Tumour necrosis factor alpha (pg/mL)3.4 ± 4.53.3 ± 1.80.9283.5 ± 4.12.7 ± 1.70.659 Biomarker differences according to hospitalisation and all-cause mortality Table 4Multivariable Cox regression analyses for clinical outcomesOutcomeVariableHazard ratio95% confidence intervalsp valueAll-cause hospitalisation6-minute walk distance (m)0.990.98–0.990.004GDF-15 (pg/mL)1.011.00–1.010.007All-cause mortalityLA stiffness index1.811.07–3.060.028NT-proBNP (pg/mL)1.011.00–1.020.013GDF-15 growth differentiation factor 15, LA left atrial, NT-proBNP N-terminal brain natriuretic peptide. Multivariable Cox regression analyses for clinical outcomes GDF-15 growth differentiation factor 15, LA left atrial, NT-proBNP N-terminal brain natriuretic peptide. Associations between left atrial stiffness, 6-minute walk distance, circulating biomarkers, and adverse outcomes in HFpEF over >2 years of follow-up. In multivariable Cox models, left atrial stiffness (HR: 1.81 per unit increase) and NT-proBNP (HR: 1.01 per unit increase) were associated with mortality, while six-minute walk distance (HR: 0.99 per metre increase) and GDF-15 (HR: 1.01 per unit increase) were associated with hospitalisation. Hazard ratios reflect the relative hazard per unit increase. Findings are exploratory. All-cause mortality Five patients (11.4%) died in the follow-up period. All deaths occurred in patients who had experienced hospitalisation during follow-up. The all-cause mortality group had a significantly lower systolic blood pressure (122 ± 18 mmHg vs. 148 ± 29 mmHg, p = 0.028) and 6MWD (170.0 ± 89.3 m vs. 311.2 ± 143.3 m, P = 0.04), and a higher NT-proBNP (2256.8 ± 3042.8 pg/mL vs. 632.7 ± 586.8 pg/mL, p = 0.004) compared with patients who remained alive at follow-up. Analysis of LA and LV mechanics also demonstrated several significant differences, including a lower LA reservoir (9.6 ± 6.0% vs. 16.3 ± 8.7%, p = 0.05), LA conduit (4.8 ± 2.9% vs. 8.7 ± 3.9%, p = 0.038), LV GLS (13.5 ± 2.5% vs. 17.1 ± 3.7%, p = 0.023), basal and mid LV GLS (11.6 ± 2.7% vs. 15.5 ± 3.5%, p = 0.02) and a higher LA stiffness index (2.1 ± 1.8 vs. 1.1 ± 1.0, p = 0.032), in the all-cause mortality compared to the alive patients. As seen in Table 3, there was a significantly higher Fibroblast factor 21 (1243 ± 2183 vs. 397 ± 245 pg/mL, p = 0.024) and GDF-15 (4825 ± 4603 vs. 2399 ± 1316 pg/mL, p = 0.02) in the all-cause mortality group compared to the alive patients. There were no other significant differences between the all-cause mortality and the alive patients. LA stiffness (HR: 1.81; 95% CI: 1.07–3.06, p = 0.028) and NT-proBNP (HR: 1.01; 95% CI: 1.00–1.02, p = 0.013) were independent predictors of all-cause mortality (Table 4 and Fig. 1). Correlation analysis There was a statistically significant association of LA stiffness (r = −0.43; p = 0.007), LA coupling index (−0.51; p = 0.011) and relative regional LV strain ratio (−0.42; p = 0.008) with the 6MWD. There was also a statistically significant association between NT-proBNP with LA reservoir strain (−0.44; p = 0.004); LA contractility (−0.53; p = 0.003); GLS (−0.34; p = 0.031); LA volume (0.43; p = 0.006); left atrial coupling index (LACVI) septal (0.76; P < 0.001); LACVI lateral (0.64; P < 0.001); LACVI Average (0.71; p < 0.001); basal mid strain (−0.40; p = 0.012); LA stiffness (0.56; p < 0.001); and GDF-15 (0.80; p < 0.001). There was a statistically significant association between IL-6 and mitral E/E’ (0.37; p = 0.0460); DD (0.53; p = 0.042); ventricular arterial coupling (0.42; p = 0.022); LACVI Septal (0.63; p = 0.010); LACVI Lateral (0.80; p < 0.001); LACVI Average (0.75; P < 0.001); and LV Strain ratio (0.61; p < 0.001). Finally, there was a statistically significant association between GDF-15 with mitral E/E’ (0.64; p < 0.001); TAPSE (−0.48; p = 0.008); LA stiffness (0.44 l; p = 0.014); and BNP (0.80; p < 0.001). There were no other statistically significant correlations. Discussion This study aimed to investigate the prognostic utility of LA stiffness, exercise tolerance and selected clinical biomarkers of inflammation, extracellular fibrosis, iron deficiency and renal function in patients with HFpEF. Importantly, our results confirm LA stiffness and NT-proBNP as independent predictors of all-cause mortality and show that 6MWD and GDF-15 are independent predictors of all-cause hospitalisation. LA stiffness, LA coupling index and LV relative regional LV strain ratio were significantly associated with 6MWD. NT-proBNP, IL-6 and GDF-15 were also associated with a range of clinical parameters. We present