Echocardiography was performed according to current recommendations of the European Association of Cardiovascular Imaging and the American Society of Echocardiography by experienced cardiologists on the same day of the right heart catheterization (19), using a Vivid E9/E95 scanners (GE Vingmed, Horten, Norway). LV and left atrial (LA) volumes were measured using biplane disk summation algorithm on dedicated 4-chamber and 2-chamber apical views, taking care to avoid chamber foreshortening. Right atrial (RA) volume was measured in a right ventricle (RV) -focused 4-chamber view. Echocardiographic evaluation of STR severity was based on an integrative approach considering multiple qualitative and quantitative parameters (6, 20, 21). Due to the retrospective nature of the study, RV dimensions were qualitatively estimated in all patients.

Patients were studied on optimized medical therapy and in euvolemic state, in non-fasting state, without sedation, and in supine position. They wore a non-rebreathing Hans-Rudolph mask connected to the V-MAX metabolic cart (Vmax SensorMedics 2200, Yorba Linda, CA, USA) to directly measure gas-exchange data and ventilation (17). A 7-F fluid-filled Swan-Ganz catheter was placed in the pulmonary artery through the right internal jugular vein under fluoroscopic guidance. Proper pulmonary artery wedge positioning was confirmed by the appearance of a typical PAWP trace as well as by an oxygen saturation >94% sampled at the tip of the catheter. The right radial artery was cannulated with the Seldinger technique. The transducer was zeroed at the midthoracic line, halfway between the anterior sternum and the bed surface. Hemodynamic measurements were performed at rest, after 1 min of passive leg raise (feet on the pedals), and during the last minute of each step of a symptom-limited, maximal exercise test (18). Two milliliters of blood were sampled at the same time from the tip of the Swan-Ganz catheter and from the radial artery for blood gases analysis. The increment in workload was personalized in order to obtain at least three steps of exercise before exhaustion. Subjects were encouraged to exercise up to their maximal volitional effort. Pulmonary artery pressure, PAWP and RA pressure (RAP) were reported as an average of several respiratory cycles (18). Hemodynamic data reflect the agreement of two readers who visually reviewed all pressure traces. CO was calculated by direct Fick method, solving the oxygen consumption (VO₂) equation as follows: CO = VO₂/C(a-v)O₂, where C(a-v)O₂ is the oxygen arteriovenous difference. Furthermore, to evaluate the relative contribution of the elements of the Fick equation to exercise capacity, we plotted CO as a function of C(a-v)O₂ on which we represented VO₂-isophlets (22).

LV trans-mural pressure (LVTMP), as a measure of LV preload, independent of right heart filling and pericardial restraint, was calculated as PAWP–RAP (23). We plotted the relationship between pulmonary vascular resistance (PVR) and stroke volume (SV) to explore whether low anterograde SV may be linked to pulmonary vascular derecruitment and higher than normal PVR (24).

Key ergospirometric measurements included standard breath-by-breath cardiorespiratory and breathing pattern parameters. Peak VO₂ was measured as the highest 30-s value obtained at the end of the effort. The slope of the relationship between minute ventilation and carbon dioxide production (VE/VCO₂ slope) was calculated over the linear component of VE vs. VCO₂ (25).

Continuous variables are reported as mean and standard deviation (SD) or median and interquartile range if the data did not follow a normal distribution. The categorical variables are shown as absolute frequencies and proportions. For data at rest, unpaired T-test (or Wilcoxon signed rank sum test in case of non-normal distribution) was applied to compare the continuous variables between HFpEF-STR and HFpEF-controls, while Chi-square test (or Fisher test) was used to compare the categorical variables. For each hemodynamic variable measured during exercise, an ANOVA model for repeated measures was fitted, considering an unstructured variance-covariance matrix to take into account the correlation among measurements on the same subjects. The included covariates in each model were group (HFpEF-STR or HFpEF-controls), time (Rest, Leg Raise and Peak) and their interaction. The statistical significance of interaction term suggested a different trend of the hemodynamic variables between groups. Moreover, we tested the least square means (LS-means) differences among groups at each time by means of unpaired t-test applying the False Discovery Rate (FDR) approach to control the inflation of the type I error.

The relationship between continuous hemodynamic variables at specific time-point was investigated by means quadratic B-spline with 1 knot. The goodness of fit of the model was estimated by means of R2. This value, ranging from 0 to 1, represents the proportion of total variance of an independent variable explained by a dependent variable.

All analyses were performed using SAS version 9.4 software (SAS Institute, Cary, NC, USA). Statistical significance was set at the 0.05 level. All P-values were two-sided.

Patients' selection flowchart is depicted in Figure 1. HFpEF-STR represented 14% of our HFpEF population who had undergone exercise right heart catheterization during the study period. All HFpEF-STR patients but one (who had a typical HFpEF profile, but presented with a systolic PAP >50 mmHg) fulfilled the criteria for atrial-STR. In order to provide results on a more homogeneous patients' phenotype, we excluded from the analysis the patient that presented with a ventricular-STR. Indeed, the analyses focused on HFpEF with atrial-STR as compared with HFpEF-controls.

The clinical characteristics of our study groups are summarized in Table 1. They represented a quite typical elderly, overweight and predominantly female HFpEF population. Of note, 83% of patients with HFpEF-STR had permanent atrial fibrillation, as compared with only 8% of permanent atrial fibrillation in HFpEF-controls (p < 0.01). The rhythm at the time of right heart catheterization was atrial fibrillation in 83% of HFpEF-STR and 8% of HFpEF-controls. Among the other cardiovascular comorbidities, arterial hypertension was the most common affecting 75% of HFpEF-controls and 83% of HFpEF-STR patients (p = 1.000).

Representation of HF signs and symptoms did not significantly differ between HFpEF-STR and HFpEF-controls (Table 1). Brain natriuretic peptide values were slightly elevated but similar between the groups. Most of the blood test data reflecting secondary organ damage (renal, hepatic) showed similar values in HFpEF-STR as compared with HFpEF-controls. Only alanine aminotransferease values resulted significantly higher in HFpEF-STR patients than in HFpEF-controls. Estimated glomerular filtration rate was lower than expected, especially in HFpEF-STR, but did not significantly differ between the two groups (Table 1).

LV geometry and function were similar between the two groups, while RV dilation was non-significantly more frequent in HFpEF-STR than in HFpEF-controls (58 vs. 11%, p = 0.067). As expected, because of the incidence of atrial fibrillation, patients with HFpEF-STR presented larger LA and RA volumes that were, respectively, 43 and 84% larger than in HFpEF-controls (respectively, p < 0.01). No patient with HFpEF-STR presented with leaflet tenting.

Complete hemodynamic data are shown in Table 2 and Supplementary Table 1, while relative changes in hemodynamics from rest to feet on the pedals and between rest and peak are shown in Table 3.

At rest, patients with HFpEF-STR had higher mean pulmonary artery pressure (24 ± 1 mmHg vs. 18 ± 1 mmHg, p = 0.006) and PAWP (17 ± 2 mmHg vs. 11 ± 2 mmHg, p = 0.006) than HFpEF-controls, consistent with mild post-capillary pulmonary hypertension (PH). PH, defined by a mean pulmonary artery pressure >20 mmHg, was present in 83% of HFpEF-STR and in 25% of HFpEF-controls. SV was lower in HFpEF-STR than in HFpEF-controls (28 ± 4 mL vs. 43 ± 4 mL/m², p = 0.03), which was compensated by higher heart rate (81 ± 4 bpm vs. 66 ± 4 bpm, p = 0.022) to maintain cardiac index, which resulted on average at the lower limit of normal in HFpEF-STR (2.2 ± 0.2 L/min/m² vs. 2.8 ± 0.2 L/min/m², p = 0.147). PVR was slightly increased in HFpEF-STR patients but not significantly higher to that of HFpEF-controls (2.1 ± 0.3 WU vs. 1.5 ± 0.3 WU, p = 0.196). RAP was twice as high in HFpEF-STR than in HFpEF-controls (10 ± 1 mmHg vs. 5 ± 1 mmHg, p = 0.007), with tall V waves (13 ± 1 mmHg vs. 6 ± 1 mmHg, p < 0.001) and a high RAP/PAWP ratio (0.64 ± 0.08 vs. 0.45 ± 0.08, p = 0.079).