important data to demonstrate that these parameters are sensitive indicators for clinical outcomes in HFpEF and should be studied in greater detail in future observational and interventional studies. The effects of therapeutic interventions on these clinical physiological biomarkers require further study, as they may represent early features of benefit with such interventions. Our data demonstrate that LA stiffness is an independent predictor of mortality. Specifically, higher LA stiffness at baseline was associated with a significantly increased hazard of all-cause mortality (HR 1.81 per unit increase). This result confirms previous findings that LA stiffness is an important parameter in the risk stratification of HFpEF patients. Continuous or intermittent elevations in LV filling pressure can trigger fibrotic alterations in the LA, leading to decreased LA compliance and increased LA stiffness. While abnormalities in LA function are often more predictive of negative consequences compared to those in the LV, it is noteworthy that LV dysfunction typically precedes LA dysfunction in the majority of HFpEF cases2,7. Nonetheless, a subgroup of patients with HFpEF may develop primary LA myopathy, leading to elevated LA pressure and reduced LV filling, ultimately resulting in insufficient cardiac output augmentation during exertion15. Our findings are similar to those of Kim et al.13 who recently observed an independent, multivariable adjusted association between LA stiffness index and composite outcome in a larger HFpEF sample. Combined with that of Kim et al.13, this work conflicts with previous suggestions that an association between LA dysfunction and worse outcomes in HFpEF are explained by concomitant impairments in LV systolic and diastolic function16, in particular for heart failure hospitalisation. Further, these results potentially suggest the utility of LA stiffness as a physiological biomarker in monitoring responses to therapeutic interventions, including non-pharmacological interventions such as exercise training. Although the effects of exercise training on cardiac parameters in HFpEF are generally underwhelming17, to our knowledge, no published research to date has investigated its effects on LA stiffness. Similarly, NT-proBNP was independently associated with all-cause mortality in this cohort. The strong association between NT-proBNP and outcomes in HFpEF is best highlighted in the PARAGON-HF trial, where the highest NT-proBNP quartile (≥1617 ng/l) had almost 3 times the risk of mortality compared to patients in the lowest quartile (≤464 ng/l)18. The reason for the smaller effect size in this work is unclear, but it is likely influenced by the small population analysed and inter-participant heterogeneity, given the influence of confounding patient comorbidities on NT-proBNP19. Regardless, this study further supports the prognostic capacity of this biomarker in patients with HFpEF. This study also evidences the independent predictive capacity of the 6MWT in all-cause hospitalisation. The prevalence and clinical significance of poor functional capacity in HFpEF patients are well-established. Previous work from Zotter-Tufaro et al.20 also reported the 6MWD to be an independent predictor of composite outcome (hospitalisation and/or cardiac death); however, the lack of predictive capacity in all-cause mortality in this study is likely attributable to the smaller sample size. Notwithstanding this limitation, our findings provide further support for the implementation of therapeutic interventions aimed at improving exercise capacity. Recent meta-analyses have reported significant improvements in exercise capacity following exercise training in HFpEF21, with particular support for the role of high-intensity interval training22. Specifically, Edwards et al.22 reported a significant increase in 6MWD following exercise training by 35.4 m, a figure that is considered clinically relevant23 and thus, according to the present analysis, may ultimately translate into fewer hospitalisations. Furthermore, 6MWD is also closely associated with improvements in quality of life in HFpEF24. As such, improvements in functional capacity should remain a priority in HFpEF patient management2,7. Lastly, patients with heart failure define quality of life as the ability to perform activities, pursue happiness and relationships with others25, suggesting efforts to improve functional capacity are cognisant of their preferences. The correlation analysis in this study demonstrated a significant inverse association between LA stiffness, LA coupling index, and relative regional LV strain ratio with exercise capacity. This is supported by previous work by Von Roeder et al.26, who reported a significant correlation between LA stiffness and aerobic capacity, which further highlights the importance of the LA in overall cardiac performance and functional capacity. Separately, longitudinal assessment of LA structure and function has been shown to provide incremental prognostic value in heart failure, reflecting the dynamic interaction between filling pressures, atrial compliance and therapeutic response27. In this context, LA stiffness may represent not only a marker of disease severity, but we hypothesise that it could also be a potential mechanistic determinant of left atrial reverse remodelling, as reductions in filling pressures and myocardial fibrosis could improve atrial compliance and reservoir function28. Our data suggest that further prospective evaluation of whether contemporary HFpEF therapies or exercise therapies modify LA stiffness, thereby resulting in clinical benefits, needs to be considered in the future. GDF-15 was an independent predictor of all-cause hospitalisation, supporting the utility of this emerging biomarker in the prognostication of patients with HFpEF29. This finding is generally in line with previous work, which has evidenced its independent prognostic capacity in a range of cardiovascular diseases, including HFpEF29–31. Indeed, Fernandez et al.30 evidenced GDF-15 as a superior prognostic marker to NT-proBNP for HFpEF patients. Furthermore, recent work from Lyu et al.10 supported its utility as an HFpEF-specific risk factor, suggesting its integration into risk score-based models to enhance prognostic capability. Interestingly, despite its prediction of all-hospitalisation, this study did not identify GDF-15 as an independent predictor of all-cause mortality. Similarly, galectin-3 was not evidenced as predictive of either hospitalisation or all-cause mortality, generally conflicting with previous findings11,32. Like the smaller effect size observed for NT-proBNP, these findings are possibly a result of the small sample size and co-concomitant inter-individual variability of such sensitive biomarkers. Ultimately, a larger-scale study with greater statistical power is needed to understand the role of these promising biomarkers in this under-recognised population. This work is limited by a small sample size, with subsequent implications for multivariate analysis. Given the modest sample size and low number of outcome events relative to the number of candidate variables, the statistical power of the multivariable models is substantially limited, and the risk of type II error is increased. In addition, the events-per-variable ratio was below that typically recommended for Cox regression, raising the possibility of model overfitting and unstable effect estimates with wide confidence intervals. Although the multivariable models were constructed cautiously with a limited number of clinically relevant covariates to reduce the risk of overfitting, the findings should be interpreted as preliminary rather than definitive. Consequently, the independence and magnitude of the identified predictors require confirmation in larger, adequately powered cohorts before generalisation to the wider HFpEF population. The HFpEF patients analysed in this study are also heterogeneous and vary amongst sub-domains of HFpEF classifications. As such, larger-scale observational trials with the capacity for sufficient participant sub-grouping via HFpEF phenotype are needed to further explore the prognostic capacity of the explored parameters. It is unclear how the variability and reliability of LA stiffness may have influenced these findings and the relevance of this to clinical application. There are also inherent limitations with correlational analyses, which are exacerbated by the small sample size. In conclusion, LA stiffness, exercise intolerance, NT-proBNP and GDF-15 are independent predictors of adverse events in patients with HFpEF. Elevated LA stiffness may be an important haemodynamic signal of reduced compliance that negatively impacts cardiovascular physiology. The addition of LA stiffness calculations during echocardiographic assessment may provide additional prognostic information in HFpEF, pending validation in larger cohorts. Larger studies are required to determine if the addition of LA stiffness better predicts outcomes than NT-proBNP or 6MWD alone. Additionally, larger studies with greater