During exercise (Figure 2, Table 3, and Supplementary Table 1), PAWP increased less in HFpEF-STR patients than in HFpEF-controls. Indeed, despite higher PAWP in HFpEF-STR than in HFpEF-controls at rest, at peak exercise PAWP was not different in the two groups (28 ± 2 vs. 29 ± 2, p = 0.584). RAP V waves displayed an opposite behavior, increasing more in HFpEF-STR patients than in HFpEF-controls (interaction p-value = 0.06), with persistently higher RAP and RAP/PAWP ratio in HFpEF-STR patients than in HFpEF-controls (group p-value < 0.01). Accordingly, LVTMP increased only in HFpEF-controls during exercise (interaction p-value = 0.006), coherent with LV underfilling in HFpEF-STR patients.

Exercise capacity, as evaluated through peak VO₂, was not different at peak exercise between HFpEF-STR and HFpEF-controls, who reached the 71 and 77% of predicted values, respectively (p = 0.446), as shown in Table 1. However, the determinants of VO₂ (Table 2, Supplementary Table 1, and Figure 3) behaved differently in the two groups, with a lower increase in CO and a higher C(a-v)O₂ in HFpEF-STR patients as compared with HFpEF-controls (group p-value < 0.05 for both variables). In particular, HFpEF-STR and HFpEF-controls roughly lay roughly on the same VO₂ isophlets at each step of exercise, with HFpEF-STR rightward and downward shifted (Figure 3). Accordingly, the CO/VO₂ slope, as a measure of CO reserve, was lower in HFpEF-STR patients than in HFpEF-controls (3.9 vs. 5.5, p = 0.024). Exercise hyperventilation (VE/VCO2 slope) did not differ between HFpEF-STR and HFpEF-controls.

The SV was lower in HFpEF-STR than in HFpEF-controls (group p-value = 0.018). In particular, at peak exercise, it resulted 40 mL lower in HFpEF-STR (p = 0.03). The rate of increase of SV in the two groups however did not significantly differ (Table 3). Similarly, PVR similarly decreased by 0.4–0.5 WU in both groups during exercise. SV and PVR were inversely correlated both at rest and peak, with higher PVR at lower SV (Supplementary Figure 1). However, the low R² values (0.12 and 0.19 for rest and peak timepoint, respectively) suggested high heterogenity in the relationship.

This is a small, single-center retrospective study, conducted on a highly selected patients' population. To investigate the effect of STR on HFpEF pathophysiology, we sought to limit potential confounders by applying a 1:1 matching for age, sex, and body mass index, as well as rigorously identifying patients with HFpEF using exercise invasive hemodynamics. The careful and precise hemodynamic evaluation we performed, which could provide significant and physiologically-meaningful results, might at least partially compensate for the small sample size, albeit we cannot exclude the possibility of a type II error. However, we could not control for atrial fibrillation, which was expectedly more frequent in HFpEF-STR than in HFpEF-controls, and that might have contributed to some of the above-mentioned hemodynamic differences between these two population, which may eventually represent two extremes of the HFpEF progression (41). Additionally, the limited sample size of our well-characterized population might have hindered to highlight significant differences in some clinical variables between HFpEF-STR and HFpEF-controls. Due to the retrospective nature of the study, patients had only qualitative assessment of RV geometry and function, while three-dimensional evaluation would have been desirable for a better understanding of this complex chamber. Moreover, the use of exercise-stress echocardiography would have improved our understanding of hemodynamic behaviors, especially in terms of ventricular interdependence (42).