statistical power are needed to better understand the independent and combined prognostic utility of clinical biomarkers such as GDF-15 and galectin-3. This potentially could lead to the development of an HFpEF risk score incorporating some of these physiological parameters to identify those at risk and intensify therapeutic management. Attenuated functional capacity is a well-established marker of adverse outcomes. As such, implementing a comprehensive therapeutic approach including lifestyle modifications, medical therapy, and, if necessary, more advanced interventions to improve LA mechanics, functional capacity and clinical biomarkers may improve outcomes for HFpEF patients. Methods Study design and participants The rationale and protocol have been previously published33. In brief, OPTIMISE HFpEF was a multi-phase programme of work that sought to: (1) explore the effectiveness of multi-disciplinary management of HFpEF34; (2) understand problems in HFpEF diagnosis and management35,36; (3) investigate the feasibility of recruiting patients with HFpEF from primary care and explore the utility of contemporary diagnostic pathways; (4) to characterise and prospectively follow a sample of community-dwelling patients with HFpEF23,24. This study is an exploratory, retrospective post hoc analysis of the original OPTIMISE HFpEF cohort, an academic observational study33. This observational study had a favourable ethical opinion from the London—Surrey Research Ethics Committee (REC reference: 17/LO/2136) and approval by the UK’s Health Regulatory Agency. All participants provided written informed consent; data were handled in line with local data protection and information governance regulations, which stipulate that personal identifiable data is stored separately and securely. The study conformed with the principles of the Declaration of Helsinki. This secondary analysis utilises data relating to patients with a confirmed diagnosis of HFpEF from the latter prospective observational study. Patients recruited were assessed at baseline, and positive HFpEF cases were followed up at 6 and 12 months. Hospital Episode Statistics and mortality data were obtained from NHS England datasets (formally NHS Digital) during the entire duration of the cohort study. The HFpEF cohort was characterised by a high burden of cardiometabolic comorbidity typical of this population. Hypertension was present in 38 patients (88%), with 29 patients receiving ≥2 antihypertensive medications. Obesity (BMI > 30 kg/m²) was observed in 30 patients (70%), while diabetes mellitus was present in 17 patients (40%), including 4 with evidence of end-organ damage. Atrial fibrillation was present in 16 patients (37%), pulmonary hypertension on echocardiography (estimated SPAP > 35 mmHg) in 14 patients (33%), chronic lung disease in 13 patients (30%), and moderate-to-severe renal disease in 11 patients (26%). Patients were treated with contemporary cardiovascular therapies including beta blockers (n = 25), statins (n = 25), loop diuretics (n = 22), calcium channel blockers (n = 17), angiotensin-converting enzyme inhibitors (n = 15), angiotensin receptor blockers (n = 12), and anti-arrhythmic medications (n = 10). HFpEF probability assessed using the HFA-PEFF score demonstrated a mean score of 5.19 (median: 5; range: 2–9), with the majority of patients classified as high probability HFpEF (score: ≥5, n = 26) and the remainder classified as intermediate probability (score: 2–4, n = 16). Functional capacity Functional capacity was assessed via the 6MWT at baseline. The test was performed in accordance with American Thoracic Society guidelines37. Assessment was performed at the end of the evaluation to avoid influencing echocardiography metrics. Echocardiography and strain analysis All participants underwent comprehensive two-dimensional transthoracic echocardiography with Doppler and Tissue Doppler imaging at the time of the study enrolment, with systolic and diastolic function graded according to contemporaneous European Society of Cardiology guidelines38,39. Images were acquired at end expiration in the left lateral decubitus position using a commercially available GE Vivid E9 machine. For each acquisition, three cardiac cycles of non-compressed data at a frame rate between 60 and 90 frames s−1 were stored in cine-loop format for offline analysis. LA deformation was assessed from the septal and lateral walls of the left atrium using dedicated atrial strain software (GE EchoPac; Version 206 workstation). Apical four- and two-chamber views were acquired and optimised specifically for left atrial assessment, ensuring full visualisation of LA length and avoidance of roof truncation, even where this differed from a conventional LV-optimised view. The software automatically generates a region of interest (ROI) tracing of the left atrium. If tracking of the LA myocardium was inadequate, the ROI was manually adjusted to enhance tracking. Speckle tracking analysis was used to derive LA phasic mechanics, including reservoir, conduit and contractile strain was calculated for analysis. LA stiffness was calculated as (E/E’)/reservoir strain, and left ventricular global longitudinal strain was calculated as previously recommended40. The LA volumetric/mechanical coupling index was calculated as the ratio of left atrial volume index over late diastolic mitral annulus velocity41. Only studies with adequate image quality and sufficient endocardial border definition to permit reliable speckle-tracking analysis were included. LA strain analysis was performed offline by a single experienced and accredited echocardiographer (NJ), blinded to clinical outcomes. To assess intra-observer reproducibility, a randomly selected subset of studies (n = 7) was re-analysed. Reproducibility demonstrated excellent agreement for LA stiffness (intraclass correlation coefficient [ICC]: 0.94) and good agreement for LA reservoir strain (ICC: 0.80). Biomarker analysis A comprehensive panel of bloods was obtained to characterise the cohort, including ferritin, fibroblast factor 21, galectin-3, GDF-15, high sensitivity C-reactive protein, interferon gamma, interleukin-6 (IL-6), iron, N-terminal brain natriuretic peptide (NT-proBNP), total iron binding capacity, transferrin and tumour necrosis factor alpha sampling at baseline. Serum and plasma samples were banked for future analyses of emerging biomarkers. Statistical analysis Continuous variables are expressed as mean ± standard deviation and categorical variables as n (%). Group comparisons were based on a 2-sample t-test for continuous variables, and a χ2 test was used for categorical variables. Multivariable adjusted Cox proportional hazard models were constructed to ascertain predictors of all-cause mortality and hospitalisation for heart failure symptoms. For model building, significant univariate parameters were evaluated for their association with all-cause mortality and hospitalisation. Forward stepwise selection procedures were used to compare models for goodness-of-fit, and a P value of <0.01 was used for retention in the final model. Hazard ratios (HR) and corresponding 95% confidence intervals (CI) are reported. We used Pearson correlation coefficient analysis to examine the relationship between NT-proBNP, IL-6, iron and GDF-15 with all included parameters. The relationship between LA and LV mechanics with functional capacity was also analysed. Given the exploratory nature of biomarker comparisons and correlation analyses, no formal correction for multiple testing was applied; and our findings in this small cohort should therefore be interpreted purely as hypothesis-generating. All analyses were conducted using the Statistical Package for Social Sciences (SPSS 29 release version of SPSS for Windows; SPSS Inc., Chicago, IL, USA). Acknowledgements This research was supported by the National Institute for Health and Care Research (NIHR) Cambridge Biomedical Research Centre (NIHR203312). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. Author contributions N.J., F.F., H.P., A.H., J.H., C.M., I.B.W., M.J.Z., C.D. and J.C. contributed to the conception or design of the work. N.J., F.F., J.O., J.E., H.P., A.H., J.H., C.M., I.B.W., M.J.Z., C.D. and J.C. contributed to the acquisition, analysis, or interpretation of data for the work. N.J., F.F., J.O. and J.E drafted the manuscript. N.J., F.F., J.O., J.E., H.P., A.H., J.H., C.M., I.B.W., M.J.Z., C.D. and J.C. critically revised the manuscript. Data availability The datasets generated and/or analysed during the current study are not publicly available due to patient confidentiality and data protection regulations, but are available from the corresponding author on reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors jointly supervised this work: Christi Deaton, Joseph Cheriyan. References Associated Data Data Availability Statement The datasets generated and/or analysed during the current study are not publicly available due to patient confidentiality and data protection regulations, but are available from the corresponding author on reasonable request.